Keywords

1 Introduction

Circular economy (CE) promotes switching from the traditional take-make-use-dispose economy to a restorative and regenerative one [1]. Due to increasing regulations, such as extended producer responsibility, manufacturers also have a vested interest in exploring closed-loop CE productions [2]. Achieving full sustainability potential of closed-loop flows requires prioritizing reuse and remanufacture at the component-level (i.e., EoU parts harvesting), over conventional practices like recycling [3]. However, this is acutely challenging with increasing rate of product design generation refreshes.

A significant part of the challenge is the mismatch between product demand (i.e., production flow) and EoU returns flows. As Fig. 1 depicts, a generation's (say, Gen1) return curve lags its production curve by the product's average useful life (uavg) [4, 5]. For example, when a smartphone's average useful lifetime is passed, the demand for that generation is markedly reduced [6]. One generation's potential parts harvesting depends on the overlap between its production and return curves. However, suppose some components (used here to refer to parts, components, modules) have shared designs between the successive generations (i.e., intergenerational commonality—IC [3]). That brings additional parts harvesting potential between generations if Gen2 production and Gen1 returns curves overlap (Fig. 1). Multi-generational product designing—design considering successive product generations—capitalizes on these opportunities. For example, overall sustainability performance of multi-generational product systems (PS) can be optimized by carefully selecting the IC level [7].

However, as Fig. 1 illustrates, different product's Gen1 return flows may start at different points in time, depending on product and market characteristics. At early design stages, the problem is that the designers lack objective methods to assess a product and its PS to decide on IC. Because increasing commonality can also limit product functionality/performance (e.g., due to restrictions in using newer technology incompatible with the prior generation), these decisions must be taken considering the overall sustainability implications. Only a very few work [4, 5] discusses multi-generational sustainable product design. Therefore, this work aims to explore a potential method and an indicator for designers to make initial assessments of a PS's suitability for closed-loop production and IC. Though this paper focuses on a single product system for brevity, the idea also extends to product families. To analyze a multi-generational PS and its sustainability performance during the design stage, first, the closed-loop production must be forecasted (Sect. 2). Then, Sect. 3 discusses the analysis of this PS using the proposed indicator.

Fig. 1.
figure 1

Production and return curves in multi-generational PS

Full size image

2 Simulation Setup

We use a prior simulation framework [4] that can forecast sales and return curves and then calculate its sustainability performance. A detailed description of the framework is available in the original publication [4]. This section summarizes the pertinent calculations.

Model assumes that both newly manufactured (containing only brand-new components) and rebuilt (containing at least one reused/remanufactured component) products fulfill the market demand and are made to similar specifications. Therefore, the customers do not differentiate them. Parameters such as product's useful life, return/reusable/remanufacturable/recyclable rates, and demand are typically value distributions. Since this method is to be used at product design stage, the calculations are done using forecasts (i.e., imperfect information) of those parameters. Prior work [8] has shown how the Monte-Carlo method can introduce stochastic modeling to represent uncertainty in such cases. In this work, those parameters are assumed to be deterministic. The diffusion parameters p and q are assumed to be the same for Gen1 and Gen2 [7]. For simplicity, we model only the variable costs and GHG emissions.

2.1 Modeling the Product Demand and Returns

Previous work discussed [7] the use of the Norton-Bass (NB) diffusion model [9] to simulate the sales of two generations of a product. NB model Eqs. (1) and (2) describe demand variations D1(t) and D2(t) for Gen1 and Gen2, respectively. All symbols have usual meanings [7] (and also provided in Table 1). Assuming the interval (tintro) between generation introductions are the same (e.g., yearly refresh of smartphone generation), when \(\tau_{i}\) is introduction time of Gen-i, \(\tau_{3}\) can be approximated by \(\tau_{2} + t_{intro}\).

$$ D_{1} \left( t \right) = F\left( t \right)m_{1} \left[ {1 - F\left( {t - \tau_{2} } \right)} \right] $$
(1)
$$ D_{2} \left( t \right) = F\left( {t - \tau_{2} } \right)\left[ {m_{2} + F\left( t \right)m_{1} } \right]\left[ {1 - F\left( {t - \tau_{3} } \right)} \right] $$
(2)

When p is the coefficient of innovation and q is the coefficient of imitation [10],

$$ F\left( t \right) = \frac{{1 - e^{{ - \left( {p + q} \right)t}} }}{{1 + \frac{q}{p}e^{{ - \left( {p + q} \right)t}} }}{ },{ }t \ge 0. $$
(3)

We assume the manufacturer's marketing department provides baseline parameter values, including p, q, and m, based on their historical and market data. Estimation of these parameter values is discussed prior [4, 8], and beyond the scope of this paper.

2.2 Production-Mix Calculations

To forecast the production requirements, we use the configuration design framework [11, 12]. Product is considered to include C0 number of distinct component-types. For Gen2, the designs of these components can have two choices (indexed v in the below equations): 1) a design common with Gen1 (which allows intergenerational parts harvesting), or 2) a new design. However, at least one component of Gen2 out of the total number C0 must be a new design to differentiate the Gen2 product from the Gen1.

Product sustainability metrics are set to vary according to the component choices. Therefore, first, the component-level production plan for each production period (i.e., production-mix [7]) is calculated. Then, the sustainability metrics are forecasted based on the said production-mix. Finally, the metrics are aggregated and averaged for the entire production run for ease of analysis [7].

The number of Gen-i recovered products in time (t + uavg) is given by Eq. (4). The model assumes the forward and returns logistics time is negligible (or, can be included in uavg). Therefore, the returns curve follows the demand curve with a delay of uavg [4]. The proportion of Gen-i products returning to closed-loop production flow out of the total reaching their EoU is the recovery rate (Ri).

$$ N_{i}^{recov} \left( {t + u_{avg} } \right) = D_{i} \left( t \right)R_{i} ,\,\,\,\,i = 2 $$
(4)

Once the products are returned at EoU, the manufacturer disassembles those to the component level. Initially, the EoU components are allotted to different EoU streams (reuse, remanufacture, recycle, and dispose) according to the expected availability rates (\(\gamma_{i,c,v}^{x}\), where x can be EoU activities like reuse, remanufacture, and recycle). When c is the component-index, these component-level numbers are given by,

$$ N_{i,c,v}^{x} \left( {t + 1} \right) = { }\gamma_{i,c,v}^{x} N_{i}^{recov} \left( t \right), $$
(5)

where i = 1, 2, c = 1,2, …, C0, v ∈ {1, 2}.

Assuming that only a single instance of each component-type is present in a product, Eq. (6) gives the number of components needed in each period t.

$$ N_{i,c,v}^{prod} \left( t \right) = { }D_{i} \left( t \right),\,\,i = 1,2,\,\,c = 1,\, \ldots ,\,C_{0} ,\,and\,v \in \,\{ 1,2\} . $$
(6)

The production-mix is the combination of the number of newly manufactured, reused, and remanufactured components produced in a period [7]. Therefore, it is calculated by the general expression,

$$ N_{c,v}^{prod} \left( t \right) = N_{c,v}^{newmanu} \left( t \right) + N_{c,v}^{reuse} \left( t \right) + N_{c,v}^{reman} \left( t \right), $$
(7)

where, c = 1, …, C0; v ∈ {1, 2}.

The sustainable activity hierarchy [13] is used to find the most efficient recovery paths. Therefore, successive period's demand for each component type is tried to fulfill with the recovered components—first, with the available reuse components, and then with the available remanufacture components. Only when those two cannot fulfill the demand the new manufacture option will be used. Any excess after fulfilling the demand will be allocated to recycling as the next best recovery option.

Modeling the Product Sustainability Metrics: The calculation of sustainability performance is also done similarly to the prior literature [7, 8]. The basic idea is that the sustainability metrics of interest are calculated for each type of newly manufactured, reused, remanufactured, and recycled component for both Gen1 and Gen2. Therefore, when \(Ux_{c,v}^{activity}\) denote the unit value metric x (e.g., cost), for the component-type c, and variation v,

$$ \begin{gathered} Cost of production at t = N_{c,v}^{newmanu} \left( t \right){*}Ucost_{c,v}^{newmanu} + \hfill \\ \,\,\,\,\,N_{c,v}^{reuse} \left( t \right){*}Ucost_{c,v}^{reuse} + { }N_{c,v}^{reman} \left( t \right)*Ucost_{c,v}^{reman} + \hfill \\ \,\,\,\,\,N_{c,v}^{recyc} \left( t \right)*Ucost_{c,v}^{recyc} + N_{c,v}^{dispose} \left( t \right)*Ucost_{c,v}^{dispose} , \hfill \\ \end{gathered} $$
(8)

where, c = 1, …, C0; v ∈ {1, 2}

Similarly, the other relevant metrics can also be calculated. For economic metrics, when aggregating the per-period cost to calculate the total cost for the production, the value of money is discounted at a set rate. Its present value is calculated in reference to Gen-1 introduction time (i.e., period 1).

3 Classification Method and Discussion

We propose uavg/tintro as an initial indicator for classifying the suitability of a product for closed-loop production. Both uavg and tintro are typical PS parameters, ascertainable at design stage. The use of this indicator is presented employing the following hypothetical test case. The production simulation model detailed in Sect. 2 is utilized with the nominal values of Table 1. These values were selected to represent typical product's values. Since only nominal values were used, the discussion below focuses on the overall results trends rather than the magnitudes. The two representative metrics manufacturing cost and GHG emissions for the overall production are charted, taking uavg/tintro as the independent variable. These two metrics were selected due to the common usage and ease of calculation for this proof-of-concept study. However, more comprehensive sustainability metrics are necessary for an exhaustive evaluation.

Following assumptions are made to simplify and make the results more general. The component-types are assumed to be similar in terms of unit cost and unit GHG emissions so that their variability does not affect metrics. It also allows calculating Gen2's component-level IC (ICC) with Gen1 as a percentage of common components (Ccommon) in a product (i.e., ICC = Ccommon/C0). Compared to Gen1 component designs' nominal cost ($10) and GHG emissions (10 kgCO2) values, the new-design (i.e., Gen-2's non-common) components are assumed 10% costlier (due to production changes necessary) and 20% less GHG emitting (due to efficiency improvements in new technology). The same reusability, remanufacturability, and recyclability rates are assumed for all component-types. Below discussion uses per product averaged metric values of manufacturing cost (Cppavg) and GHG emissions (GHGppavg) to minimize the effect of number of products sold. For the discussion and visualization simplicity, these averages are for the combined production of Gen1 and Gen2.

Table 1. Parameter values used
Full size table
Fig. 2.
figure 2

Cost and GHG emissions metrics variation with uavg/tintro

Full size image

3.1 General Trends

To test if the indicator uavg/tintro can be used with different product categories, the simulation was run for different tintro values, and checked if the trends in metrics hold. Figure 2a & 2b illustrate the variation of Cppavg with uavg/tintro, for different tintro values when the ICC is 0% and 80%, respectively. For all tintro values, there is a common trend in average cost declining considerably for lower uavg/tintro ratios (< 1.25). It is understandable since lower uavg/tintro means a larger overlap between production and return curves, allowing greater parts harvesting for reuse/remanufacture. Furthermore, higher IC allows a lower cost over a greater range of uavg/tintro (e.g., ICC = 0% plateaus by uavg/tintro = 1.25, whereas ICC = 80% plateaus only after uavg/tintro = 2.0). It means higher IC allows a longer time window for manufacturer to repurpose EoU components in reuse/remanufacture. In Fig. 2c, GHGppavg also reveals a similar overall trend. Unlike GHGppavg curves, in Cppavg (Figs. 2a & 2b), each tintro curve plateaus to a distinct final cost level. It is explained by discounting of monetary value (i.e., cost) over time.

3.2 EoU Strategy and Returns Planning

Strategizing the EoU recovery and returns planning is critical to designing closed-loop and CE-focused products [13]. Consider a product where tintro is 12 months; according to Fig. 3a, if uavg is 24 months, the average per-product cost is $ 57.6; and if uavg is 18 months, it is $53.3. This is a 7.4% cost reduction if the EoU returns can be collected sooner (since some EoU products stay with customers for a long time after becoming EoU). This insight is beneficial for the manufacturer to plan product return incentives so that returns are more likely to happen when there is more opportunity for parts harvesting. It highlights the importance of considering not only how much EoU products are returned (i.e., recovery rate), but also at which time point. Consequently, product sustainability measurements must be treated dynamic (i.e., temporally-variable) [8].

Although per component GHG emissions are higher in the common design (since we assumed new component designs 20% more GHG efficient), Fig. 3b shows, from 0.5 < uavg/tintro < 1.5, there is still a potential to lower the overall average GHG emissions of production through increased reuse and remanufacture opportunities resulting from IC. Therefore, ICC and uavg/tintro indicators are vital factors for planning EoU recovery.

Fig. 3.
figure 3

Cost and emissions variation for different commonality levels (ICC = 80% and 0%)

Full size image

3.3 Planning Component Commonality with uavg/tintro

Figure 4 further illustrates the relative cost (Fig. 4a) and GHG emissions (Fig. 4b) difference between commonality level 80% vs 0%, for several tintro values. These charts show that the benefit of component commonality becomes highest during 0.25 < uavg/tintro < 2.0. For most tintro, the uavg/tintro = 1 provides the highest benefits. Notably, when uavg/tintro is small (say, < 0.25), the IC level matters less. This is because, in such cases, a considerable portion of the returns happens within the same generation's production. Thus, when a product's uavg/tintro ratio is very low, the manufacturer can focus more on devising designs with more efficient component designs (i.e., new designs) without being limited by the need to increase IC.

Fig. 4.
figure 4

Impact of increasing the inter-generational commonality (ICC) from 0% to 80%

Full size image

Overall, this case presents a preliminary view of the proposed classification approach. Additional work needs to be done with real industry case studies to validate these results. Furthermore, the preliminary work here is a reminder of the need to explore more complex strategies for closed-loop CE. Reconfigurable and cross-product fungible components can decouple the need for overlap between production and return flows of the same product's generations. For example, a module that is made common between multiple product categories can lead to parts harvesting over a longer time and more applications. However, the approach presented in this paper can also be extended to such cases by considering the module level aggregate demand.

4 Conclusions

For manufacturers and designers planning products for the CE, an objective indicator capable of providing a preliminary understanding regarding multi-generational PS's suitability for closed-loop production can be extremely useful. In this paper, we present the indicator uavg/tintro as a potential solution. By examining the variation of sustainability metrics relative to uavg/tintro, it was found to be usable with different tintro.

Based on the hypothetical example examined, a common trend showed that the overall multi-generation production's sustainability metrics (average cost and GHG emissions) show marked improvements in lower uavg/tintro ratios (significantly when uavg/tintro < 1.25). While uavg and tintro are typically set considering product and market characteristics, manufacturers can use the insights gained from the indicator to strategize EoU recovery and plan IC. For example, manufacturers can use the proposed indicator and the analysis when planning a return incentive scheme to influence uavg and obtain EoU products at the most opportune time (i.e., when there is more opportunity for parts harvesting). And, since the IC benefits are highest during a limited range (i.e., 0.25 < uavg/tintro < 2.0), manufacturer can either use the commonality level to expand that range, or focus on choosing more efficient component designs (i.e., new designs) without needing to make the generations more common. This work will be further extended by considering more variables and actual test cases to validate and develop a more robust indicator that factors additional product and PS characteristics.