*3.2. Disassembly Optimizer*

The first step in the disassembly strategy optimizer is selecting the most suitable optimization method depending on the available data and the objectives of the disassembly (see Figure 4). The technique chosen will then be applied to calculate an optimal disassembly strategy. Disassembly sequence planning (DSP) is a non-deterministic polynomial (NP) problem [25]. Here, the solution space is huge, especially for large products such as EVBs. Furthermore, the solution space becomes larger when further decisions have to be made when planning the disassembly strategy, such as the disassembly depth and the circular economy strategy at the component level. That is why primarily nature-inspired heuristic optimization methods are used in the literature to solve the DSP problem, such as genetic algorithms, particle swarm optimization, ant colony optimization, scatter search, and artificial bee colony optimization [25]. In this work, we will focus on the use of a modified genetic algorithm, since genetic algorithms are the most widely used optimization method for finding optimal disassembly strategies [25]. Furthermore, they offer multiple advantages compared to other metaheuristic methods, such as a wide application range, strong expansibility, and high robustness, since they usually do not fall into local optimal solutions [12]. In addition, genetic algorithms are attractive because, first, they quickly and cost-effectively produce high accuracy solutions, even when the solution space is huge, and second, they are easy to understand and implement since simple mathematics is involved [27].

Figure 4 shows the structure of the disassembly strategy optimizer designed in this paper. Here, the steps of the implemented genetic algorithm are shown in detail. It starts with generating the initial population of potential disassembly strategies coded in chromosomes. The chromosome structure will be described in the following subsection. Subsequently, the individuals of the first population are evaluated using an objective function, which can consist of different sub-objectives. In this context, Alfaro-Algaba and Ramirez [14] proposed a combined objective function composed of economic and environmental subobjectives to maximize the economic profit while minimizing the environmental impact during the disassembly process of EVBs. A lot of data at the component level are needed, such as the disassembly costs, the costs to recondition disassembled components in order to implement a selected circular economy strategy, and environmental data.

Next, the selection step takes place to find out the fittest chromosomes to build a mating pool. The subsequent step is the mutation phase. Here, it should be ensured that all chromosome sections have the opportunity to mutate in order to increase the chances of discovering new solutions with higher performance.

The steps of the genetic algorithm are then performed until a termination condition is satisfied, such as a predefined number of generations or the fulfillment of specified convergence criteria.

In the following subsections, we describe our methodology for the different steps of the implemented genetic algorithm to optimize disassembly strategies for EVBs in terms of the disassembly sequence, disassembly depth, and circular economy strategies at the component level.

#### 3.2.1. Generating the Initial Population

The initial population consists of feasible disassembly strategies coded into chromosomes. Thereby, one chromosome consists of three sections. The first section represents a disassembly sequence, taking into account the precedence constraints. In the second section, all battery components are assigned a circular economy strategy dependent on their condition. The third section consists of only one gene, representing the disassembly depth.
