Mechanisms of sand production, prediction–a review and the potential for fiber optic technology and machine learning in monitoring

Abstract

Sand control is an ongoing challenge in numerous hydrocarbon-producing wells in sand-rich reservoirs. Sand production in these wells can cause damage to equipment, reduce production rates, and lead to erosion that can damage subsea equipment, production equipment, well completions, and surface facilities. This problem can compromise the mechanical integrity of the well, resulting in reduced hydrocarbon production and increased operating expenses. This review evaluates various sand production mechanisms, including geological and mechanical production methodologies, and fluid-related aspects, which are thoroughly investigated to offer a thorough understanding of the complexity of the issue and the state of sand prediction approaches. Empirical correlations, numerical simulations, and analytical models are among the sand production prediction techniques critically assessed in this study. The benefits, drawbacks, and suitability of these techniques for various reservoir environments are discussed. Furthermore, the potential benefits of combining Fiber optic (FO) technologies and machine learning (ML) for real-time monitoring and mitigation are highlighted. This integrated strategy has the potential to transform sand control practices of the industry, as demonstrated by case studies and new research that highlights its effectiveness. The future vision outlined in this review includes developments in automation, data processing methods, and sensor technologies, which should improve the precision and dependability of sand production predictions and mitigation. In conclusion, this review paper provides an extensive analysis of the current level of prediction techniques, as well as the mechanisms behind sand production in oil and gas wells. This highlights how real-time, data-driven solutions for monitoring and addressing sand production problems may be provided by FO and ML, which can ultimately lead to safer and more effective hydrocarbon recovery operations.

Introduction

Sand production, commonly referred to as "sanding," is the phenomenon wherein formation sand, along with formation fluids such as gas, oil, and water, is produced as a result of the unconsolidated nature of the formation(Issa et al. 2022). This occurs when the strength of the formation is insufficient to withstand the stress imposed upon it (Mahmud et al. 2020). Most oil and gas reservoirs in the world are contained in weakly consolidated sandstone, and as a result, they tend to produce sand (Rahmati et al. 2013b). In other words, 70% of the world’s hydrocarbon reservoirs are found in formations that are unconsolidated or weakly consolidated formations (Araujo-Guerrero et al. 2022; Ikporo and Sylvester 2015). Hence, unconsolidated or weakly consolidated formations pose major problems in sand production (Hveding and Bukhamsin 2018). Figure 1 shows the three processes that occur in the downhole area, which lead to the three-step process of sand production that is observed on the surface: (a) formation failure, (b) sand erosion due to flow, and (c) sand transport (Mahmud et al. 2020).

Fig. 1

Sand production (sanding) mechanisms modified from Mahmud (Ben Mahmud et al. 2020)

Sand production remains one of the biggest technical challenges for many hydrocarbon-producing wells in sand-bearing reservoirs (Thiruvenkatanathan et al. 2016; Mathis 2007; Carlson et al. 1992). This is especially true for formations with young geological ages (Aborisade 2011; Allahar 2003). Sand or solid production during hydrocarbon production is a critical operational shortcoming that can lead to well collapse (Odigie et al. 2012). Therefore, it is important to determine the types of solids or sand produced to correctly predict the efficient sand control mechanisms (Salahi et al. 2021).

The importance of sand production prediction

Predicting sand production is crucial due to the safety, environmental, and operational concerns that can arise when produced sand particles fill and clog the wellbore, eroding it and the surface equipment (Moore 1994; Zhang et al. 2000). In addition, sand production can damage well completion, subsea equipment, production equipment, and surface facilities by causing erosion, which increases the likelihood of maintaining the mechanical integrity of the well and results in decreased hydrocarbon production and increased operational costs (Saurabh Mishra 2015; Ranjith et al. 2014). Table 1 outlines the issues stemming from sand production in the reservoir, surface equipment, and installations, along with the resultant impacts of this sanding phenomenon.

Table 1 Sand problems from the reservoir to the surface equipment (Ikporo and Sylvester 2015)

Therefore, it is critical to minimize sand production while maximizing reservoir fluid production and maintaining facility integrity because of the negative cost effects of sand production (Peretomode et al. 2022). Hence, a data-driven sand management plan requires real-time sand measurements that provide more insight into how well the asset integrity and sand protection techniques work (Orourke et al. 2020). The adage "prevention is better than cure" perfectly expresses the significance of sand production prediction, monitoring, and surveillance to safeguard well integrity, which helps secure the environment, assets, and reputation without unduly impeding production performance (Shazana et al. 2019). Various methods exist for sand production prediction, broadly categorized into empirical methods based on field observations and well data, laboratory experiments (Sect. "Empirical techniques"), and theoretical modelling techniques encompassing analytical and numerical approaches (Sect. "Theoretical modelling techniques"). Each method offers advantages and limitations, which are discussed in detail in Sect. "Sand production prediction methods", aiming to contribute to ongoing efforts to improve sand production prediction and management.

Objectives of the review

In this extensive review, we intend to explore the complex topic of sand production monitoring during the recovery of hydrocarbons from oil and gas wells. Our main goals included identifying the complex dynamics behind sand generation, assessing the effectiveness of current monitoring methods, and outlining their shortcomings. Moreover, we endeavoured to explore the rapidly developing potential of FO technology as a revolutionary tool for real-time data collection in downhole conditions and sand production rates. In addition, we aimed to evaluate how ML might improve accuracy and enable proactive mitigation methods in predicted sand production modelling.

Within this context, our study aims to offer a venue for the presentation of interesting case studies and effective applications that demonstrate the practical advantages of incorporating FO technology and ML into sand production monitoring landscapes. We examine the complex issues posed by technology integration and provide a review of the financial and environmental effects of better monitoring techniques. As we promote innovation in sensor technologies, data analytics, and ML algorithms, we will also work to uncover new research directions. The adoption of FO technology and ML to revolutionize sand production monitoring and management tactics is the main objective of our review, and it is our primary objective to provide the oil and gas industry with insightful knowledge, practical suggestions, and a forward-looking viewpoint.

Scope of the paper

This comprehensive review will undertake an in-depth examination of sand production mechanisms in hydrocarbon production operations, as well as the evaluation of cutting-edge sand prediction methodologies. The principal objective of this review is to assess the potential of FO technology in real-time sand detection and the identification of problematic areas within hydrocarbon-producing wells. Therefore, recent advancements in sand production monitoring are moving towards a transformative era that employs ML and FO technologies. FO sensors have become vital tools for measuring various parameters, including vibration, strain, and temperature. These measurements can be utilized to predict and identify sand production-related events. Furthermore, ML algorithms play a crucial role in processing and analysing substantial amounts of data generated by FO sensors to create models that predict sand production. Consequently, distributed acoustic sensing (DAS) technology, with its high sensitivity and spatial resolution, offers a unique opportunity to monitor acoustic signals resulting from sand movement within production wells.

ML techniques can be integrated with DAS data to enhance sand production detection and analysis in oil and gas wells. By training ML algorithms with large DAS datasets, patterns and correlations can be identified to accurately identify zones producing sand and estimate the volume of sand produced (Webster et al. 2013). ML algorithms can also be used to predict sand production events based on reservoir parameters, allowing proactive sand management strategies (Fujioka et al. 2019). The combination of DAS and ML provides a comprehensive solution for real-time monitoring and analysis of sand production in oil and gas wells. DAS technology allows for real-time monitoring of sand production without interrupting production rates, while ML techniques enhance data analysis and prediction capabilities. This integrated approach can help operators identify the zones producing sand, estimate the volume of sand being produced, and take proactive measures to mitigate sand production issues. By detecting and monitoring sand production in real-time, operators can optimize production, reduce equipment damage, and ensure the safety and efficiency of oil and gas operations.

Mechanisms of sand production

The challenge of sand production in oil and gas reservoirs has attracted considerable interest in the petroleum industry. As hydrocarbons are extracted from subsurface formations, the accompanying flow of reservoir fluids can mobilize loose sand particles, leading to sand production (Zaitoun et al. 2021). This phenomenon poses a multitude of issues ranging from equipment erosion and damage to reduced well productivity, ultimately affecting the economic viability of oil and gas projects. Understanding the mechanisms underlying sand production, predicting its occurrence, and monitoring its progression are crucial endeavours for reservoir engineers and operators seeking to mitigate its adverse effects.

Wellbore instability and failure of the perforation tunnel in unconsolidated and poorly consolidated reservoirs are the main causes of sand production (Zhang et al. 2000). To shed light on the significance of variables, including the degree of consolidation, reduction in pore pressure over a well's lifetime, increasing water production, increasing production rate, and viscosity of reservoir fluids, we will examine some of the major elements that contribute to sand production in hydrocarbon-producing wells. It is essential to understand these elements to maximize well operations, guarantee the sustainability of hydrocarbon production, and maintain the integrity of the wellbore. Figure 2 shows how sand production occurs because of weak rock strength.

Fig. 2

How sand production occurs because of weak rock strength (Ben Mahmud et al. 2020)

Degree of consolidation

Several factors can contribute to sand production in hydrocarbon-producing wells. One mechanism is the degree of consolidation, which is the mechanical bonding of sand grains. The probability of sand production is decreased by this consolidation process, which increases the compressive strength of the rock (Alakbari et al. 2020). Sanding is more likely to occur in poorly consolidated or unconsolidated formations (Saurabh Mishra 2015). Moreover, Poorly cemented and unconsolidated rocks are substantially weaker (Subbiah et al. 2021). Hence, unconsolidated sandstone reservoirs with permeabilities of 0.5 to 8 Darcies are more likely to produce sand (Carlson et al. 1992).

Reduction in pore pressure

The pore Pressure, also referred to as the formation pressure, is the pressure of the fluids inside a porous rock (Asfha et al. 2019). The decrease in pore pressure during a well's life is another mechanism of sand production; as the well operates, the reservoir's pore pressure drops, which may cause sand grains to become unstable and produce sand (El Mohtar et al. 2014). According to a study by Zhang et al. (2000), The formation of sand grains may eventually separate from the matrix and produce fines along with the well fluid as the reservoir pressure decreases and some of the supporting rock beneath it detaches and experiences increased stress. High-porosity rocks, such as sandstone and chalk, are instances where it is usually possible to observe the pore collapse mechanism. In addition, increasing water production can contribute to sand production. The presence of water can weaken the formation and cause sand grains to detach and be produced along with fluids (Alakbari et al. 2020). Hence, sand particles may move because of drawdown if the sand face is composed of relatively unconsolidated sand. The "threshold sand production pressure" is the pressure at which this happens. The fluid flow will cause migration of sand into the well bore once the "threshold sand production" level is exceeded, and as a result, sand is produced at the surface (Allahar 2003).

Therefore, sand production can occur when the rock deforms because of in situ stresses related to reservoir depletion, which causes effective stresses to surpass the formation strength (Aminu et al. 2019a). Consider a rock formation as a cage that encloses individual sand grains. While the fluids in the pores push outward against the cage walls (pore pressure), the surrounding rocks exert an external pressure that acts as a squeezing force. The strength and ability of the "cage" to hold sand grains in place is determined by the effective stress, which is the difference between these forces (Subbiah et al. 2021). Pore pressure drops when fluids are drawn out of the formation, which raises the effective stress on the rock matrix and weakens the "cage." First, the higher effective stress compresses the matrix, weakening the substance that holds the sand grains. Second, during the production process, fluids move through the formation and exert drag forces on sand grains. These drag forces have the potential to induce shear failure and the generation of sand if the effective stress is not sufficiently high to hold them in situ. This problem is particularly significant in oil and gas wells because the production of hydrocarbons lowers the pore pressure and can cause sand to be produced around the wellbore, endangering the stability of the well and breaking equipment. When fluids are extracted for hydrocarbon production in the context of oil and gas wells, the reservoir pore pressure is lowered. This pressure drop may cause sand to be produced around the wellbore (Rahmati et al. 2013a), which can damage nearby equipment and cause instability. Similar consequences can also result from sudden declines in the water table levels in water wells, especially in unconsolidated formations. These abrupt shifts have the potential to destabilize the formation, which might make well integrity and equipment maintenance more difficult.

Production rate

Sanding, which refers to the production of sand with hydrocarbon fluids, may increase as the rate of hydrocarbon production increases. Sanding can occur for a variety of reasons, such as erosion and mobilization of sand grains caused by higher fluid velocities as a result of increasing production rates (Papamichos et al. 2001). In the vicinity of the wellbore (perforation), a substantial fluid pressure gradient caused by an increase in the well production rate tends to pull the sand (Aborisade 2011; Saurabh Mishra 2015). Every reservoir has a threshold pressure at which a well will produce sand-free water, according to Aroyehun et al. (2018). However, because this threshold pressure is below the economic production rate, engineers frequently disregard it to maximize the output from sandstone reservoirs, which subsequently causes sanding (Mohd Rapor and Darul Ridzuan 2015).

A risky condition, called sanding, may result from higher production rates. Higher fluid velocities result from the fluid (water, oil, or gas) having to pass through the reservoir more quickly as it produces more (Wang et al. 2024). Higher production rates lead to higher fluid velocities, increasing the drag force on sand grains and increasing the risk of erosion. Additionally, turbulence in the flow can further exacerbate the sanding issues. Consider a swiftly flowing river: Similarly, the high speed of the water applies a powerful force to the sand grains, forcing them away from the rock and into the flow. Sanding is particularly problematic near the wellbore, where the pressure gradients are steeper owing to fluid extraction. The sand grains were strongly dragged by this pressure differential, which increased their susceptibility to erosion and fluid movement. In other words, sanding problems might arise from quicker fluid flow brought on by greater production rates, particularly close to the wellbore when the pressure drops dramatically.

Increase water production

Sand production during hydrocarbon production can be influenced by an increase in water production. Moreover, Increased water production in a well may cause the formation to become unstable, leading to the production of sand along with the fluids (Andrews et al. 2005). The presence of water can weaken the formation, causing sand grains to detach and be produced along with hydrocarbon fluids. Hence, sand production increases with an increase in water cut, or as water production begins, as does sand production (Aborisade 2011). In contrast, Wu et al. (2006), Sandstone's mineralogical makeup and its level of residual water saturation both influence sand production. Therefore, sandstones with a high clay content and low residual water saturation have the most impact; otherwise, clean sandstones or those with a high residual water saturation have less impact. However, Clay minerals are found in many reservoir rocks and serve as natural adhesives between sand grains. These clays may interact with water and swell or weaken, particularly if the water's chemical makeup is incompatible with the formation of brine (Wu et al. 2006). This weakens the overall integrity of the formation, making it more susceptible to erosion and the dislodging of sand grains. The produced water and formation brine must be compatible. Therefore, incompatible brines can cause clay swelling and weaken the rock matrix. Sanding is more likely to occur in unconsolidated or poorly cemented formations with weak sand-grain bonding. Rock mineralogy, grain size distribution, and clay concentration are additional factors. In particular, formations with high clay concentration and minimal residual water saturation (Liu 2023).

Reservoir fluid viscosity

Sand production is significantly influenced by the viscosity of the reservoir fluids. The resistance of the fluid to flow is referred to as its viscosity. In comparison to fluids with lower viscosities, such as water, those with higher viscosities are "thicker" and more resistant to flow (Verst et al. 2022). The viscosity of the reservoir fluid may affect the amount of sand produced. Sanding is more likely to occur when fluids with a higher viscosity because they often transport more sand grains (Quiñones-Cisneros et al. 2004). Furthermore, the drag force and reservoir fluid viscosity are directly correlated because a more viscous fluid with more drag force produces more sand movement toward the wellbore, and high-viscosity reservoir fluid has a stronger ability to drag the sand particles than low-viscosity fluid (Atashi et al. 2018).

• Relationship between viscosity and sand transport When two fluids have the same flow velocity but different viscosities, the higher-viscosity fluids drag on the sand grains. This is because there is a greater drag on the grains owing to the higher resistance to movement provided by high-viscosity fluids (Song et al. 2017). Stokes' law is a well-known equation in fluid mechanics that describes the drag force (Fd) and is related to particle size (expressed by its radius, r), fluid velocity (v), and fluid viscosity (μ). Hence, the formula is Fd = 6πηrv. There is a direct correlation between the drag force and the viscosity (μ).

• Impact on sand mobilization Sand grains that are loosely cemented may become dislodged in weak formations due to the drag force of the flowing stream. Nonetheless, the viscosity of the fluid also affects its capacity to move these dislodged grains. Sand grains are more effectively suspended and carried by higher-viscosity fluids than by lower-viscosity fluids (Mishra and Ojha 2016). Higher viscosity fluids can create a more cohesive force around the sand grains, allowing them to be carried within the flow more readily.

Sand production prediction methods

The significant difficulty of sand production during the production of hydrocarbons from wells must be addressed for efficient well management and sand control. Predicting sand production is essential in the early stages of field development planning for well-completion planning and later for production management (Gharagheizi et al. 2017a). There are various approaches to predicting sand production. Production data, well logs, laboratory tests, acoustic, intrusive sand monitoring devices, and analogies are some of the techniques used (Khamehchi and Reisi 2015). Therefore, the currently available sand prediction methods are either based on in-field sand production measurements, laboratory sand production trials, theoretical sand production modelling, or the latest distributed acoustic sensing (DAS) seismic. Figure 3 is a diagram showing sand production prediction methods.

Fig. 3

Sand production prediction methods

Empirical techniques

In contrast to solely theoretical or abstract ideas, empirical procedures relate to methods or approaches that are based on observation, real-world data, and practical experience. Empirical methods have been used to assess reservoirs' capacity for producing sand-free hydrocarbons (Ajayi et al. 2022). The likelihood of sand influx can be estimated using elastic factors like shear modulus, bulk compressibility, and the shear modulus to bulk compressibility ratio. The sanding potential of reservoirs may be evaluated using these factors, which can also aid with sand production prediction.

Additionally, investigations have concentrated on evaluating the pore structure of sand-conglomerate reservoirs (Zhou et al. 2018). Understanding the capacity of these reservoirs to produce sand requires a quantitative analysis of the pore structure. However, there are currently no reliable methods for analysing the pore structure of tight sand-conglomerate reservoirs. Overall, these studies show how difficult it is to forecast and control sand production rates during hydrocarbon production without considering a variety of variables, including rock characteristics, fluid flow, and stress distribution. The suggested models and experimental results aid in the development of successful sand control approaches and well-management procedures. To predict sand production rates and comprehend the underlying mechanisms, several techniques have been suggested.

Concisely, during the production of hydrocarbons, sand production is predicted using a variety of methods. These include choosing the best sand filters, using AI approaches, developing prediction models, using empirical methodologies, and assessing reservoir pore structures. These methods can offer insightful information about the possibility of producing sand and provide recommendations for sand management measures by merging field observations and well data.

Field observations and well data

This method depends on establishing a correlation between well data from sand production and field operation parameters (Mohd Rapor and Darul Ridzuan 2015). One approach is to select optimal sand screens that effectively retain sand while maximizing hydrocarbon production (Ahad et al. 2020). The sand-retaining precision of screens plays a significant role in sand control and production rates. By choosing the appropriate sand screen aperture and evaluating the limitations during sand retention tests, sand production can be predicted and managed effectively. Table 2 below shows the parameters affecting sand production associated with formation rock, reservoir properties, completion type, and production.

Table 2 Parameters affecting sand production (Veeken et al. 1991)

The possibility of producing sand is influenced by several variables in four categories: formation rock, reservoir, completion, and production. It is essential to comprehend these characteristics and their implications to design production strategies that reduce this risk. The properties of the formation rock are important. The rock's inherent strength, which is impacted by cementation and grain size, defines how well it can withstand forces that could cause it to disintegrate and produce sand. In-situ stresses, the natural forces pushing on the rock from all directions, can change as fluids are extracted, potentially weakening the rock structure (Askaripour et al. 2022). Depth also plays a part, as deeper formations experience greater pressure that can compact and weaken the rock. Reservoir characteristics like pore pressure, and the pressure of fluids within the rock pores, are equally important. Declining pore pressure during production reduces the force counteracting the in-situ stresses, potentially leading to sand production (Subbiah et al. 2021). Permeability, a measure of how easily fluids flow through the rock, can also be an indicator. Higher permeability often suggests weaker rock with larger pore spaces, making it more susceptible to sand movement (Tiab and Donaldson 2012). The type of fluids in the reservoir, particularly the presence of water, can further weaken the binding forces between sand grains and increase the risk of sand production (Mahmud et al. 2020).

Beyond formation properties, completion design and production practices significantly impact sand production management. Wellbore design considerations include factors like orientation, diameter, and completion type (open hole vs perforated) (Renpu 2011). These elements influence how stresses are distributed around the wellbore and the potential for sand production. Sand control methods, such as screens, gravel packs, and chemical consolidation, are crucial techniques used to prevent sand from entering the wellbore (Abduljabbar et al. 2024). Production parameters like flow rate, drawdown, and flow velocity all exert forces on the formation. Higher values can increase the risk of dislodging sand grains (Nemati et al. 2024). Careful management of production operations, including bean-up/shut-in procedures, artificial lift selection, and consideration of reservoir depletion, is essential to minimize stress on the formation and prevent sanding (Geilikman et al. 2005). Three methods: one parameter, two parameters, and multi-parameters can be used to measure the effects of these parameters. For instance, for one parameter the prediction tool uses cut-off depth criteria (Sulaimon and Teng 2020).

Laboratory experiments

Experimental studies are one method for learning about sand production (Lu et al. 2018), conducted an experimental investigation on the generation of sand during the exploitation of hydrates and discovered that sand production is a serious issue that compromises the safe and effective extraction of hydrates. This method is also frequently used to establish a relationship between the likelihood of sanding and observable factors like stress, flow rate, and rock strength and to gain an understanding of the mechanism of sanding in the formation in question (van den Hoek et al. 2000). Hence, observing and simulating the production of sand in a controlled environment helps in the development of an understanding of the mechanics behind sand production as well as the impact of various field and operational parameters on sand production (Veeken et al. 1991). Both unconsolidated sand and friable-consolidated sandstone were used in the study. Table 3 shows the factors causing sand production in different types of formations.

Table 3 Factors causing sand production in different types of formations (Mohd Rapor and Darul Ridzuan 2015)

In unconsolidated sand, where grains are loosely packed, sand production is primarily driven by flow rate and capillary forces. High flow rates can overpower the weak attraction between grains, causing them to be dislodged and produced along with the fluids. For friable-consolidated sandstone, the picture is more complex. Here, the rock offers some resistance, but the key factor is the stress acting on the formation around the wellbore. As fluids flow, they can erode the rock, creating a cavity. The size of this cavity is influenced by the flow rate. If the flow rate is high enough, the cavity becomes unstable and collapses, triggering sand production. Interestingly, research suggests there's a critical flow rate (around 5–10 barrels per day) at which this collapse occurs (Mohd Rapor and Darul Ridzuan 2015), regardless of factors like the specific sand mixture, existing cavity size, or even the pressure conditions within the formation.

• Role of laboratory experiments in sand control

Laboratory experiments play a vital role in developing effective sand control strategies and well management practices. These controlled environments allow researchers to simulate downhole conditions, including pressure, stress, and fluid flow dynamics. This controlled setting provides a safe and cost-effective platform to assess the stability of rock formations and their susceptibility to sand production (Chin et al. 2023). Through simulated conditions, various sand control methods like screens, gravel packs, and chemical consolidation can be rigorously evaluated in the lab. These experiments analyse the effectiveness of each method in preventing sand movement under varying loads, mimicking real-world wellbore scenarios. The insights gained from these experiments directly translate to improved well management. By understanding the mechanisms of sand production, engineers can design production parameters like flow rates and drawdowns to minimize formation stress and reduce the risk of sand production. Additionally, lab data can be used to develop sand production prediction models. These models assist in real-time monitoring and proactive well management adjustments, ultimately preventing sand production issues (Gharagheizi et al. 2017b).

Theoretical modelling techniques

Analytical and numerical methods are the cornerstones of predicting sand production during hydrocarbon production. These methodologies aim to not only identify the conditions that trigger sand production but also provide valuable insights for effective sand control and well management strategies. By employing both analytical and numerical methods, engineers can gain valuable insights into the factors influencing sand production and make informed decisions regarding sand control measures and well management practices (Rahmati et al. 2013a). The subsequent sections will delve deeper into analytical and numerical methods.

Analytical modelling

Due to its straightforward computation, easily implementable calculations, and convenience in running many realizations to evaluate various scenarios, this method has grown in popularity in the petroleum industry (Aborisade 2011). This approach requires mathematical formulation of sand failure mechanisms such as compressive failure, tensile failure, and erosion (Veeken et al. 1991). For confined sandstone oil reserves (Guo et al. 2021), suggested a model for predicting sand production. The model considers both the drag force brought on by fluid flow and the strength of the binding between sand particles. Sand is produced when the binding force is less than the drag force. This model highlights how crucial it is to comprehend the hazards and elements involved in sand production in tight sandstone oil reservoirs. It allows for the application of sand control measures and efficient well management by forecasting sand production rates.

Analytical methods, such as the model proposed by (Li et al.2021 provide insights into the fundamental factors influencing sand production and offer a simplified approach to predict sand production rates based on key parameters such as bond strength and fluid drag force. These analytical models can be useful for initial assessments and quick estimations. Below, in Fig. 4, the mechanisms of sand failure are illustrated, encompassing far-field stress, drawdown-induced failure, tensile failure induced by drawdown, and erosion induced by flow.

Fig. 4

Sand failure mechanisms adapted (Mohd Rapor and Darul Ridzuan 2015)

The mechanical failure of a rock (reservoir) depends on several elements. They include (a) the rock's strength (UCS), (b) the mean-effective stress acting on the rock, (c) the stress distribution near the wellbore, and (d) the drawdown caused by excavation (drilling) and flow rate (production) (Subbiah et al. 2021).

• Pore structure and sand control strategies

The analysis of pore structure in sand-conglomerate reservoirs offers valuable insights for sand production prediction and control strategies. Pore characteristics like size distribution, connectivity, and tortuosity influence sand grain mobility and fluid flow paths within the reservoir (Ecay et al. 2020). Understanding these features helps predict the likelihood of sand grain dislodgement and identify areas susceptible to higher flow velocities that could increase sand production risk. Knowledge of pore structure informs the selection of appropriate sand control measures. For instance, large, poorly connected pores might necessitate stronger methods like gravel packs or screens, while reservoirs with well-connected smaller pores might be managed through production rate optimization or specific well-completion techniques.

Numerical modelling

The amount of oil and gas produced depends heavily on the production process's ability to predict unconsolidated sandstone sand production accurately (Sun et al. 2023). Numerical approaches have been the most significant approach to date for predicting the production of sand. These techniques can function more effectively with the support of the experimental findings and analytical correlations. The numerical models do have limitations too (Shahsavari et al. 2021). Numerical simulation techniques can be classified into four categories: discrete element method, discrete difference method, and discrete element-finite element hybrid method.

• A. Finite element Method

The computation of thick wall cylinders, equivalent plastic strain analysis, critical drawdown pressure difference evaluation, and sand production prediction are the primary applications of the finite element approach (Garolera et al. 2020). To examine sand generation in brittle and compacted sandstones, Papamichos et al. (2001) conducted experiments and built a volumetric sand production model. They discovered that the generation of sand is related to the plastic nature and decohesion of a zone around the wellbore, which can then be mobilized by the fluid flow's hydrodynamic forces.

It was observed that when external stress was introduced in addition to the fluid flow rate, the rate of sand generation increased. The importance of characteristics such as permeability, porosity, solids density, and sand production coefficient in predicting sand production was further underlined by the study. Moreover, Li et al.'s (2020a, b) research concentrated on exploiting seepage and in situ stress distribution to predict sand production in tight sandstone oil reservoirs. The model takes the stress distribution close to the well and the formation fluid seepage law into consideration when calculating the rate of sand production. To properly manage wells and control sand, the authors underlined the significance of bond strength in influencing sand erosion and the need for a model that anticipates sand production. Field data from the Daqing oilfield was used to validate the model, yielding a relative error of 4.38% furthermore, Li et al. (2020a, b) examined the production of sand in shale gas wells at different stages of production in a distinct study. They found that the roundness of the sand worsens with decreasing sand size and that the sand's composition is comparable to that of fracturing proppant and shale minerals. Sand is produced because of both proppant failure and shale fracture, as shown by the organic chemicals that are adhered to the surface of the sand. This study provides insights into the source of produced sand and theoretical guidance for understanding the sand production mechanism.

• Drawdown pressure difference

This method is validated by finite element software and simulates the single-hole sanding behavior under genuine triaxial stress and fluid flow conditions using the true triaxial stress chamber (TTSC) (Younessi et al. 2013). Research indicates that in the sand-producing zone surrounding the borehole, a critical drawdown pressure differential is required to initiate sand production (Song et al. 2022; Sun et al. 2023).

• Equivalent plastic strain analysis

Equivalent plastic strain analysis is a method used to analyze sand production conditions and improve the accuracy of sand production prediction (Muller et al. 2011; Sasaki et al. 2021). It involves introducing the equivalent plastic strain to simulate the failure near the wellbore during production. The failure threshold is determined according to the critical equivalent plastic strain to predict sand production. This method has been used since the 1990s and is more accurate than other methods such as the Drucker-Prager method (Pradel et al. 2011). It has been used in conjunction with other numerical simulation methods such as the finite difference method and the discrete element-finite element hybrid method to predict sand production.

• Thick wall cylindrical calculations

Thick wall cylindrical calculations are a method used to optimize the matching between laboratory tests and finite element model simulation results. This method involves testing the experimental curve of a thick-walled cylinder (TWC) to calibrate the strength and plasticity of the material. When the numerical test values match the experimental results, the failure threshold is determined according to the critical equivalent plastic strain to simulate the failure near the wellbore during production (Deng et al. 2019). This method has been proposed as a workflow for sand production prediction and has been shown to have high accuracy with relatively simple calculations. It has been used in conjunction with other numerical simulation methods such as the finite element method, finite difference method, and discrete element-finite element hybrid method to predict sand production.

• B. Finite difference method

This method is grounded in the erosion criterion proposed by (Vardoulakis et al. 1996) and employs a finite difference model to investigate the onset and rate of sand production. By simulating the fluid flow phenomenon and observing the sand erosion process, the numerical model predicts the amount of sand in the wellbore. (Rahmati et al. 2013a) focused on validating the predicted cumulative sand and sand rate against physical-model tests, emphasizing the significant costs associated with sand production in the oil industry and the benefits of predicting and managing sand production rates. The numerical model developed captures the erosional mechanics involved in sand production, accounting for the coupling between fluid flow and mechanical deformation. The model predicts the sanding rate, which increases with higher flow rates, and provides accurate predictions of external deformations observed in experiments. Numerical methods, such as the one developed by (Rahmati et al. 2012), offer a more comprehensive understanding of sand production mechanisms by considering the complex interactions between fluid flow and mechanical deformation. These numerical models simulate the erosion process and provide detailed predictions of sand production rates and deformations. They can be valuable for more accurate and detailed sand production predictions, especially when dealing with complex reservoir conditions.

• C. Discrete elements methods

A computer program that combines the Discrete Element Method (DEM) with the Lattice-Boltzmann Method (LBM) to simulate the production of sand has been developed. This program successfully replicated the various stages of sand production observed in trials. Researchers found that the tool was effective in investigating the parametric effects on sand formation, such as the influence of fluid flow and stress levels. Honari et al (2021), developed a computer program that simulates sand production by combining the Discrete Element Method (DEM) with the Lattice-Boltzmann Method (LBM). Their simulations accurately represented the various phases of sand formation seen in trials. The software has shown to be a useful tool for researching the parametric impacts on sand production, including the effects of stress level and fluid flow.

A numerical approach was also used by Wang et al. (2019a, b, c), to model the production of sand. They employed a mesoscopic bound particle lattice Boltzmann approach to simulate the fluid–solid interaction at the pore/grain level. By cutting the connections between the particles that mimicked cementation, they were able to precisely represent the process of transient particle erosion. According to their findings, sand generation occurs when the tensile failure area develops close to the wellbore cavity's edge during decline.

• D. Discrete element-finite element hybrid method

This method combines the continuum theory to simulate small deformations away from the wellbore with the discontinuous characteristics of discrete elements to analyse sand production behaviour near the wellbore (Wu and Choi 2012). This can further improve the accuracy of sand production prediction. Overall, the Discrete Element-Finite Element Hybrid Method is a valuable numerical simulation method for predicting sand production in unconsolidated sandstone reservoirs, but more research is needed to fully understand its potential and limitations.

Overall, these investigations shed light on the processes that lead to sand production during the production of hydrocarbons. Experimental studies demonstrate the gravity of the issue and the requirement for efficient prevention measures. The mesoscopic bonded particle lattice Boltzmann method or the combination of DEM and LBM are two examples of numerical modelling approaches that provide useful tools for investigating intricate fluid–solid interactions and comprehending the variables affecting sand production. Sand production is influenced by several factors, such as the degree of consolidation, decrease in pore pressure, increase in water production, production rate, and reservoir fluid viscosity. Below is a summary of the description, advantages, and limitations of all existing methods for predicting sand production, presented in Table 4.

Table 4 Sand production prediction methods, advantages, and limitations

Sand production monitoring

Sand production is a significant challenge in hydrocarbon production from oil and gas wells, particularly in unconsolidated sandstone reservoirs (Ahad et al. 2020). Sand particles can separate from the reservoir rock and be transported by hydrocarbons to the well, causing erosion and corrosion of downhole and surface equipment. Hence, this can lead to production interruptions and the need for costly repairs (Aminu et al. 2019a). Therefore, effective monitoring of sand production is crucial for sustainable hydrocarbon production and to minimize operational risks and costs.

There are various methods and techniques for monitoring sand production in oil and gas wells. One approach is to use vibration sensors to detect the vibration response characteristics of sand-carrying fluid flow impinging on the pipe wall (Li et al. 2021). This method provides real-time monitoring of sand production and can help ensure the safety and long-term production of oil wells. Another approach is the use of computational intelligence-assisted design frameworks for real-time quantitative monitoring of sand in gas flowlines (Aminu et al. 2019a). This method utilizes artificial neural networks to optimize the design of sand monitoring systems and improve the accuracy of sand detection.

Another method for monitoring sand production is using vibration sensors (Wang et al. 2015). This approach involves the application of time–frequency analysis, characteristic sand frequency band filter method, and peak searching-denoising method to enhance the detection ability of sand vibration signals in the presence of background noise. By comparing the time–frequency domain and power spectrum of the vibration signals, it is possible to distinguish between sand and non-sand production wells.

Sand management methodologies are also important for sustained hydrocarbon production in the presence of sand or proppant particles in well fluids (Rawlins 2013). These methodologies aim to minimize the impact of produced solids on surface equipment and ensure the operability of surface facilities. By controlling the particle size and concentration of formation sand or proppant, sand management strategies can optimize production and reduce the need for costly interventions.

Traditional sand production monitoring methods

Operators in the oil and gas industry have been extremely concerned about traditional sand production monitoring methods because of the possibility of mechanical equipment wear, safety hazards, and negative effects on hydrocarbon production rates (Aminu et al. 2019a). Minimizing operational risks and expenses and promoting sustainable hydrocarbon production depends on efficient sand monitoring. The study of vibration in response to weak shocks is a traditional technique for evaluating sand production. This method entails keeping an eye on the features of the fluid flow-carrying sand that impinges on the pipe wall and its vibration response. Sand production in offshore oil wells can be detected and tracked by examining the vibration signals. (Li et al. 2021). Another vibration sensor approach has been used to monitor sand production in Bohai Bay, demonstrating the feasibility of real-time monitoring using a special broadband sensor (Wang et al. 2015).

While there has been use of traditional methods for monitoring sand production, it is crucial to remember that these approaches may have drawbacks and difficulties. For instance, water chemistry and mineral properties may have an impact on how well a process works at lower temperatures (Harjai et al. 2012). The escalation of oil sand extraction can lead to extensive environmental impacts (Rosa et al. 2017). Challenges in sand production management during marine natural gas hydrate exploitation and the need for innovative solutions have also been identified (Wu et al. 2021).

To summarise, the oil and gas sector has long used vibration analysis, temperature-based analysis, and the examination of variables affecting sand production during various phases of production as a means of monitoring sand production. These techniques offer insightful information about the production of sand and can assist operators in reducing risks and optimizing workflows.

Surface acoustic sand detectors (ASD)

The acoustic energy propagated when a particle strikes the inside of a pipe at a bend will be detected by an acoustic detector placed on the outside of a pipe (Haugsdal 2017). A moving particle's kinetic energy is partially transformed into thermal energy and partially into acoustic energy when it collides with a stationary object (Foster and Linville 1979). As a result, utilizing commercially available acoustic sand monitors that clamp to the outside of the pipe wall and purportedly detect sounds from a sand collision with the pipe wall is one of the appealing approaches for sand detection (McLaury and Shirazi 2017).

A sand detector is a device that can be used to determine the amount of sand present in a flowing stream traveling across a certain location. Acoustic sand detectors come in both intrusive and non-intrusive varieties. The non-intrusive method involves 'listening' to the flow line for sand-related impact noise (Fig. 5), whereas the intrusive type of detector detects sand by measuring the degrees of erosion on a probe introduced into the flow stream (Allahar 2003).

Fig. 5

Non-intrusive acoustic probe (Allahar 2003)

Limitations of ASD

Detecting the presence of sand in oil and gas production wells is a critical aspect of preventing equipment damage and maintaining uninterrupted operations. However, the current methods employed for this purpose have limitations, relying primarily on stationary acoustic sensors positioned on the surface of the production flowline. These sensors have several shortcomings, as they face challenges in identifying the precise location where sand enters the flowline and provide information that is delayed due to their design, which is affixed to the surface. furthermore, It may be challenging to distinguish between real sand signals and false alarms due to ASD's calibration (Khan and Okotete 2015). Two wells, Well-X1 and Well-X2, with confirmed sand occurrences are shown in Figs. 6 and 7 below. Sand signals have occasionally been seen in Well-X1 (Fig. 6), a Frac Pack (FP) completion well, primarily in conjunction with an increase in water cut. Beginning in June 2013, there was a noticeable increase in the frequency of sand signals. Sand indications start to develop if the liquid rate surpasses 15,000 bpd, according to the Step Rate Test. Since June 2013, the liquid rate has been progressively decreased from 18,000 to 14,000 bpd to minimize sand signals.

Fig. 6

Well-X1 Confirmed Sand Event (Khan and Okotete 2015)

Fig. 7

Well-X2 confirmed sand event (Khan and Okotete 2015)

However, during the first ramp-up in October 2012, big sand signals were detected in Well-X2 (Fig. 7), an Expandable Sand Screen (ESS) completion well. The Step Rate Test verified that the appearance of sand signals occurs when the liquid rate surpasses 6,000 bpd. One noteworthy finding was that, despite the sand detector showing enormous signals, no sand production was seen on topside equipment. Just before the Full Field Shut Down (FFSD), Well-X2 was ramped up, and the Acoustic Sand Detector picked up enormous sand signals during the well's initial ramp-up. Thus, it is believed that Well-X2 is the source of all the sand retrieved during the loop pigging. After conducting comprehensive tests on Well-X2, it was determined that there was a possibility of ESS compromise during the well's completion operation. Therefore, to establish that sand signals are indeed caused by real sand events, the ASD signals must be supported by further data. To provide real-time sand monitoring based on the raw signal and to historicize raw signal data, ASD raw signal was integrated with the real-time historian (Khan and Okotete 2015).

Recommendations

To overcome these constraints, a novel approach incorporating DAS technology and ML has been proposed. DAS employs a FO cable that is installed throughout the entire length of the well. This method presents several substantial benefits. For one, DAS supplies real-time data, making it possible to immediately detect sand production events. Secondly, by examining the extracted seismic features, the technique can precisely pinpoint the exact location of sand entry within the wellbore. Furthermore, analysts can quantify the frequency and relative duration of sand production events by studying these features. Gaining insight into sand production patterns is vital for optimizing well production and preventing damage. In conclusion, DAS constitutes a significant advancement in sand detection capabilities compared to traditional methods. It enables real-time, precise detection and, most importantly, the quantification of sand being produced. In Sects. "DAS and DTS technology for sand production monitoring" and "ML techniques for sand production monitoring", we will delve into the potential of FO technology and ML in sand production monitoring during hydrocarbon production.

DAS and DTS technology for sand production monitoring

Real-time sand production monitoring could be revolutionized by DAS, distributed temperature sensing (DTS) technology, and ML. A common application of DAS technology is the structural health monitoring of civil infrastructure (Ye et al. 2014). It has been employed in monitoring the safety conditions of high-rise structures during construction and in-service stages. FO technologies have been developed for intervention applications in the oil and gas industry (Gardner et al. 2015). They can be used in vertical and highly deviated well sections, providing important information about reservoir dynamics. Furthermore, in wells completed with open-hole gravel, sand production has been detected using FO-distributed vibration sensing (Mullens et al. 2010). High-risk interventions are avoided when operators can determine the depth and interval of sand entrance by analysing the vibration profiles on FO. Long-term reservoir monitoring at the sand face is made possible by an FO support device, which enables high-resolution early detection of subsurface movements (Earles et al. 2010). Operators can use this information to schedule early preventative measures and track the results of those activities in real-time. Furthermore, Monitoring well integrity, evaluating injection and production flow profiles, analysing cross flow behind casing, and other downhole events may all be accomplished with FO technology (Al-Qasim et al. 2021). Continuous and real-time measurements over an FO cable's whole length are possible with the use of distributed FO sensors (Ashry et al. 2022). In addition, to monitoring pipelines that carry hydrocarbons over great distances, these sensors are extensively used in the oil and gas sector and can yield important data at any point in a well's life cycle.

The technique monitors the production process of oil and gas wells, the process and effects of hydraulic fracturing, and the oil and gas transmission pipelines by using the FO itself as a sensor to detect the temperature, sound waves, and strain data of each tiny layer underground (Yao et al. 2023). This makes it possible to track the production of sand in gas and oil wells with greater accuracy and dependability. Furthermore, since FO technology does not involve the insertion of instruments into the well, it is a non-invasive monitoring method that lowers the possibility of well damage and the need for expensive repairs. All things considered, FO technology is a useful instrument for tracking the production of sand in oil and gas wells.

Discrete sensing technology, distributed temperature systems (DTS), and DAS are the three primary FO monitoring technologies (Chen et al. 2016). Pressure and temperature measurements taken at the desired depth are used in discrete sensing technology. DTS and DAS are examples of distributed sensing technologies that measure temperature and acoustic data dispersed throughout the wellbore. None of the three systems are susceptible to electromagnetic interference and don't require any downhole electrical supply. Monitoring of sand production has also been done with vibration sensors. In offshore oil production, real-time sand production monitoring is essential because excessive sand production can result in sand burial in the pay zone, wear on downhole and surface equipment, and a decrease in production (Wang et al. 2015). Effective sand production monitoring systems may consistently and successfully gather real-time data on sand production in oil wells by examining the vibration response to slight shocks (Li et al. 2021). FO technology can be utilized for well-integrity monitoring during hydrocarbon production, in addition to monitoring sand production. Fluid migration may result from cement seals deteriorating over time; this can be observed with DAS monitoring (Raab et al. 2019). This technology allows for real-time monitoring of well integrity and can help ensure the sustainable operation of subsurface reservoirs for hydrocarbon production. Figure 8 shows a Schematic description of FO data acquisition and processing. FO data collection and processing is a technology that uses optical Fibers, which are thin, flexible, and effective at transmitting data over great distances—for data transmission and processing. The main elements and procedures involved in FO data collection and processing are shown in the following schematic diagram. Source data, optical sensing cable, distributed sensors across the entire length of the FO cable, processing unit, and data output (display).

Fig. 8

Schematic description of FO data acquisition and processing (Al Hashemi 2022)

DAS technology

DAS is a technology that records sound and vibration signals along a FO cable (Harris 2017). It records dense seismic data for tens of kilometres with high spatial and temporal resolution (Lellouch and Biondi 2021). Therefore, it is advantageous for high-resolution, continuous, and real-time measurements (Li et al. 2022). DAS technology is a strain-sensing technology that utilizes the Rayleigh scattering of laser pulses in FO cables (Kuvshinov 2016). It provides a cost-effective alternative to conventional seismic acquisition systems and has gained increasing attention in seismic recording. It is based on the principle of continuously monitoring external acoustic or vibration signals detected by the sensing fiber in real-time (Wang et al. 2019a, b, c). DAS systems employ various sensing technologies such as interference sensing, optical backscattering, optical coupling detection, and optical nonlinear parameter detection.

The working principle of DAS

The DAS system employs a method known as coherent optical time domain reflectometry (C-OTDR), which entails the observation of very low levels of a backscattered signal by transmitting successively brief pulses of highly coherent light down an optical fiber (Hill 2015). Furthermore, a small amount of the light that is naturally scattered during the passage of an optical pulse through the fiber and back to the sensor unit is called Rayleigh scattering (Soroush et al. 2022a). One can calculate the amount of dispersed light produced throughout the fiber by measuring the returning signal against time (Johannessen et al. 2012). DAS has applications outside of conventional sectors. It has been investigated in tele seismic research, showcasing its usefulness in receiver function computation and tele seismic waveform analysis (Yu et al. 2019). Using the Rayleigh scattering of laser pulses in fiber optic cables, DAS technology is a strain-sensing method. It provides an affordable substitute for traditional seismic acquisition systems and finds use in several industries, including data mining, tele seismic research, geophysics, geothermal energy, and the oil and gas sector. DAS systems detect external vibration or acoustic signals in real-time using a variety of sensing technologies. Figure 9 shows a DAS system consisting of an optical fiber that is connected to an interrogator. The interrogator sends a pulse of light down the fiber, and the light is scattered back by acoustic waves that are present in the fiber. The interrogator measures the time it takes for the light to travel back and forth and then uses this information to calculate the location and strength of the acoustic waves.

Fig. 9

Schematic diagram of DAS measurement using the OTDR principle to detect the seismic wave incident

DAS is an acoustic detection technology that has shown promise in various applications, including production and geophysical settings (Webster et al. 2013). DAS utilizes FO cables to detect acoustic signals along the entire length of the cable, allowing for continuous monitoring of the wellbore (Hornman 2017). Hence, by deploying DAS systems in oil and gas wells, operators can detect and locate sand production events in real-time, enabling proactive measures to mitigate sand production issues. This technology provides a cost-effective, low-risk, and efficient solution for monitoring sand production without interrupting production rates in oil and gas wells. As a result, it is an attractive option for monitoring sand production.

DAS has emerged as a promising technology for monitoring and analysing various environmental dynamics. DAS utilizes standard FO cables with attached laser interrogator units to measure vibrations and strain rates along the entire length of the fiber, effectively transforming it into an array of pseudo-seismometers (Zhu et al. 2021). This technology has been successfully applied in geophysical applications, such as seismic monitoring for oil and gas exploration, CO2 sequestration, and landslide detection.

The use of DAS in seismic monitoring programs has been demonstrated in various projects, including the CO2CRC Otway Project, where a buried 3D DAS array and a fiber-optic cable on production tubing were utilized (Correa et al. 2017). The Otway installation offered comprehensive seismic data coverage, making it an ideal environment for testing the performance of DAS as a monitoring tool.

DAS technology has also been used in a variety of fields. DAS has been applied in geophysics to monitor fracture hydromechanical reactions and to quantify downhole seismic and acoustic signals (Becker et al. 2017). Additionally, it has been applied to distributed sensing of teleseisms and microseisms, yielding insightful information on seismology (Williams et al. 2019). DAS has been applied to high-temperature Carbon Capture, Utilisation, and Storage (CCUS) projects in the geothermal energy area to monitor seismicity (Stork et al. 2020).

The oil and gas sector has also made use of DAS technology. With benefits including high sample density and resistance to severe conditions, it has been employed to gather vertical seismic profile data (Li et al. 2023). In wellbore flow interpretation, DAS has shown reliability and accuracy in determining production contributions and fluid phase changes (Wu et al. 2022). Additionally, DAS has been applied in pipeline monitoring and management for the oil and gas industry (Wilson et al. 2019). Figure 10 below shows on the left the Fiber is installed outside tubing while on the right the Fiber is installed outside casing. Hence, applications for distributed Fiber optic sensing (DFOS) go beyond well monitoring for production and reservoir optimization; they also include well integrity risk identification and other well completion failure detection (Guo et al. 2021). Because it converts the Fiber optic cable into a dispersed array of acoustic sensors, distributed acoustic sensing (DAS) is a rapidly developing Fiber optic technology with several applications for wellbore monitoring and geophysical surveillance (Sulaimon and Teng 2020). Furthermore, it allows continuous, real-time, and permanent monitoring over the full length of the Fiber line (Rahmati et al. 2012). In‐well pressure, temperature, flow, and casing or completion deformation can be monitored through a permanent Fiber optic installation in a well using time-lapse vertical seismic profiles (VSP) (Lellouch and Biondi 2021).

Fig. 10

Different types of DAS installation. Elements of the well are marked, and the Fiber is yellow (Lellouch and Biondi 2021)

Case studies of DAS for sand detection

The petroleum industry has become interested in adopting DAS technology to monitor sand production during hydrocarbon production from oil and gas wells (Daley et al. 2016). DAS allows seismic data acquisition without discrete sensors, making it a promising tool for wellbore monitoring. By deploying FO cables, DAS can provide real-time monitoring of sand production in oil and gas wells. One study conducted field testing of modular borehole monitoring using simultaneous DAS and geophone vertical seismic profiles (Daley et al. 2016). The research demonstrated the feasibility of using DAS for monitoring sand production in wells. The study highlighted the advantages of DAS, such as its ability to provide continuous and high-resolution data, which can aid in detecting sand production events and assessing their impact on wellbore integrity.

Overall, FO technology offers a range of applications for sand production monitoring during hydrocarbon production. From FO sensors for continuous measurements to vibration sensors for real-time monitoring, these technologies provide valuable information for maintaining well integrity and optimizing production processes.

Case-study 1 FO technologies have been successfully utilized for sand detection in the oil and gas industry. An operator in Southeast Asia came upon sand production in offshore gas wells, according to one of the case studies described in the paper (Bale et al. 2020). Although conventional wireline logging equipment was first taken into consideration, they proved to be useless due to the discrete timing of tool runs and the variable nature of sand production. Instead, the operator decided to detect sand using DAS. DAS technology allows for in-situ and continuous monitoring across the entire well during production, providing an accurate understanding of sand production. By utilizing FO cables, the operator was able to detect and visualize sand ingress locations across the entire interval of the producing zones. This real-time monitoring helped the operator plan and execute sand shut-off operations more effectively. The successful application of FO techniques like DAS for sand detection demonstrates the value of FO technologies in addressing the technical challenge of sand ingress in weakly consolidated sandstone reservoirs. Therefore, while DAS technology can provide accurate and continuous monitoring of sand production, there may be limitations in terms of data analysis and interpretation that need to be addressed. Figure 11 to effectively separate the DAS-based acoustic energy spectrum from gas generation, sand production, and sensor noise, a sophisticated data mining method was developed. Considerable scientific work was done to distinguish the continuous and stable properties of gas from the discontinuous and variable qualities of sand. The information also includes a methodical computation of how the total acoustic energy was divided into the domains of contributions from gas and sand.

• Limitations

• 1. Resolution and accuracy The DAS technology may have limitations in accurately identifying the precise locations of sand ingress within the well. The resolution of the acoustic measurements may not be sufficient to precisely identify small-scale sand production events.

• 2. Interference The acoustic energy spectrum obtained from DAS may be susceptible to interference from other sources, such as sensor noise or overlapping acoustic signals from different downhole activities, which could impact the accuracy of sand detection.

• 3. Complex data analysis The process of distinguishing between acoustic energy related to sand production and other downhole activities (such as gas production) might be complex and require sophisticated data analysis algorithms. The accuracy of this differentiation could be a potential limitation.

• 4. Dynamic sand production DAS technology may face challenges in accurately capturing the dynamic nature of sand production, especially in wells where the sanding rate varies over time. Real-time monitoring of sand production events might be challenging.

• 5. Validation The research does not explicitly mention the validation of DAS-based sand detection against other conventional methods. Therefore, the limitations related to the validation and accuracy of DAS-based sand detection are not explicitly addressed.

Fig. 11

DAS used for sand detection (Bale et al. 2020)

Case-study 2 The integration of the DAS system with other surveillance data has been crucial in achieving successful remediation and sand control in the Azeri Chirag- Gunashli (ACG)field (Sadigov et al. 2018). By distinguishing sand ingress signals from other noise sources, the DAS system has proven to be a novel signal processing technique for the real-time diagnostic and remediation of sand production problems. This has not only improved production but also enhanced safety in the field. The DAS system has been effective in detecting sand ingress 'signals' from background 'noise' and has been used in over 30 surveys to identify sand entry zones along the wellbore. This has allowed for targeted remediation using expandable patches to address the sand production issues. The well EX28 DAS, which monitors sand entry over completed intervals and correlates the data with a surface acoustic sand detector (ASD), is shown in Fig. 12 as having real-time data visualization. The lowest two sand intervals without gravel in the picture have been subjected to high-intensity sanding.

Fig. 12

Well EX28 DAS real-time data visualization showing high-intensity sanding across the bottom two sand intervals without gravel (Sadigov et al. 2018)

Figure 13 below is the gravel pack log and wash pipe gauge data used in validating the DAS sand log. The gravel pack log provides information on the effectiveness of the gravel pack job, which is important in preventing sand production. The wash pipe gauge data, on the other hand, provides information on the pressure changes during the gravel pack job, which can be used to identify incomplete gravel packs.

• Limitations Several limitations and challenges are associated with sand detection and quantification using DAS technology in this case study. These limitations include:

• 1. Signal processing complexity A robust signal processing procedure is necessary to identify and distinguish sand ingress signals from other noise sources. This complexity arises due to the capture of all background "noise" by DAS systems, including fluid flow and instrumentation noise.

• 2. Data volume handling DAS systems produce large data volumes, which can create complexities in data handling and increase the analysis and interpretation turnaround time, hindering the ability to make real-time decisions.

• 3. Relative concentration Measurement Effective remediation requires not only a binary 'Yes' or 'No' output for detecting in-well sanding events but also a measure of the relative concentrations of sand entering the wellbore at different sections along the reservoir interval at different production rates.

• 4. In-situ information Existing sand detection methods provide a delayed indication of sanding events and help manage sand risk accordingly, but they do not provide in-situ information about the location of the sand ingresses across the reservoir. Unambiguous identification of the zones contributing to sanding and their relative concentrations is essential for successful sand shut-off operations.

Fig. 13

Evaluation of EX28 Well gravel pack through integrated data: DAS log view, GP log, and wash pipe gauge data interpretation (Sadigov et al. 2018)

These limitations highlight the need for further development and refinement of DAS technology to address the complexities associated with sand detection and quantification in real-time reservoir management.

Case-study 3 Acoustic 'point' sensors that are fixed to the production flowline at the surface are the conventional method of detecting sand. To provide a relative measure of sand ingress across the reservoir section during production, DAS is a novel digital signal processing scheme that separates sand ingress ‘signals' from fluid flow and other background 'noise' in real-time. The results are presented as 'sand ingress logs' (Thiruvenkatanathan et al. 2016).

The first stage of developing the real-time sand detection system was to physically comprehend the "acoustic fingerprint" left by sand entry into wells that produce hydrocarbons. BP conducted a multi-phase flow loop experiment from which experimental data was used to derive the auditory fingerprint of sand entry. As a result, a downhole completion assembly has also been obtained by modeling the results using basic principles. Hard discs were used to store the collected raw data, which was then transported from the platform to the beach for data interpretation. After that, the data was manually analysed to verify and examine the information acquired throughout the reservoir section to pinpoint areas that had acoustic signals that resembled the sand ingress fingerprint that was modelled and experimentally seen during the flow loop experiments. This showed that throughout the production period, the reservoir had numerous depth zones with real-time acoustic concentration. A whole reservoir section's "sand log" was created by averaging the filtered sand acoustic signals across time. Figure 14 is a typical example of the sand log that was therefore produced.

Fig. 14

Schematic illustrating an example DAS ‘sand log’ computed across the reservoir interval to identify sand entry points (Thiruvenkatanathan et al. 2016)

A sample DAS sand log derived from data obtained in field trial 1 is displayed in Fig. 15. The sand log shows the depth-dependent acoustic amplitude over the reservoir interval, filtered for sand infiltration. A review of the log reveals five separate zones of sand ingress, with zones 3 and 4 showing comparatively greater noise levels associated with sand ingress. A greater understanding of the chokes/drawdowns at which each of the sand-producing intervals starts contributing sand is made possible by the drawdown-lapsed sand logs, which provide a better picture of the temporal behaviour of sand ingress across the sanding periods defined downhole. This information helps determine the well drawdown envelope for maximum oil production as well as improve assessments of the degree of sanding in each of the sand ingress zones.

• Limitations

• 1. Single experimental setup The study describes experiments conducted in a horizontal flow loop, which may limit the generalizability of the findings to other flow configurations, such as vertical wells or complex well geometries.

• 2. Limited sand types The study focuses on specific sand types and sizes, potentially limiting the applicability of the findings to a broader range of sand characteristics encountered in different reservoirs

• 3. Surface detection While the study emphasizes the real-time monitoring of sand migration patterns, it primarily focuses on wellbore sanding conditions. The applicability of DAS for sand detection in other reservoir zones or surface facilities has not been extensively explored.

• 4. Environmental factors The influence of environmental noise and other background effects on the accuracy of sand detection using DAS is not extensively discussed. Factors such as temperature, pressure, and fluid composition could impact the reliability of sand quantification.

• 5. Repeatability and variability The study mentions repeatability using multiple flow rates and sand slurry ingress locations. However, the variability of results under different operational conditions and sand characteristics is not extensively addressed.

• 6. Data handling and interpretation The study briefly mentions signal processing methods to reduce data size and improve interpretation. However, the specific challenges related to data handling, interpretation, and real-time monitoring are not thoroughly discussed.

Fig. 15

Representative example of DAS sand log computed from data acquired in field trial 1 (Thiruvenkatanathan et al. 2016)

Addressing these limitations could enhance the applicability and robustness of DAS technology for sand detection and quantification in oil and gas production operations.

Case study 4

The findings of examining the flow of gas, water, and sand in various pipe layouts are covered in a study by Wang et al. (2019a, b, c). It describes how sand particle movement in a fluid-moving medium produces detectable and measurable acoustic waves. An acoustic sensor, measurement tools, and a sand conveying system were all part of the lab setting employed for the investigation. Gas was injected into the pipeline during the experiments, and the flow velocity was adjusted. The sand particles striking the pipe wall generated acoustic waves, which the acoustic sensor picked up on. The goal of the study was to establish a technique for sand particle detection in the flow, which is crucial for the oil and gas sector. The investigation gave rise to the idea of just looking for sand patterns at pipe bends on a surface that wasn't naturally distribute. The study provides a foundation for future research in this area.

• Limitations

The current study has certain limitations. Firstly, its focus on laboratory settings might not accurately capture the intricate nature of offshore oil and gas environments in real-world scenarios. Additionally, the study primarily focuses on the acoustic sensor approach, potentially overlooking other viable techniques for sand particle detection in pipe flow. It is possible that the experimental design and conditions do not accurately represent the diverse range of challenging circumstances encountered in offshore oil and gas operations. Furthermore, the study does not address how variables such as fluid composition, temperature, and pressure may impact the effectiveness of the acoustic sensor technology.

Case study 5

An inventive technique based on DAS is used by Li et al. (2020a, b). To provide a thorough investigation on the measurement of sand concentration in sand-water two-phase flows. By using an optical fibre distribution acoustic sensor (DAS) system, the researcher shows how to identify acoustic signals produced by sand particles hitting the pipe wall. The study uses statistical analysis to describe various sand conditions, specifically focusing on the standard deviation of the acoustic emission (AE) data. Sand concentration ranges from 0 to 0.14 weight percent, with a step of 0.02 weight percent. The experimental results show a strong linear relationship between the standard deviation of the AE signals and sand concentration. This result highlights the possibility of the suggested technique for non-invasive, long-distance pipeline sand content monitoring. The work presents a novel strategy to measuring sand concentration since it places more focus on using statistical techniques than on conventional frequency domain analysis.

• Limitations

The DAS method for sand detection may have a disadvantage in terms of sensitivity to external noise. This is due to its reliance on detecting acoustic waves, which could be interfered with by external noise sources, affecting the accuracy of the sand concentration measurement, especially in environments with high levels of background noise or other acoustic disturbances. Additionally, the size and composition of sand particles may also impact the effectiveness of this approach. Changes in material composition and particle size may affect the acoustic signals produced, making it more challenging to precisely characterize the concentration of sand using this technique.

Below is a summary in Table 5 of case studies focusing on DAS for the detection of sand. The table includes information on the application of DAS, its configuration in wells, key findings, as well as its advantages and limitations. Limitations: the limitations of this work on sand detection and quantification using distributed acoustic sensing (DAS) include the following:

Table 5 Summary of Case Studies of DAS for Sand Detection

DTS technology

DTS technology has been widely used in the oil and gas industry to monitor in-well parameters (Baldwin 2014). It allows for continuous temperature profiling along entire well paths, providing valuable information about flowing profiles, fluid properties, and the condition of artificial lift equipment (Williams et al. 2000). By tracking temperature variations in the wellbore, DTS can identify sand production. Sand has a cooling impact that lowers the temperature when it mixes with the fluid and passes through the wellbore throughout the production process (Yamamoto et al. 2023). These temperature variations, as well as the location and intensity of sand generation, can be detected using DTS. DTS can offer real-time data on sand production by continuously measuring temperature variations. This enables the early diagnosis and prevention of damage to downhole and subsea equipment. Furthermore, DTS can be combined with pressure sensors to offer a more thorough understanding of the dynamics of sand production. Therefore, by permanently installing FO-distributed temperature monitoring systems, sand production can be monitored in real time without interrupting production rates (Brown et al. 2005). Figure 16 shows the principles of DTS in well applications.

Fig. 16

The principles of DTS in well applications

ML techniques for sand production monitoring

Machine learning, a subfield of artificial intelligence, involves the development of algorithms capable of learning from data and improving performance without explicit programming (Orourke et al. 2020). As an alternative to traditional, costly, and ineffective methods, machine learning techniques have proven to be a valuable tool for sand production monitoring. By leveraging advanced algorithms and big data, machine learning models can analyse various types of sensor data, uncover sand production patterns, and predict future production levels. This section delves into the application of machine learning for sand production monitoring, examining the data utilized, the algorithms employed, and the significant advantages it offers for well integrity and production optimization. Machine learning can be broadly categorized into two primary types: supervised and unsupervised learning. Additionally, sand management and geomechanically characterization have made use of ML techniques (Abdelghany et al. 2023). Based on empirical equations or ML models, these methods can forecast crucial characteristics in the energy sector, like sand production (Abdelghany et al., 2022). ML algorithms can offer insights into sand production patterns and aid in the optimization of sand management strategies by evaluating data from borehole images and multi-arm callipers. For efficient mitigation measures and early identification of subsurface movements, real-time monitoring of sand control completions is essential (Earles et al. 2010). To monitor strain and temperature on sand screen products over an extended period, FO support devices have been created (Earles et al. 2010). This technique enables real-time monitoring of the efficacy of mitigation measures in addition to early identification of sand production when combined with optical wet connections.

Supervised learning techniques

A labelled dataset, consisting of matched input and output labels, is used to train the algorithm in supervised learning (Bandura et al. 2018). For the algorithm to be able to predict or classify fresh, unknown data, it must first learn the mapping from input to output. In contrast, unsupervised learning requires that the algorithm identify patterns or structures in unlabelled data without the need for human intervention. Common tasks in unsupervised learning are dimensionality reduction and clustering. DAS has proven to be an effective method for monitoring sand production when ML techniques are used (Mendoza et al. 2022). Real-time detection of sand production is achieved through the analysis of continuous data streams from the field using ML models, which are provided by DAS. To predict multiphase downhole inflow, supervised (ML) combines thermal and acoustic characteristics from DAS. This method has been effectively used in well integrity surveillance, production, injection profiling, and sand production detection. DAS and DTS are incorporated into standard analysis methods for Formation Evaluation by the end-to-end analytics platform that has been documented in the literature. Experts can now limit uncertainty and provide thoughtful recommendations for well remediations or operational modifications thanks to the application of ML in the monitoring of sand production.

Classification

Sand production prediction models have been developed for specific types of reservoirs, such as tight sandstone oil reservoirs (Liu et al. 2021). These models are based on the derivation of the sand production rate and have been validated against field data. Factors such as flowing bottom-hole pressure, sand production radius, and permeability influence the sand production rate. Understanding these relationships can aid in predicting sand production in tight sandstone oil reservoirs. Figure 17 is the general ML processing applied to DAS data and its output and Fig. 18 shows the model training procedure and its subsequent use for DAS -based real-time downhole event classification. To classify downhole occurrences more accurately, the picture illustrates how hybrid ML models incorporate characteristics from DAS and DTS. A decision tree classifier is also depicted in the illustration, with samples from the training set divided into nodes based on features-based conditions. Three steps make up the staged flow characterization process: identifying input intervals in step one; defining fluid phase in step two; and evaluating the relative interval contribution of various fluids in step three. Experts can now thoughtfully offer well remediation or operational adjustments by making use of the application of ML in the classification of downhole events.

Fig. 17

General ML processing workflow on DAS data

Fig. 18

Model training process and subsequent application for real-time downhole event classification modified from Mendoza et al. (2022)

Step 1 Extract relevant features from the detected seismic events such as amplitude, frequency content, energy distribution, and duration of the events will be used to differentiate between events related to sand production and other sources of seismic activity.

Step 2 Prepare a labelled dataset by associating the extracted features with known instances of sand production and non-sand production events. This dataset forms the basis for training a machine learning model.

Step 3 Train a machine learning model, such as a support vector machine (SVM), to classify seismic events as either sand production-related or not.

• Physics –informed ML models

Physics-informed ML models bridge the gap between traditional scientific models and ML algorithms (Hao et al. 2022). While ML excels at finding patterns in data, it cannot often incorporate known physical laws. Conversely, traditional physics models can be highly accurate but require extensive upfront knowledge and may struggle with complex or uncertain situations. Physics -informed ML models address this by combining both approaches. They leverage machine learning's ability to learn from data with the established principles of physics. In essence, the model is trained on two types of information: (1) real-world data like sensor readings or historical observations, and (2) the known physical laws governing the system under study. This can be expressed through mathematical equations or established physical relationships. By incorporating this physics knowledge, physics -informed ML models are guided toward solutions that are not only statistically likely based on data, but also physically plausible.

Regression

Support Vector Regression (SVR), which is grounded in the principles of Support Vector Machines (SVM) for regression analysis, predicts sand production. SVR, as a supervised learning algorithm, examines data and recognizes patterns to generate predictions. When it comes to sand production, SVR is trained with input features, such as well type, permeability, porosity, and other pertinent parameters, as well as corresponding sand production data (Song et al. 2022). The SVR algorithm is designed to locate the best-fitting line (or hyperplane in higher dimensions) that illustrates the relationship between the input features and the sand production output. It accomplishes this by transforming the input data into a higher-dimensional space and pinpointing the hyperplane that maximizes the margin between the data points and the hyperplane. This margin signifies the confidence level in the predictions made by the SVR model. After the SVR model has been trained, it can be used to forecast sand production for new input data by mapping the input features into the higher-dimensional space and identifying the predicted sand production output based on its position relative to the hyperplane. This quantitative prediction capability of SVR allows for more informed decision-making regarding well management strategies and sand production mitigation techniques.

Artificial neural networks (ANN) for sand production prediction and analysis

ANNs are becoming an increasingly valuable tool for forecasting and analysing sand production in oil and gas wells. Conventional methods may be either intricate or lack precision. ANNs, modelled after the structure of the brain, are capable of learning from historical well data that encompasses reservoir properties, production parameters, and instances of sand production. The study by Azad et al. (2011) details various techniques for predicting sand production in oil wells, including empirical models, numerical approaches, and ANNs, and underscores the significance of accurate input data. The research emphasizes the advantages of utilizing prediction models for sand control and management, focusing specifically on the application of ANNs to forecast critical bottom-hole flowing pressure to prevent sand production. By doing so, it highlights the importance of developing precise prediction techniques to comprehend the onset of sanding. This facilitates more efficient well management and production optimization, enabling proactive measures to be taken to avert sand production and its associated economic and safety risks.

Deep learning (convolutional neural networks)

Machine learning is a subfield of artificial intelligence that involves the use of neural networks to analyse and learn from data. In particular, deep learning is a subset of machine learning that utilizes artificial neural networks to learn and comprehend complex patterns in large datasets (Sircar et al. 2021). In the field of sand production monitoring, deep learning has made significant strides, particularly with the use of convolutional neural networks (CNNs). Unlike traditional methods that rely on hand-crafted features, CNNs can automatically extract relevant patterns from well data, including well-log images, pressure readings, and acoustic emission data. By processing this data through convolutional layers, CNNs can learn to identify subtle features that correlate with sand production events. This capability is particularly valuable for analysing complex downhole environments where traditional methods may struggle. For instance, research by Liu et al. (2021) demonstrated the effectiveness of CNNs in identifying sand production from well-log images, achieving superior accuracy compared to conventional ML techniques. As deep learning continues to evolve, the use of CNNs for real-time sand production monitoring has the potential to enable proactive interventions and improved well management.

Random forest

Random Forest can be effectively utilized for predicting sand production during hydrocarbon production by leveraging its ability to handle complex, nonlinear relationships and interactions between multiple variables (Zahedi et al. 2018). In the context of sand production prediction, various geological, mechanical, and operational parameters such as reservoir properties, fluid flow characteristics, wellbore stability, and production rates can serve as input features. By training the Random Forest model on historical data where these parameters and corresponding sand production rates are known, the model can learn to predict the likelihood and volume of sand production under different conditions. The ensemble nature of Random Forest, which combines multiple decision trees, helps in improving the robustness and accuracy of the predictions by reducing the risk of overfitting and enhancing generalization.

Unsupervised learning techniques

Algorithms that employ unsupervised learning examine data that has not been assigned labels or categories. Unsupervised learning enables the model to discern hidden patterns and structures within the data without external guidance, in contrast to supervised learning, which provides input with pertinent outputs. (Khanum et al. 2015).

K-nearest neighbors (KNN) for real-time sand detection

A study conducted by (Wu et al. 2022) aimed to interpret vibration signals captured by DAS technology for wellbore flow analysis. The researchers applied the K-means clustering algorithm to differentiate signals associated with sand production, enabling real-time monitoring and detection of sand influx in the wellbore. This technique classifies new sensor data based on its similarity to past instances of sand production events. By continuously analyzing data streams from pressure sensors, vibration monitors, and acoustic emission detectors, KNN can identify patterns indicative of sand movement in real-time. The key advantage of KNN lies in its simplicity and computational efficiency, making it suitable for real-time applications with limited processing power. Studies by (Ranjan et al. 2019) demonstrate the effectiveness of KNN in real-time fault monitoring systems, achieving high accuracy in anomaly detection [1]. However, KNN's performance can be sensitive to the selection of the "k" value (number of nearest neighbors) and the quality of training data. Further research is needed to optimize KNN for sand detection in real-time scenarios, considering factors like data dimensionality and noise reduction techniques. This approach has the potential to improve wellbore flow analysis and enhance the safety and efficiency of oil and gas operations (Table 6).

Table 6 Machine learning techniques for sand production monitoring

Challenges and limitations of FO technology

Role of FO technology in well applications

FO technology, specifically DAS and DTS, is emerging as a powerful tool for various well applications, including real-time sand production monitoring. DAS utilizes a downhole fiber optic cable to detect and locate acoustic emissions (vibrations and sound) along the entire wellbore (Soroush et al. 2022a). These emissions can be caused by sand production, with characteristic signatures allowing for real-time identification and location of sand events within the well (Sindi 2023). Additionally, DAS can detect changes in fluid flow patterns, potentially indicating an increased risk of sand production (He and Liu 2021). DTS, on the other hand, utilizes the fiber optic cable to measure temperature variations along the wellbore length (Fakiris et al. 2023). This data is valuable for sand production monitoring in several ways. Changes in temperature profile can indicate fluid flow anomalies associated with sand production events (Ukil et al. 2012). Furthermore, DTS data can help identify zones of high fluid flow or water influx, factors that can contribute to sand production (Zhao et al. 2023). Additionally, temperature variations can sometimes point towards well integrity issues that could indirectly contribute to sand production. Compared to traditional methods relying on periodic well shut-in or sampling, FO technology offers a significant advantage: continuous real-time data for enhanced sand production detection and monitoring. The distributed nature of DAS and DTS, providing data along the entire wellbore length, offers a more comprehensive picture of downhole events compared to point-based measurements (Fernández-Ruiz et al. 2020). There are several types of FO sensing and applying them to the oil and gas industry will expand and improve our understanding of wells. FO technologies can be used for injection management, well integrity monitoring, well stimulation and production performance optimization, thermal recovery management, sand detection, and more (Al-Qasim et al. 2021). Hence, the data quality of FO technology is generally high due to its ability to provide real-time data across the area of sensing during production or injection change. This allows users to observe and analyse the dynamic reservoir properties, which enhances the process of decision-making. In addition, using FO technology saves time and cost due to the low installation time required before launching the tool. FO technology, particularly DAS and DTS, is revolutionizing data acquisition and interpretation in well applications beyond just sand production monitoring. This technology offers valuable insights for enhancing the understanding of wells and enabling functionalities such as injection management, well integrity monitoring, and thermal recovery management.

• Injection management

DAS can detect and localize fluid flow patterns within the wellbore, providing crucial information for optimizing injection strategies. For instance, by identifying zones receiving less injected fluid due to channeling or uneven distribution, operators can adjust injection rates or wellbore configurations to ensure a more uniform sweep and improve reservoir performance (Fakiris et al. 2023).

• Well integrity monitoring

Both DAS and DTS can contribute to well integrity monitoring. DAS can detect leaks or casing defects through changes in acoustic signatures caused by fluid movement outside the wellbore (Soroush et al. 2022a). Similarly, DTS can identify temperature anomalies indicative of leaks or heat transfer associated with wellbore integrity issues (Ukil et al. 2012). Early detection of these issues allows for prompt intervention and helps prevent wellbore failures, environmental risks, and production losses.

• Thermal recovery management

DTS plays a crucial role in thermal recovery processes like steam injection. By measuring temperature profiles along the wellbore, operators can monitor steam conformance, identify zones with poor steam distribution, and optimize injection strategies for efficient heat delivery to the reservoir (Bao and Chen 2012). Additionally, DTS can help detect potential issues like steam breakthrough into unwanted zones, allowing for adjustments to injection parameters to ensure optimal reservoir heating. FO technology, through real-time and distributed data acquisition, paves the way for a more comprehensive understanding of well behaviour and enables informed decision-making for various well management functions.

A schematic deployment for land Vertical seismic profile (VSP) acquisition is shown in Fig. 19. One unique benefit of DAS is that it provides excellent spatial sampling over the whole well's length.

Fig. 19

Schematic layout of a typical land DAS VSP acquisition system (Rafi et al. 2024)

Challenges with FO technologies

The development of affordable, simple-to-install FO cables that are suited for a range of FO measurements is one of the primary challenges. Furthermore, managing, storage, and interpreting the massive amounts of data produced by FO sensing technology can be quite difficult. Data quantities in certain applications might reach up to 1 TB/well/day, necessitating sophisticated methods of data processing and interpretation (Koelman et al. 2012).

Data analysis and interpretation

While Fiber Optic (FO) technology offers a wealth of valuable data for well applications, data analysis, and interpretation present certain challenges. Unlike conventional point-based measurements, FO technology like DAS and DTS generates vast quantities of continuous data from distributed sensors along the entire wellbore. Hence, FO data analysis and interpretation demand sophisticated analytics techniques and models. An accurate correlation between downhole events and complex fluid physics and the collected data is required (Bale et al. 2020). Developing scalable and reliable analysis frameworks that can manage the massive amount of data and offer insightful information for decision-making is a challenge (Johannessen et al. 2012). The development of appropriate tools or algorithms to help with or automate the process is necessary for the effective interpretation of this data (Van Der Horst et al. 2013). Hence, large volumes of data are produced by FO technologies; these data must be prepared and arranged correctly to be analysed (Bale et al. 2020). Addressing problems like data collection, storage, and standardization falls under this category. The complicated nature of the data format and structure can make it difficult to comprehend the data and integrate it with other systems. For real-time or nearly real-time analysis, efficient transmission and storage of the collected FO data are necessary. This entails dealing with problems about data management, data security, and transmission rates. It can be difficult to guarantee the data is delivered from downhole systems to the analysis platforms in a dependable and timely manner. Furthermore, there are several significant obstacles in the processing and interpretation of a wide range of complementary FO data. Extracting meaningful insights from this massive dataset requires advanced data analytics techniques. Traditional methods might not be sufficient to handle the complexity and real-time nature of the data. Machine learning algorithms and pattern recognition techniques are increasingly being explored to automate data processing, identify anomalies, and correlate sensor readings with downhole events (Ma et al. 2023). Another challenge lies in accurately correlating the collected data with specific downhole events. Interpreting the vast amount of information requires a thorough understanding of wellbore physics, reservoir behaviour, and the characteristic signatures of various downhole phenomena within the FO data (Zhao et al. 2023). Integrating data from FO technology with other wellbore measurements and production data can further enhance the accuracy of event identification and interpretation (He and Liu 2021). Addressing these data analysis and interpretation challenges is crucial to unlock the full potential of FO technology for well monitoring and management.

Cost

The high cost of FO sensors poses a major challenge to the development of this market. Depending on the applications, operating conditions, and the type of FO cable used, the cost of a DAS system can be very high and still not affordable by every company that requires real-time monitoring and sensing (Ashry et al. 2022). Furthermore, the installation and maintenance of these systems are also costly (Otchere et al. 2024). One of the challenges is the cost of deploying FO cables, which needs to be reduced to levels significantly below typical current costs (Koelman et al. 2012). However, Mad Zahir et al. (2019) discussed the advantages of adopting FO technology, including cost-effectiveness and safety, and gave instances of its successful application in leak detection and well performance monitoring. The development of new subsurface technologies and the requirement for passive monitoring instruments to spot leaks and inadequate cement works during well decommissioning processes are also covered in the paper. While sensor costs are decreasing, downhole FO cables designed for harsh environments are expensive. These costs, along with wellbore factors, can limit FO technology adoption. However, overcoming these challenges could lead to wider use, increased data acquisition, and the development of new well management applications (Soroush et al. 2022b). Researchers are exploring cost-effective cable materials and deployment methods to make FO technology a more accessible and cost-competitive solution for well monitoring and optimization.

Reliability

Although DAS offers benefits for geophysical monitoring, the measurement's signal-to-noise ratio still needs to be improved to match with geophone systems' performance. DAS specifically faces limitations in signal-to-noise ratio (SNR). Extraneous noise from downhole sources like pumps, fluid flow, and even earth tides can obscure the weak acoustic signatures of events like sand production (Soroush et al. 2022a). Improving SNR is crucial for accurate data interpretation and reliable event identification. This improvement would increase the range of applicability of DAS in terms of the depth of targets and time-lapse sensitivity (Van Der Horst et al. 2013). Another reliability challenge lies in the electronic components used for data acquisition and processing in FO systems. These electronics can be susceptible to downhole environment factors like high temperature and pressure, potentially leading to system malfunctions and data loss (Fakiris et al. 2023). Furthermore, in high-temperature environments, electronic systems historically have higher failure rates, which has created a need for more reliable alternatives (Hou and Kheong 2010). However, the application of permanent FO monitoring systems, such as FO pressure and temperature gauges, has been successfully demonstrated in various scenarios, including deepwater applications, sand face profiling, injectivity loss identification, and well monitoring. These FO systems have proven to be robust and reliable, capable of withstanding high temperatures and hostile operating conditions. Another challenge is the collection and use of data from permanent installations of FO cables in dry tree wells. The issue is not with the deployment step but with the reliability of wellhead connectors and the need for integrated systems for true real-time data acquisition capability and adapted software for data visualization (Allanic 2012). Additionally, practical problems related to splicing methods, wet connects, and making perforations without damaging pre-installed FO need to be solved. Another challenge is related to the robustness and reliability of the system. Early failures and degradation are currently attributed to leaking connectors and nipples or mechanical stress caused by poor fiber immobilization (Cheng et al. 2013). Developing more reliable alternatives to these electronic components, potentially using high-temperature and pressure-resistant materials, could significantly enhance the overall reliability of FO monitoring systems.

Sensitivity to environmental factors

FO cables and sensors can be sensitive to environmental factors, such as temperature, humidity, and vibration. This can cause the cables to expand or contract, which can misalign the fiber and reduce signal strength. Additionally, environmental factors can also damage the FO sensors, making them unreliable (Maier et al. 2023). FO technologies face several challenges, especially when deployed in high-temperature reservoir environments. Although FO cables are often made to withstand tough conditions, very high or low temperatures might have an impact on the way signals are transmitted. Exceeding the cable's specified temperature rating in high-temperature reservoirs may result in signal attenuation and perhaps compromise data integrity (Fakiris et al. 2023). One significant challenge is the lack of installation experience in these environments, which makes it difficult to quantify the full cycle benefits of implementing these technologies (Hou and Kheong 2010). Moreover, difficult to install fiber in certain well types to acquire meaningful injectivity or flow profiles, the FO cable needs to be installed over the injection or production interval. However, this can be challenging in subsea wells or wells with liner sections. Further work is required to develop low-cost deployment techniques for installing fiber in these types of wells (Van Der Horst et al. 2013). One of the challenges is the deployment of FO in subsea wells and multi-stage completions. The optical continuity must be ensured through all parts of the completion and the Christmas tree through optical connectors (Allanic 2012). Optical connectors have been tested, but some configurations have failed for different reasons. However, Earles et al., (2010) described the development of new techniques and equipment to install and connect FO in sand control completions. Another challenge is extreme operating environments, physical space limitations, and/or unique measurement requirements (Johannessen et al. 2012). Existing electrical/quartz sensing technology has always had a basic physics temperature limitation of around 150 °C and does not readily meet extreme temperature requirements. However, FO sensing systems have the potential to deliver well and reservoir data along the entire length of the well and use that data for production optimization to a level that is not possible with other technologies alone. The challenges include rigorous qualification of components, designing systems to accommodate a diverse range of reservoir conditions and completion types, managing high rates of data under conditions of great water depth and significant offset distances, and internal company hurdles presented by critical cross-functional activities (Patni and Dria 2014). Furthermore, Moisture entry into the cable can deteriorate the signal and possibly permanently harm the fiber core. Using materials that are resistant to moisture and maintaining adequate cable sealing is essential for dependable performance in high-humidity conditions (Ukil et al. 2012). Finally, overly strong vibration may interfere with the DAS signal with noise, making it more difficult to detect events and interpret data accurately. It is crucial to reduce vibration by using appropriate cable deployment and wellbore completion methods (Soroush et al. 2022b).

The technology is still at an early stage, and its full potential has not been fully understood yet. However, FO technologies can withstand extreme conditions imposed by the downhole environment, unlike other conventional technologies. There are several types of FO sensing, and applying them to the oil and gas industry will expand and improve our understanding of wells (Al-Qasim et al. 2021). Despite the challenges, advancements in FO sensing technology have improved the selection of equipment and techniques for deployment in challenging environments, giving confidence to operators to utilize FO technologies in difficult high-temperature reservoirs such as fractured basement granite (Hou and Kheong 2010). It will need significant work to develop strong interpretive approaches in addition to additional advancements in our understanding and modeling of the underlying physics to overcome these challenges. The industry is about to witness a significant advancement in well and reservoir monitoring capabilities, regardless of these challenges. To produce value-driven solutions and fully realize the promise of FO technology in the oil and gas sector, these issues must be resolved.

Future research directions

The integration of DAS technology and machine learning (ML) holds tremendous potential for enhancing our capabilities in this critical area, as the landscape of sand production monitoring during hydrocarbon production is constantly changing. The development and application of ML algorithms for sand production prediction is one of the major future paths we envision. ML models can be developed to predict sand production events as well as their intensity and duration using the vast amount of data collected by the DAS system. By allowing operators to proactively implement mitigation measures, these predictive capabilities can help reduce costly downtime and equipment damage. Furthermore, if ML techniques continue to advance, such as deep learning and reinforcement learning, there is a possibility that complex patterns in the data that were previously difficult to discern will be uncovered. This would improve the accuracy of sand production prediction. Some potential future research areas, but not limited to, are outlined below.

• A. Developing advanced data analytics techniques The integration of machine learning (ML) and artificial intelligence (AI) in the application of DAS for sand detection during hydrocarbon production can be considered a noteworthy tool that can be incorporated to improve operations and ensure the safety of the oil and gas industry. This section discusses potential future research directions in this area.

• 1. Integration of DAS data with machine learning models The most recent research findings have been integrated with real-time data from DAS using machine learning (ML) methodologies to improve the precision of sand detection. Future studies will focus on employing more advanced ML models, such as deep learning neural networks like Convolutional Neural Networks (CNNs) or Recurrent Neural Networks (RNNs), which can more accurately detect structures in DAS data by identifying concealed patterns.

• 2. Real-time monitoring and decision support systems Developing an efficient and reliable monitoring system equipped with artificial intelligence capabilities can facilitate continuous monitoring of sand production. Such systems can detect early warnings of sand influx or equipment malfunctions. This information can be utilized to make informed decisions promptly to mitigate risks in the production system and enhance the effectiveness of operational processes. Studies in this area can focus on the development of AI models that are continually refined using real-time data updates to improve the accuracy of forecasts.

• 3. Data fusion and multi-sensor integration Integrating DAS data with other sensor data, such as pressure sensors and temperature gauges, can provide a more comprehensive understanding of the reservoir's condition and the degree of sand erosion. Future efforts could involve the integration of processes that combine data, such as Bayesian inference or Kalman filtering, to enhance the dependability of sand detection algorithms by incorporating information from multiple sources.

• 4. Anomaly detection and predictive maintenance The utilization of an AI-based anomaly detection algorithm allows for the identification of abnormalities in compressed air source data, which can indicate instances of sand production or equipment malfunction. Through the analysis of past data and the application of machine learning techniques, predictive maintenance models can estimate impending device failures, enabling proactive measures to be taken and reducing downtime. This approach can lead to enhanced equipment performance and overall efficiency.

• 5. Optimization of data acquisition and processing The research area under consideration pertains to enhancing the parameters of data acquisition for DAS, specifically focusing on the improvement of sampling rates and spatial resolutions, to achieve more favourable outcomes for the detection of fine-grained sand. In addition, the utilization of advanced tools such as noise removal algorithms and feature extraction techniques that are continuously being refined, will significantly contribute to the efficiency and precision of data analysis algorithms.

• 6. Explainable AI and uncertainty quantification The increasing complexity of AI-powered systems necessitates the development of explainable AI techniques that can provide insights into the decision-making process and decode model predictions. Additionally, the ability to quantify uncertainty in AI predictions is crucial for assessing the reliability of sand detection methods and managing risk levels. Future research may explore the possibility of integrating explainable AI and uncertainty quantification methods into DAS data analytics frameworks.

• B. Research opportunities for cost reduction Reducing the financial expenses associated with fiber optic (FO) sensors, cables, and installation is crucial for broadening the acceptance of FO technologies across multiple industries, such as oil and gas, infrastructure monitoring, and healthcare. Investigating novel manufacturing methods, materials, and designs offers several promising research avenues for achieving this objective.

• 1. Material innovation Research can offer materials with properties that are suitable for use with cable sensors or wiring, promoting both feasibility and cost-effectiveness. Moreover, discovering alternative materials such as polymers or specialized coatings that possess or surpass the functionalities of metallic materials at a more affordable cost would constitute a significant breakthrough in minimizing production expenses.

• 2. Manufacturing process optimization For instance, discovering alternative production methods like additive manufacturing (3D printing) or roll-to-roll technology can result in the recreation of entire production streams and cause a surge in materials savings. Taking in the manufacturing profiles optimally and deploying automation technologies help to maintain a constant supply of high-quality products at a lower cost.

• 3. Integration of nanotechnology Employing nanomaterials and nanofabrication technologies will be key to achieving miniaturized fiber optic sensors and cables having less demand for material with simpler fabrication processes. Nanomaterials also contain improved exposure to the sensors as sensitivity and durability are increasing with cost reduction through economies of scale.

• 4. Flexible and wearable designs Designing FO sensors and fibers with flexible and wearable configurations may help for the reduction of installing costs by just placing them in different environments including places difficult to deal with or integrating them perfectly with existing infrastructure. Categorically, R&D can investigate new materials and fabrication techniques to make flexible substrates and impregnating materials.

• 5. Standardization and scalability Manufacturers of such products will be able to benefit if they come up with standardized designs and manufacturing processes which may drive the economies of scale to reduce the production cost. Research collaborations between sectors and academia drive the adoption of similar technical specifications and standards and production therefore becoming a repetitive process eventually leading to cost reduction.

• 6. Robotic installation technologies Studying robotic or autonomous technologies installation for FO sensors and wires considering minimization of labor costs and effectiveness in high-risk environments is beneficial in not only enhancing operations but also reducing labor costs. Research can examine building robotic platforms that will use fewer humans to deploy and maintain FO sensor networks but can function with practically no human interference.

• 7. Lifecycle cost analysis Decision-makers who use total life cycle cost analyses to find ways to cut costs wherever possible throughout the value chain, from the manufacturing system all the up to operation and upkeep. Gaining the insights that help in identifying cost drivers, and trade-offs between design choices and manufacturing processes can be consequential for strategic decision-making and optimization's effectiveness.

• C. Research areas for improving reliability A major challenge for FO systems is to make their networks more reliable, and this may pose the largest barrier to FO systems from wide adoption across various industries. Here are several research areas aimed at enhancing the reliability of FO systems, including developing more robust sensors and cables and enhancing the signal-to-noise ratio of FO sensing systems.

• 1. Sensor and cable design optimization Research can be maneuverer to enhance the configuration of FO sensors and wires in such a way as to preserve maximum stability and resistance to the harshest environmental factors. Such material exploration focusses on enhanced mechanical properties such as strength, corrosion resistance, and temperature stability to achieve efficient operation in oil and gas wells, structural monitoring applications, and other areas of high temperature and humidity.

• 2. Environmental sealing and protection By creating more sophisticated environmental sealing methods and industry-standard FRP materials for FO sensors and cables, the ingress of moisture, chemical exposure, and mechanical damage can be prevented, which guarantees longer operation life and reliability. Researchers may work with solidifying materials that are not used frequently, as well as methods of encapsulation that are suitable for application of interest.

• 3. Failure prediction and prognostics The implementation of a predictive maintenance system through continuous monitoring and analysis of key functional parameters, such as temperature, pressure, and capacity, of the FO system will enable early detection of potential failures and facilitate proactive interventions. This can involve the development of machine learning algorithms and data analysis techniques that can identify indicators signalling the impending failure of sensors or cables, thereby ensuring timely maintenance and minimizing plant downtime.

• 4. Redundancy and fault-tolerant architectures The use of built-in redundancy and a fail-proof architecture in FO systems is designed to minimize the likelihood of negative component failures and improve overall reliability. This study will examine current algorithms for fault detection and isolation, sensor redundancy, and network criticality to ensure the dependability of operations and the integrity of data in the face of malfunctions or shifts in patterns.

• 5. Signal processing and noise reduction The capacity of a FO sensor system to differentiate between a useful signal and background noise is of paramount significance to its ability to achieve high accuracy and efficiency, particularly in high-noise environments. The development of signal processing methods, such as adaptive filtering, spectral analysis, or coherence-based techniques, can aid in the suppression of interfering sources and the enhancement of signal quality in FO sensor measurements.

• 6. Calibration and performance validation Developing standard out-of-the-box calibration and performance assessment plans for FO sensors and cables is essential to ensure reliability across different manufacturing batches and environments. By doing so, it will guarantee that the sensors and cables meet the required specifications for the intended application. It is important to note that the calibration standards, uncertainty quantification, and validation methods used in research tasks should be comprehensive and reliable. In addition, in-situ validation should be implemented to ensure that the sensors and cables perform optimally in actual field conditions.

• 7. Long-term stability and aging effects Analysing the long-term dependability and aging effects of FO sensors and cables may be advantageous in determining the functionality of the links over extended periods of use. Examining the aging process through advanced testing methods, environmental exposure studies, and degradation modelling may be useful in simulating various environmental factors, such as temperature fluctuations, mechanical stress, and radiation. This research could be beneficial in ensuring the reliability of the links in various applications.

Conclusions

• 1. Distributed acoustic sensing (DAS) technology has the potential to be utilized for sand detection in oil and gas wells. The basis for this application lies in the fact that the vibrations and strain caused by sand particles flowing through the wellbore can be detected and analysed using DAS. By analysing the vibrations and strain patterns along the fiber optic (FO) cable, it is possible to identify the presence of sand production.

• 2. The utilization of machine learning (ML) techniques is essential in analysing the vast amount of acoustic data gathered by DAS. These algorithms are trained to recognize the distinct indicators of sand production. The ability of these algorithms to differentiate between sand signals and other auditory sources, such as fluid flow and background noises, enables accurate sand detection. Additionally, advanced models may even calculate the quantity of sand produced based on the properties of the acoustic signals. This information can be invaluable for real-time detection and quantification, assessing the integrity of the wellbore, and implementing appropriate mitigation measures to prevent sand-related issues, such as pipeline damage and flow restrictions.

• 3. To implement DAS for sand detection, it is essential to establish a baseline and assess the variations in vibration and strain patterns over an extended period. This process necessitates precise calibration and validation to guarantee the reliable and accurate detection of sand production.

• 4. Additional investigation and advancement are essential to enhance the application of DAS for the detection of sand in oil and gas wells. This involves refining the methods of data analysis, enhancing the signal-to-noise ratio, and tackling any obstacles connected to the unique circumstances and traits of sand production in various well settings.