Reman Co-Design: A Combined Design and Remanufacturing Optimization Framework for the Sustainable Design of High-Value Components

To balance between progress and sustainability, the idea of a circular economy has been emphasized in recent decades to drive the optimization of resources, reduce raw material consumption, and recover waste by recycling or repurposing. Unlike a linear economy that follows the “take-make-dispose” schema, a circular economy encourages repair, reuse, remanufacture, recycle, and recover (also known as Re-X options). These options keep resources in use as long as possible, extract maximum value during usage, and recover products and materials at the end of each service life [1]. Apart from optimizing maintenance and repair decisions to prolong the effective usage length [2–4], most studies have been conducted on the end-of-life and end-of-use decision-making to help generate another useful life effectively. Guo et al. [5] reviewed the existing modeling and optimization approaches in disassembly sequence planning, like graph-based modeling and matrix-based modeling, to achieve economic and time efficiency. With appropriately disassembled components, one of the Re-X options can be selected based on the component condition. Mishra et al. [6] proposed a stochastic programming model to optimize the Re-X options based on end-of-use conditions considering uncertain product demand. Recycling treatments have mainly been investigated for lithium-ion batteries and plastic package wastes to reduce energy consumption, improve recovery efficiency, and avoid safety issues [7,8]. Compared with recycling and reuse, remanufacturing offers better preservation of the value added to a product as it can restore end-of-life products to an “as-new” state [9,10].

The remanufacturing process often consists of disassembling, cleaning, inspecting, repairing, replacing, and reassembling the components of a product. There are also chances that new and better functionalities can be added to a used product during remanufacturing [11]. Existing studies [12–14] show that remanufacturing can reduce more than 50% of gas emissions and 25% of energy consumption compared with new product manufacturing. Although remanufacturing has been conducted in various product areas like automobile, heavy-duty equipment, machinery, and IT products and has critical driving forces like sustainability, it also has numerous challenges related to core (used product or its part) availability, timing, and quality [15]. To tackle the uncertainty in the core collection process, acquisition strategies are proposed to make the trade-off between cost and quality [16], and methods for acquisition control, return strategies, and return forecast are reviewed in Ref. [17]. To recover used products with various qualities back to an as-new condition, remanufacturing requires more than just repairing. Thus, nanotechnology, biotechnology, and advanced surface technology have been applied in remanufacturing processes [18]. Moreover, there is a growing emphasis on fostering collaborations among different participants within closed-loop supply chains [19].

Proper decision-making during a product’s early design stage can also facilitate remanufacturing and ease some of the challenges that remanufactured goods face [11,20]. Various rules and guidelines for the product design stage have been proposed to help better realize the true remanufacturing potential, which forms the basis of design for remanufacturing (DfRem). Yang et al. [21] developed a DfRem assessment tool by considering four major design aspects: material selection, material joining methods, structure design, and surface coating methods. With a review of published literature discussing DfRem rules, industry knowledge base, and pockets of detailed industry remanufacturing information, Hilton [22] summarized a set of descriptive design guidelines and integrated the guidance into a product designer’s toolkit.

In addition to qualitative rules, quantitative design changes and optimization efforts, considering the operations and end-of-use actions, have gradually gained attention in the literature. The design decisions including joint type, material processing, and product configuration are improved by evaluating the value-recovery efficiency at the end-of-use stage in Ref. [23]. Reliability model parameters of product components are optimized in Ref. [24] to achieve sustainability objectives considering warranty and circular practices. The architecture of a modular-based product family is optimized in Ref. [25] to allow the protection of intellectual property during the remanufacturing process. Kwak and Kim evaluated and optimized product design considering both pre-life and end-of-life stages in Refs. [26,27] by mainly using a transition matrix to capture the link between design alternatives and recovery scenarios. Joshi and Gupta [28] proposed a model to evaluate different designs of a product for the ease of disassembly and remanufacturing with the help of the sensors and radio frequency identification tags embedded in the products.

While these studies incorporate the consideration of remanufacturing-related actions into design optimization problems and vice versa, the focus remains solely on either optimizing the design decisions or the remanufacturing processes. Nonetheless, it is necessary that the product design and end-of-life recovery processes are considered simultaneously to fully realize the benefits of remanufacturing [29]. This is somehow analogous to the challenge faced in the design of integrated dynamical engineering systems, where the plant design is strongly connected to the control design and vice versa. This issue in the domain of systems design was effectively addressed via the combined plant and control design (control co-design) formulation, where the plant and control decisions are optimized simultaneously. By considering all the interactions in one design formulation, control co-design is generally able to find the superior optimal decisions that could not be realized otherwise [30–33].

The design and remanufacturing decisions for a product in DfRem are in some way akin to the plant design and control decisions in dynamic systems. Using this similarity and the concept of simultaneous design and control optimization from control co-design, we propose a new combined design and remanufacturing optimization (reman co-design) formulation, where the product design and its remanufacturing decisions are optimized simultaneously. This new formulation aims to address all the interactions between the two domains in one optimization problem to maximize the benefits of remanufacturing throughout the entire lifespan of a product. It should be noted that the proposed formulation can only address the design and remanufacturing interactions that are considered/modeled within the optimization. Therefore, the effectiveness of reman co-design is highly dependent on the level of design and remanufacturing details provided.

This work is a continuation of the previous study [34] on DfRem and offers several new contributions that are mentioned below:

1. We propose a new DfRem formulation, referred to as combined design and remanufacturing optimization (reman co-design), that optimizes the design and remanufacturing decisions for the entirety of a product’s lifespan simultaneously. This new formulation is developed using the design concepts from control co-design [30,31].

2. The new framework explores the whole feasible product and remanufacturing design space. This is different from and a step above the previous work [34], where only a handful of designs and remanufacturing options were considered and analyzed.

3. To showcase the benefits of reman co-design, we compare its results to the ones from an approach where the remanufacturing options are optimized for a fixed product design. The latter approach resembles a decoupled DfRem approach where remanufacturing is designed after the product is fully realized. We refer to this design methodology as “reman design” in this work.

4. Additionally, the trade-off analysis in the previous work was returned state (damage level)-unaware and life cycle-independent, i.e., the remanufacturing actions in the design options were assumed to be taken regardless of the damage level of the returned parts and their life cycle. In this work, the reman co-design and reman design formulations are both life cycle dependent. In addition, we develop the damage-aware (DA) and -unaware (DU) variants of these formulations to investigate the potential benefits that a damage-aware formulation might provide.

5. Finally, a hydraulic manifold, (re)manufactured by John Deere, is used as a case study to showcase the utility of the proposed reman co-design formulation. First, we identify the most crucial failure mode of this manifold during its remanufacturing by performing a new remanufacturing failure modes and effects analysis (RFMEA). Thereafter, we use the proposed framework to optimize the design and remanufacturing decisions suggested for that failure mode to maximize the benefits of remanufacturing for this part.

In summary, the major contributions of this work compared to the previous study [34] are as follows:

1. The proposed formulation in this work explores the whole feasible design space in order to find the optimal solutions.

2. The new reman co-design formulation is life cycle and damage level dependent, i.e., the remanufacturing decisions are optimized for every given life cycle and damage level.

3. A new RFMEA framework for the analysis of the failure modes has been developed that helps identify the most critical failure modes.

The rest of this paper is organized as follows. In Sec. 2, we present the general formulation of reman co-design and provide the details on its damage-aware and -unaware variants. Then, we introduce the case study of the John Deere hydraulic manifold and develop the specific reman co-design optimization model with the insights from RFMEA in Sec. 3. Next, in Sec. 4, we reflect on and discuss the numerical results from the case study. Finally, in Sec. 5, we provide the concluding remarks and potential future work directions.

The proposed reman co-design formulation can be state-aware or -unaware. In the former variant, the reman action variables are state-dependent u = u(k, x(k)), and consequently, the reman action variables are optimized for every possible state of the returned parts. In the latter form, the reman action variables are state-independent u = u(k); therefore, these actions are applied to all of the returned parts regardless of their damage severity levels or failure modes. For the sake of completeness, we are also including g(·) vector which denotes any inequality constraints that might be included in a DfRem problem. The initial design and reman actions generally affect the state of the returned parts, i.e., some initial designs and reman actions taken might lead to parts with more resilience to wear and degradation during their operation; this relationship is presented via x = f_x(·) in the above problem formulation. Finally, n_rem is the total number of the reman cycles. Note that from now on, we refer to “damage severity levels” as “damage levels” for the sake of brevity.

In this section, we illustrate how the proposed reman co-design framework can help better achieve remanufacturing success with a hydraulic manifold case study.

The infinitely variable transmission system (re)manufactured by John Deere includes a hydraulic manifold assembly that regulates the fluid flow among various other components of the transmission. Note that we only focus on the hydraulic manifold assembly as our case study in this work. The hydraulic manifold assembly is primarily made of aluminum and is manufactured by die casting followed by machining/finishing processes. Figure 2 illustrates the cross-sectional and side views of the manifold assembly, along with its location in a John Deere tractor. During the current remanufacturing process of used hydraulic manifold assemblies, after the disassembly process, manifold housings (the major part of the hydraulic manifold assemblies) passing the initial inspection will be cleaned and then reassembled for reliability tests. Since basic cleaning does not provide effective repair to the used manifold housings, only as-new manifold housings can be reused and the reuse rate will decrease with remanufacturing cycles. Thus, the original equipment manufacturer (OEM) has a critical need to explore enhancements in the remanufacturing process to attain better sustainability outcomes across multiple remanufacturing cycles. Through close collaboration with the design and remanufacturing engineers at the OEM, we collect their engineering expertise and manifold housing data spanning from the design phase to the end-of-use stage. This allows us to adopt our proposed reman co-design framework to this specific case study.

To gain insight into the reasons why some of the hydraulic manifold assemblies are not reused during remanufacturing, and to identify the most crucial failure modes for the remanufacturing process of this part, we are using a slightly modified version of failure modes and effects analysis (FMEA) [44,45] in this work. In the next section, we provide a detailed description of the modified analysis, which we refer to as reman FMEA or RFMEA.

RFMEA is different from the traditional, industry-standard FMEA in that, aside from identifying the failure modes, their severity (SEV), and occurrence (OCC) levels, it also includes possible design/repair changes to address each failure mode during remanufacturing. Our new RFMEA framework includes one new index score related to the feasibility (FEA) of the potential design/repair changes for each failure mode. This index shows how feasible, both economically and technologically, the potential design/repair changes would be at the OEM’s facilities. The risk priority number (RPN) is the product of the SEV, OCC, and FEA indices. RPN here is used as an index to rank the failure modes based on their overall severity, occurrence, and feasibility of potential solutions. Failure modes with higher RPN values in RFMEA are considered the most critical failure modes for enhancing the remanufacturing process. In the following, we provide a summary of the remanufacturing process for the hydraulic manifold assembly and discuss briefly its major failure modes and their potential design/repair changes that were provided to us by the OEM.

When a transmission fails and is received at the OEM’s facility, the hydraulic manifold assembly is detached from the transmission. The process for the different parts of the hydraulic manifold assembly is as follows:

• Manifold housing: The main component of the hydraulic manifold assembly is the manifold housing. After removing all the bolts and bearings from the manifold housing, it proceeds to the removal of its gasket. Depending on the level of gasket adherence to the part, a variety of increasingly aggressive processes may be used for removal. As the aggressiveness of these processes increases, the opportunity for damage does as well. After all traces of the gasket are completely removed, the gasket area is inspected and if it passes the inspections (in terms of dents and scratches), it will proceed for further cleaning and inspections. If all the inspections are passed, the manifold will be reused, otherwise, it will be scrapped.

• Wear components: The OEM never reuses the wear components, such as the gaskets, O-rings, bearings, washers, etc. When these parts are disassembled from the assembly, they are scrapped.

Based on our communications with the OEM’s engineers, we have identified three major failure modes for the manifold housing. Here, we define these failure modes and analyze their severity and occurrence levels. Additionally, as a part of the new RFMEA, we explore the potential design and repair modifications that could address these failure modes. Based on the comments from the OEM’s engineers, we also provide a ranking for the feasibility of each design change. The major failure modes for the manifold housing are as follows:

• Deep dents and scratches: The gasket surface of the manifold housing is prone to dents and scratches during the gasket removal process. Also, some dents and scratches can occur on the gasket surface due to improper handling and transportation. A sample view of such dents and scratches is shown in Fig. 3. The reuse of manifold housings with this failure mode can cause a “moderate” level of decline in the hydraulic performance of the manifold assembly. This failure mode is identified as “very probable” for its occurrence level. The OEM does not repair this failure mode and scraps the manifold housings with this type of damage. One of the potential design/repair options to address this failure mode is to design these manifolds with a thicker gasket surface such that dents can be machined off during remanufacturing. The alternative option is to use additive manufacturing to repair the gasket surface. Improving the handling process at the OEM’s facility is also considered for preventing this failure mode. The recommended actions to address this failure mode are the same as the potential design/repair changes since these options have a “moderate” level of feasibility.

• Spool bore damage: This failure mode is caused by a failed transmission and its floating debris in the transmission fluid. These particles can cause pitting and wear on the spool bore valve. This issue can lead to “high” levels of transmission fluid contamination and assembly misalignment problems. This failure mode is also ranked as “very probable,” and is not repaired. The potential design change would be to change the cartridge type of the spool bore or to hard anodize the part. The OEM’s engineers rank the feasibility of this design change as “low,” and the recommended action remains as scrapping.

• Valve housing damage: Over-torquing during the diagnosis stage can cause damage to the valve housing’s threads. Corrosions and cavitations can also trigger this failure mode. If this damage is left unaddressed, it can cause unintentional clutch pressures and leakages at a “moderate” level. This failure mode is also ranked as “very probable” and is not repaired by the OEM. One potential design change to address this failure mode is to reinforce the valve housing threads, but this design change has a “low” level of feasibility; therefore, the recommended action is scrapping the damaged manifolds.

Henceforth, in this work, we only focus on this particular failure mode to leverage its design options in order to improve the remanufacturing process. Since it is a challenging task to quantify “handling process improvements” for the “dents and scratches” failure mode, we are only focusing on the remaining two design options for this failure mode. The first option is to have a thicker gasket area in the initial design of the manifold housing to allow machining during remanufacturing, and the second option is to perform additive manufacturing to recover the damaged gasket surface area. Note that the abovementioned options are not mutually exclusive and a hybrid approach can also be used. For instance, for a given gasket surface thickness, a manifold housing might go through “machining” for its first two remanufacturing cycles and an additive manufacturing process for its third remanufacturing cycle. Identifying the optimal gasket surface thickness and the optimal combination of the remanufacturing decisions is not an easy task and requires rigorous analysis as there is an intricate interplay between the two domains. On one hand, having a thicker gasket area escalates the costs and environmental footprint of the initial design, and on the other hand, it allows the OEM to perform the more cost-effective operation of machining rather than additive manufacturing.

In the next section, we use our proposed reman co-design formulation to find the optimal design and remanufacturing decisions for this particular failure mode. This will help us assess the utility of our proposed reman co-design formulation when applied to a real-world DfRem problem.

The design and remanufacturing decisions are evaluated for the entire lifespan of the part and in terms of costs, energy consumption, and CO₂ emissions. The formulations to calculate these three performance metrics for different actions are summarized in Table 1. We maintain continuity by carrying over the assumptions, material specifications, and processing data from the previous study, where the parameters/data for this case study were obtained from the OEM. Note that the costs, energy consumption, and CO₂ emissions for manufacturing are dependent on the design decision d and those for machining are dependent on the dent/scratch depth. The reman co-design problem formulation for this case study is solved using the exhaustive search algorithm shown in Algorithm 1. We adopt a sample size of N_s to evaluate the expected costs, energy consumption, and CO₂ emissions for all the possible combinations of design d and remanufacturing [u(1), u(2), u(3)] decisions. Thereafter, we obtain the Pareto set that optimizes the objective functions. For each sample, at each remanufacturing decision point, we generate a random dent/scratch depth from the fitted distribution. There is a probability of (1 − p_d/s)(1 − r_rate,k) that the sample suffers from other failure modes which cannot be solved by the gasket area design change. In this scenario, the actual remanufacturing action becomes u′(k) = S, which stands for the “Scrap” action. Besides, if the planned action is u′(k) = B, basic cleaning, but the dent/scratch depth X(k) is larger than the threshold θ, the actual action becomes u′(k) = S. Similarly, if the planned action is u(k) = M, machining, but the remaining machining stock allowance is smaller than the dent/scratch depth X(k), the actual action becomes u′(k) = S. Overall, instead of providing constraints to limit the reman action selection, we replace the actual action with scrapping if the selected action is infeasible to perform. Then, if the actual action is machining, the machining stock allowance will be reduced by the dent/scratch depth. With the costs, energy consumption, and CO₂ emissions calculated for each sample throughout its lifespan, the expected objective values for each combination of design and remanufacturing actions can be derived, which are then used for obtaining the Pareto set.

We also implement a damage-aware variant of the reman co-design problem that optimizes the remanufacturing actions for each cycle and each damage level. We categorize the manifold housings’ conditions into four damage levels as shown in Table 2, where m_k is the remaining machining stock allowance at reman cycle k. The assessment of damage levels is determined by two key factors: the reusability of the manifold housing after undergoing the gasket removal process, and the feasibility of employing machining as the current remanufacturing decision. Therefore, in the damage-aware problem, at each remanufacturing cycle k, instead of selecting one optimal remanufacturing action u(k), we need to select the best action u(k, l) for each damage level l. We follow the same optimization workflow as shown in Algorithm 1 for the damage-aware problem variant. Note that the parameters used in this case study are listed in Appendix  A.

In this section, we present the numerical results derived from implementing the proposed reman co-design framework on the manifold housing case study. Additionally, we explore various scenarios and discuss the potential savings in terms of costs, energy consumption, and CO₂ emissions that can be achieved using our proposed formulation. All the reman co-design problem formulations have been solved using the python scripting language and on a machine equipped with an Intel(R) Core(TM) i5-10505 CPU @ 3.20 GHz. The parameters for the reman co-design problem formulations and their reliability analyses are mentioned in Table 3.

In this study, we establish the baseline for our investigations by considering the current design and remanufacturing decisions of the OEM. The current practice for the manifold housing is to not have any machining stock allowance and to perform basic cleaning without any major repairs at all of the remanufacturing cycles (part’s life cycles). Figure 5 showcases the breakdown of costs, energy consumption, and CO₂ emissions associated with this baseline practice and for each life cycle of the part. To facilitate our comparisons and to protect the OEM’s data, we normalize the results and present them as percentages relative to those from manufacturing a new, base manifold housing (with no design changes).

We first adopt the damage-unaware variant of reman co-design to jointly optimize the design decision and remanufacturing actions at three reman cycles. Then, to provide adaptive remanufacturing actions for manifold housings at different damage levels, we apply the damage-aware variant to the case study and analyze the resulting life cycle performance as well. The optimal objective values are summarized in Fig. 6. As mentioned in Algorithm 1, we assign equal importance to the three performance metrics and generate a Pareto set that encompasses optimal decisions. With the current parameter settings, we obtain one Pareto optimal solution, which is labeled as RemCoD (reman co-design) P#1 in Fig. 6. Results noted by RemD (reman design) P#1 are the Pareto optimal solutions that we obtained when optimizing the remanufacturing decisions only. In this case, the design is fixed to have no machining stock allowance, resembling the current design of the manifold housing. For the damage-unaware scenario, the life cycle performance with respect to costs, energy consumption, and CO₂ emissions can be improved by 20%, 30%, and 29%, respectively, from the current practice when using the reman design optimization process. Using reman co-design also leads to 9%, 6%, and 6% additional improvements from the reman design approach for the metrics, respectively. Additionally, note that the three objective values increase along remanufacturing cycles due to the updated dent/scratch generation distribution reflecting decreased reuse rates. Thus, at a later remanufacturing cycle, more of the returned parts need to be scrapped, which leads to higher costs, energy consumption, and CO₂ emissions. Further life cycle performance enhancement is achieved using the damage-aware formulation; however, we observe minimal improvements compared with the damage-unaware results. This outcome can be attributed to the concentrated damage level in this specific case study, as elaborated in greater detail in Sec. 4.3.

Here we deep dive into the optimized actions and manifold housing condition evolvement over remanufacturing cycles and derive insights from the optimization process. Solutions in part (a) of Fig. 7 are the optimized remanufacturing actions with the current design without additional machining stock allowance. Generally, when machining is not feasible, additive manufacturing is a better choice than basic cleaning as it can recover the condition of used manifold housings to avoid scrapping the returned cores. Since no machining stock allowance is added, damage levels 1 and 3 cannot be observed. The benefit of damage-aware reflects in the last remanufacturing decision point where basic cleaning and additive manufacturing can be selected for damage level 2 and damage level 4 correspondingly. Solutions in part (b) of Fig. 7 are the optimized design and remanufacturing decisions from the reman co-design process. Since machining leads to an even lower reuse rate reduction according to our assumption in Eq. (5), a maximum amount of machining stock allowance has been allocated to the gasket surface of the manifold housing to ensure machining feasibility. Consequently, machining is selected as the optimal decision at all remanufacturing cycles for the damage-unaware scenario. However, when the return rate changes, we may not choose to add the maximum amount of machining stock allowance, which is explained in more detail in Sec. 4.4. Similar to part (a), the benefit of damage-aware reflects in the last remanufacturing decision point. Because the last decision only needs to minimize objective metrics without caring about reuse rate reductions, the optimal action for damage level 1 is basic cleaning in the last reman cycle. Due to the significant advantage of additive manufacturing and machining in recovering the manifold housing condition compared with basic cleaning, basic cleaning is not selected for damage level 1 or 2 as the optimal action in the damage-aware variant for the first two cycles.

In both part (a) and part (b), the consistency in optimal reman action decisions from the damage-aware and -unaware formulations for the first two cycles makes the additional improvement from damage-aware optimization minimal. However, in other applications, when the condition-recovery effect of performing different remanufacturing actions is comparable, the distinction between damage-aware and damage-unaware variants will become more noticeable.

The previous analyses are conducted under the assumption of a 100% core return rate, although in actual remanufacturing processes, a lower core return rate may be observed. Consequently, we now delve into investigating the impact of core return rate on two specific matters: the benefits derived from remanufacturing and the optimal design and remanufacturing decisions. The results will aid the OEM in striking a balance between the investment in core return incentives and the resultant reductions in life cycle cost and environmental impacts. Note that the core return rate refers to the ratio of the returned parts for remanufacturing to the total number of manufactured parts. This is different from the reuse rate which refers to the portion of the returned parts for remanufacturing that are reusable.

In this section, we have calculated the total costs, energy consumption, and CO₂ emissions over the entire life cycle of the manifold housings for three different scenarios and different core return rates. The first scenario is when the OEM does not repair any of the returned manifolds and replaces them with new parts; we refer to this scenario as “all-new” in Fig. 8. In the second scenario, the abovementioned metrics are evaluated for the current design and remanufacturing practice of the OEM, which corresponds to having no machining stock allowance and performing basic cleaning only. The results associated with this scenario are labeled as “current” in Fig. 8. In the third scenario, we have used the damage-unaware reman co-design formulation to optimize the design and remanufacturing decisions for different core return rates. These results are denoted by the “RemCoD” label in Fig. 8. We use the damaged-unaware variant of reman co-design here as, based on the previous analysis, the damage-aware and -unaware variants of this formulation yield highly similar optimal metrics in this case study. Note that here, in order to find the optimal solutions, we have assigned equal weights to the metrics. It can be seen that at any given core return rate, the “all-new” scenario results in the highest costs, energy consumption, and CO₂ emissions. The “current” remanufacturing practice yields lower metrics than the “all-new” scenario; however, as expected, its benefits diminish as the return rate decreases. The RemCoD results follow the same trend, nonetheless, they generate the lowest metrics at any given core return rate when compared to the other two approaches. The second row of Fig. 8 illustrates the percentage reductions in costs, energy consumption, and CO₂ emissions that can be achieved using the RemCoD method. These reductions are from the “all-new” and “current” scenarios. When modifying the weights assigned to costs, energy consumption, and CO₂ emissions to 0.5, 0.1, and 0.4, respectively, as suggested by the OEM, the impact on the established remanufacturing benefits is minimal; however, we do observe changes in the optimized decisions which are explained next by comparing the results in Figs. 9 and 10.

To showcase the effect of the core return rate on the optimal decisions resulting from the reman co-design formulation, we are using Figs. 9 and 10, where the optimal machining stock allowance and remanufacturing actions have been presented for different core return rates. The results in Fig. 9 are from the case where the weights of the performance metrics are set as equal, and the results in Fig. 10 are from the case where the weights of 0.5, 0.1, and 0.4 are set for costs, energy consumption, and CO₂ emissions, respectively. The weights in the latter are suggested by the OEM as they reflect the priorities of the OEM for this specific case study. We have conducted this analysis for two scenarios to further investigate the effects of metric weightings on the optimal solutions. The foremost crucial aspect regarding these results is that different core return rate values result in different optimal design and remanufacturing decisions. Specifically, in the equal-weight scenario, for core return rates ≤ 10%, the optimal design is to have no machining stock allowance and to perform additive manufacturing to repair the damaged manifolds. For the core return rate of 15%, the optimal machining stock allowance has been identified as 0.762 mm and the optimal remanufacturing actions have been found to be machining for the first remanufacturing cycle and then performing additive manufacturing for the consequent cycles. On the other hand, the optimal machining stock allowance for core return rates ≥ 20% is given as 2.286 mm, and the optimal remanufacturing action is to machine the dents off at all of the remanufacturing cycles. Note that the machining process for remanufacturing is economically and environmentally more attractive than additive manufacturing; however, in order to be able to use the former, the initial design must accommodate, i.e., it should have an additional machining stock allowance. Nonetheless, the additional machining stock allowance escalates the costs, energy consumption, and CO₂ emissions of the initial design. Our findings in this section indicate that investment in the additional machining stock allowance is economically and environmentally viable only if the core return rate is relatively high, i.e., the ratio of remanufactured parts to the total number of manufactured parts surpasses a certain threshold. This threshold is the point where the economic and environmental benefits gained from sole machining, as opposed to additive manufacturing, outweigh the initial costs incurred due to the additional machining stock allowance. In the equal-weight scenario, this investment starts to seem viable for core return rates ≥15%. This aligns well with the prevailing knowledge that opting to invest in design for remanufacturing is financially and environmentally sound under the condition that the number of remanufactured parts is a significant portion of the total number of manufactured parts. When different weights are assigned to the three objectives, we notice that the aforementioned threshold changes to 10%, and an optimal machining stock allowance of 1.524 mm is selected for core return rate of 15%. By incorporating the OEM’s suggested weights, the focus is shifted toward costs and CO₂ emissions. The updated design decisions at core return rates of 10% and 15% indicate that investing in the initial design can better achieve the reductions in costs and CO₂ emissions. The steady remanufacturing benefits and different optimized decisions obtained from two “weighting” scenarios indicate that the proposed reman co-design model can help adjust the design and remanufacturing decisions under different core return rate situations to realize remanufacturing benefits by aligning them with OEM’s distinct focus on performance metrics.

This article focuses on the crucial interactions between product design and remanufacturing decisions and proposes a novel DfRem framework called combined design and remanufacturing optimization, or reman co-design. With this framework, the optimization of design changes and remanufacturing actions becomes a unified process, aimed at addressing existing remanufacturing failure modes and achieving optimal outcomes in terms of costs, energy consumption, and CO₂ emissions over the entire product lifespan. To account for the inherent uncertainty in the condition of returned cores and to ensure the robustness of the optimal decisions, a probabilistic formulation of reman co-design is also established. This framework identifies the optimal remanufacturing decisions for each cycle and enables the consideration of both damage-aware and damage-unaware remanufacturing decisions.

We illustrate the efficacy of the proposed framework with a case study of the hydraulic manifold used in John Deere’s infinitely variable transmissions. We first investigated the failure modes observed in the remanufacturing process of this part and engaged in discussions with engineers at the OEM. This collaborative effort led us to identify potential design and remanufacturing action improvements for this case study. It is important to note that a new RFMEA was developed for this case study which provided us with an appropriate direction for the life cycle decision improvement. Then, we formulated the specific reman co-design problem for this case study to identify the optimal design changes and corresponding remanufacturing actions. In comparison to the current practice of the OEM, the optimized design and remanufacturing decisions resulted in remarkable reductions of 30–36% in costs, energy consumption, and CO₂ emissions. This level of improvement cannot be achieved by solely optimizing either the design or remanufacturing process. Lastly, we explored the impact of core return rate on sustainability improvement resulting from the optimized design and remanufacturing decisions. The quantification of costs, energy consumption, and CO₂ emissions savings in our study could potentially assist the OEM in making informed decisions regarding core return incentives. Additionally, we observe that the design decision regarding machining stock allowance adjusts according to the core return rate. This adjustment reflects the trade-off between the initial design investment and the remanufacturing benefits derived from applying the machining process to the returned cores.

In the case study examined in this work, the focus was on addressing a critical failure mode observed during the remanufacturing process by applying our reman co-design framework. While in future applications, a larger design change space can be explored to tackle multiple failure modes. However, the current solution methodology, exhaustive search combined with crude Monte Carlo simulations, might become computationally expensive and even intractable for larger problems. Therefore, more efficient optimization algorithms, such as genetic algorithm or particle swarm optimization, combined with more efficient sampling methods, such as importance sampling, should be considered. Additionally, design validation processes (e.g., tolerance stack-up analyses or computational fluid dynamics) can be embedded in the framework to ensure the functionality and reliability of products after design changes and remanufacturing activities. Moreover, the current product degradation process (e.g., dent/scratch generation) is derived inversely from the reuse rate records. To further improve this, a degradation process model that reflects the impact of design decisions and remanufacturing actions can be developed to establish a stronger connection between the design phase and subsequent remanufacturing stages.

This material is based upon work supported by the U.S. Department of Energy’s Office of Energy Efficiency and Renewable Energy (EERE) under the Advanced Manufacturing Office Award Number DE-EE0007897 awarded to the REMADE Institute, a division of Sustainable Manufacturing Innovation Alliance Corp.

The authors of this paper would also like to express their sincere gratitude to Rick Lopez, Corey Smith, Kevin Bishop, and Navaid Ahmed at Deere & Company for their immense guidance, unwavering support, and invaluable assistance throughout this work.

This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

There are no conflicts of interest.

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

k =

remanufacturing cycle number

d =

design variables

g =

inequality constraints in reman co-design

u =

remanufacturing actions

x =

state of returned parts

A =

additive manufacturing in remanufacturing

B =

basic cleaning in remanufacturing

$E$ =

expected value operator

M =

machining in remanufacturing

S =

scrapping in remanufacturing

C =

random cost

G =

random inequality constraints in reman co-design

X =

random state of returned parts

m_k =

remaining machining stock allowance on the gasket surface at reman cycle k

n_rem =

number of reman cycles

p_d/s =

portion of the returned parts that suffer from dents and scratches after cleaning

r_rate,k =

reuse rate at reman cycle k

f_x =

state prediction function

F_LN =

cumulative density function of lognormal distribution

N_s =

sample size for Monte Carlo simulation

R^target =

reliability target

Pr =

probability operator

EN =

random energy consumption

EM =

random CO₂ emissions

α =

reuse rate reduction percentage

η =

shape parameter for lognormal distribution

θ =

maximum dents/scratches threshold

λ =

scale parameter for lognormal distribution

ψ =

objective function of deterministic reman co-design

Ψ =

objective function of probabilistic reman co-design

The parameters used in the hydraulic manifold case study are listed in Fig. 11.