Keywords

1 Introduction

With the aim of increasing the sustainability, circular economy and resource efficiency of fiber-reinforced plastics, a new process variant of the pultrusion process is being developed as part of the CaproPULL project. The so-called “in-situ pultrusion process” to produce thermoplastic profiles based on nylon 6 offers decisive advantages over the processing of non-recyclable and non-functionalizable thermoset profiles that is common in the state of the art.

To produce the thermoplastic profiles, dry fibers stored in large racks are continuously pulled through a preheating and drying unit before entering a steel die where they are impregnated with the matrix. The highly reactive two-component matrix, which is stored in a metering and mixing unit, immediately begins to polymerize in the die at defined temperatures in several heating zones. When the fully cured profile leaves the die, it cools until it reaches the puller and the saw, where it is cut to the desired length.

Due to the high number of process and material parameters of the very sensitive material, a data-based process development and optimization would significantly increase the development time and product quality and thus shorten the time to market and enable a wide use and substitution of the current non-sustainable products.

Since the pultrusion process is very simple in its principles, digitization including machine and sensor data acquisition and processing is not common today [1]. Taking into account the Industry 4.0 maturity index in Fig. 1 and the classification of [2], pultrusion may currently be assigned to the “Industry 3.0 stage” (step 1–2: computerization - connectivity), as there are predominantly no or only a few connections between the main systems pultrusion line, injection unit, preheating unit and various sensor systems.

Fig. 1
A bar graph plots the Industry 4.0 maturity index versus the development path. The increasing bars are computerization and connectivity in Industry 3.0, and visibility, transparency, predictability, and adaptability in Industry 4.0. Project Capro PULL leads from computerization to predictability.

Industry 4.0 maturity index and objective in Capro-PULL project. Fig. adapted from [2]

Full size image

When working with limited process data without a central unit, the synchronization of different files with different and sometimes missing timestamps, different sampling rates and multiple data sets with the same information leads to unstructured datasets and makes an accurate and time-dependent interpretation of process-induced effects time-consuming and complex, so that many effects and findings are not determined. Consequently, there is an urgent need to work out the basics and steps to prospectively implement ML methods to improve the quality of the process and profiles.

The question therefore arises as to how a data acquisition system can be designed that provides heterogeneous data in a standardized and reliable manner for data-based process development and optimization. The aim is therefore to develop a digitized and standardized data acquisition prototype for the continuous production of sustainable profile structures so flexible that the various components of the production line can be adaptively combined in a central data acquisition system.

The paper discusses the state of the art in industrial communications, highlighting standardized communication protocols that will be considered in the subsequent concept development of the standardized data acquisition prototype. This includes defined requirements for data acquisition, processing, and future storage of the given systems. Necessary modifications, extensions, and the use of retrofits of the given systems are presented based on a standard protocol, followed by a final concept evaluation.

2 Industrial Communication – State of the Art

Choice of sensors as well as transmission options and protocols heavily determine the capabilities of the data acquisition process. While the sensor technology is mostly determined by the specific experimental conditions, transmission protocols are often determined by the manufacturers of the sensor technology or the data acquisition systems. This is mainly due to the historical development of industrial communication.

Starting with parallel cabling of sensors, actuators and controllers, industrial communication has been developed since the 1980s. Initially, serial interfaces such as MODBUS RTU or PROFIBUS enabled to set up networks for exchanging information in production while maintaining real-time capability. However, the limits were the number of network participants and the communication in the IT systems [3].

The development of Industrial Ethernet intended to overcome these limits and enable communication across multiple company levels. In this context, new industrial transmission protocols and fieldbus successors such as PROFINET or EtherCAT emerged.

One of the biggest challenges of the fieldbus is the lack of protocol compatibility among each other. High investment cost for machinery requires long usage times leading to heterogenous protocol environments being the norm. Gateways or similar individual solutions are therefore necessary for the integration of devices in cross-company applications [4]. With rising awareness for the necessity of standardization of data exchange in industrial control systems standards like OPC UA emerged, as well as extensions to network protocols like MQTT. There are many common industrial protocols, but due to the advanced standardization, this paper will focus on OPC UA and MQTT.

Message Queuing Telemetry Transport (MQTT)

MQTT is a communication protocol for machine-to-machine communication. It is considered to be particularly lightweight and easy to implement. MQTT is based on the publish-subscribe model, in which messages can be distributed from one instance (publisher) to many interested parties (subscribers), usually via middleware (the broker). Messages are written to so-called topics, which can be organized in a tree structure. Clients can either subscribe to these topics, i.e. receive messages in case of updates, or publish them, i.e. publish messages themselves. The protocol offers various security mechanisms such as encryption and authentication [5]. Simplified data modeling and standardization of this is introduced with the Sparkplug B specification, which defines topic namespaces, making it suitable for industrial data exchange in IIoT scenarios [6].

Open Platform Communications Unified Architecture (OPC UA)

OPC UA is a communication standard that defines a comprehensive data model and mappings to various communication protocols. Data transport and data modeling are largely independent of each other. Both client–server and publish-subscribe models can be used for communication, with the former currently being used more frequently. A mapping to MQTT also exists. OPC UA stands out especially due to its data modeling. While fieldbuses only transmit values, these can be extended in OPC UA with the corresponding semantics. Thus, a value can be supplemented with information about the measured variable and the unit in which it is represented [7]. Key benefits of OPC UA compared to data models developed on top of existing protocols like MQTT are the fully developed and standardized information models for specific machine and equipment types, enabling vendor independent development of applications.

3 Concept Development for Machine Data Acquisition

In order to develop a concept for the collection of data, first requirements are documented, then standards are selected and consequently applied in the concept design.

3.1 Requirements for a Standardized Data Acquisition

To develop a feasible data acquisition concept for pultrusion, key requirements for such a system must be defined. Requirements result from interviews at research level, literature research and were supported by an industry questionnaire for pultruding companies to fulfill estimated future needs and demands. Requirements regarding the raw data can be deducted from the list of parameters that drive the pultrusion process. Technical requirements regarding the data acquisition, processing and storage must be carefully tailored to the peculiarities of the pultrusion process to enable analysis in the future. Resulting requirements are listed in Table 1.

Table 1 Requirements for data acquisition, processing, and storage
Full size table

3.2 Selection of Preferred Standards

Aspects like time synchronization, adequate resolution, accuracy, acquisition rate and completeness of acquired data are heavily dependent on the chosen data model and protocol, hence their selection is essential. When weighing up the various communication standards, OPC UA and MQTT stand out above all because of their built-in security mechanisms and data modelling possibilities and their high reliability.

With the so-called sessions, OPC UA enables to resend information sent in the meantime after a connection has been lost and successfully re-established. MQTT offers Quality of Service, which can be configured in three levels. The highest level ensures that the receiver has received the information exactly once. Therefore, one of these two protocols will be used at the interface between data acquisition and processing.

OPC UA as well as MQTT provide built-in data modelling possibilities. While OPC UA standardizes a comprehensive and at the same time extensible information model, the latest MQTT Sparkplug specification B allows a simplified data model to be built using a topic structure (like a folder path). Likewise, the OPC UA specifications offer a mapping of its data model to PubSub protocols like MQTT in OPC 10,000–14.

For the data model, the information model of OPC UA is used due to the extensive modeling possibilities, the already profound standardizations, and the possibility of transmission via MQTT. The novel combination of these two standards (OPC UA over MQTT) enables a new comparison of the two standards. The project will examine which transmission path (MQTT vs. OPC UA) is the more reliable and suitable approach. Thus, the goal for data acquisition is defined based on the interface for data processing. In the next step, the existing systems and components are analyzed, and integration steps are defined.

3.3 Retrofitting a Standardized Data Acquisition System

The pultrusion plant at Fraunhofer ICT exemplarily reflects a common production as it can be found in manufacturing companies as the three main components of the system show different interfaces and degrees of digitalization. The overall system is shown in Fig. 2 and consists of a pultrusion system including mold, a mixing and metering system, and a heating chamber for pre-drying, where the former is most digitized and interconnectable whereas the latter has no digital interface at all.

Fig. 2
An architecture diagram. Scenario 1 with a pultrusion plant, a mixing and metering plant and an industrial P C, scenario 2 with mold and a data acquisition box, and scenario 3 with a heating chamber and P L C, connect to middleware followed by data processing, databases, cloud, and further services.

Schematic overview of machine data acquisition and transmissions within the in-situ pultrusion line

Full size image

We define those three sub-systems as distinct scenarios as they impose different requirements on our system that is supposed to eventually integrate data from the heterogenous sources in a uniform way.

Scenario 1, The pultrusion plant and the mixing and metering plant use PROFINET for communication. At the same time, the systems are capable of implementing standards such as OPC UA or MQTT due to their state-of-the-art control technology.

Scenario 2, in a data acquisition box, sensor signals for temperatures and pressures in the mold are currently processed with an amplifier and a transmitter. Latter has a serial interface which can be processed on a standard computer using Labview.

Scenario 3, the heating chamber for pre-drying the fibers currently has no data interface. An expansion of the sensor technology is desired since the drying process influences process stability and product quality.

Digitization of Production Line Components

According to the degrees of digitization, the production line components must be integrated into a data acquisition system via several steps. In the first step, the transmission protocols and the information transmitted in the process are standardized to be able to supply constant, homogeneous data quality for each device in the future.

Scenario 1, the pultrusion line as well as the mixing and metering unit can be upgraded to common communication standards via their PLCs. Nevertheless, it is planned to use a PROFINET-capable industrial PC as a gateway to separate the production network from the data acquisition network. Due to security considerations, they are separated with distinct hardware. The industrial PC publishes the process parameters of the two machines via OPC UA to obtain standardized access to the variables.

Scenario 2, the current Labview program can be extended with Labview’s OPC UA toolkit, which allows the measurement data to be provided in an OPC UA server to the middleware. The OPC UA interface replaces the Windows solution and secures it with common security mechanisms.

Scenario 3, in a retrofit an additional data acquisition box will be installed at the heating chamber, for implementation of additional sensors (humidity and temperature). The acquisition system also consists of a transmitter and amplifiers for the individual sensors. Depending on the transmitter there are various scenarios for data recording. A transmitter with serial MODBUS interface is selected to minimize costs. This requires pre-processing of a signal to transform to common interfaces (such as OPC UA or MQTT). This is realized via programmable logic controllers (PLCs). In this project, the aim is to compare common, powerful PLCs with compact Arduino-based PLCs.

Standardized Middleware

After these adjustments the production line components can supply process data to other systems. In this state, the data is only partially standardized and synchronized in time. In order to meet the requirements and to provide data in a standardized format, a middleware is implemented as the center of data acquisition. It abstracts from the proprietary or semi-standardized protocols and represents a generic approach to data provision. It is implemented on the edge to pick up data from the components, synchronize it in time via NTP and provide it in an aggregated data model according to common OPC UA standards. So the task is to combine data and prepare it in a standardized way. It is still to be tested if data will be transmitted through OPC UA or MQTT using the OPC UA data model. This provides standardized and easy access not only for the following data processing but also to other services developed in future.

Data Processing and Data Persistence

Signal data or variables of the protocols are stream processed. They trigger a data stream, which can be modified and further processed. The standardized information models of the OPC UA and MQTT interfaces simplify the integration of data acquisition into data processing and allow the reuse of processing mechanisms. The challenge is to find out at which level such data processing can take place - directly on a PLC, on an edge device within production or in a cloud environment.

After processing, the data shall be stored permanently for documentation. For this purpose, two approaches are compared: The standardized formats of OPC UA allow the use of NoSQL databases that enable fast storage of the data. At the same time, the data model based on OPC UA also offers a basis to model a schema for a relational database that stores the data for later analysis. Parallel use is therefore possible, and the two approaches are to be evaluated in direct comparison.

3.4 Concept Evaluation

Table 2 takes up the requirements defined in Sect. 3.1 and considers their fulfillment by the developed concept.

Table 2 Evaluation of the concept against the given requirements
Full size table

4 Summary and Outlook

A concept for data acquisition, processing, and storage for the in-situ pultrusion process was developed as part of the CaproPULL project. Data acquisition is based on and at the same time uses novel combinations of state-of-the-art communication standards. In future trial campaigns, the developed concept will be validated in the existing production environment. For this purpose, hardware and software adaptations are currently being carried out. This includes the digitization of the various components, the setup of the database as well as its database modeling based on common OPC UA standards. Within the scope of the work, recommendations for actions to control the process will be derived. In addition, the development of a uniform data model provides the basis for the development of future ML and AI applications.

This work not only serves to optimize the in-situ pultrusion process, but also offers the potential to optimize already established process variants with other material combinations (e.g., thermoset matrices) in terms of economic efficiency and process robustness. This is possible as the high number of parameters and their known, but also partly unknown interactions can be structured and efficiently processed by statistical methods.