Research on Optimizing Passive Filter Weight in Automotive Applications While Maintaining Performance

Passive filters are widely used in high and low-voltage electricity to remove unwanted frequencies in power supplies, and communication devices, ensuring EMI-free operation which stands for an operation without Electromagnetic Interference. These filter designs protect sensitive equipment from being damaged by high levels of electrical noise. Despite their effectiveness, passive filters are typically heavy and expensive. We want to design a new configuration of passive filter that optimizes the weight and maintains performance levels. By having a quasi-3D simplified model capable of importing finite element mesh files, we categorize the mesh elements corresponding to ferromagnets and air with labels '1' and '0', respectively. This approach is known as topology optimization and draws inspiration from the Ising model in statistical mechanics, where the goal is to achieve equilibrium by minimizing the system's energy, governed by the Hamiltonian operator. By utilizing this framework, we aim to identify the minimum energy configuration, analogous to the behavior observed in ferromagnetic systems within the Ising model. The aim is to maximize the number of zeros, corresponding to mesh elements made of air, thereby reducing the amount of ferromagnets used to decrease weight and cost. This approach is a combinatorial optimization which is a method used to find the optimal solution from a finite set of discrete solutions. This problem belongs to the NP-hard class in complexity computation and does not have an effective polynomial-time solution. In this work, the problem is approached using three distinct AI-based methods: a Random Agent, a Simulated Annealing Agent, and a Simulated Annealing Agent with tunneling. The Random Agent serves as a baseline model, making decisions randomly without considering energy minimization or the state of the system. While simplistic, it provides a reference for evaluating the effectiveness of more advanced strategies. The Simulated Annealing Agent, inspired by the physical annealing process, iteratively explores the solution space by allowing occasional increases in energy, thereby avoiding local minima and gradually converging to a low-energy configuration. This method aligns closely with the behavior of systems seeking equilibrium, as modeled in the Ising framework. Finally, we extend the Simulated Annealing approach by introducing quantum tunneling, a mechanism that enables the agent to overcome high energy barriers by "tunneling" through them, rather than escaping slowly via probabilistic jumps. This allows the system to better explore complex energy landscapes and increase the likelihood of reaching the global energy minimum. By comparing these methods, we aim to identify the most effective strategy for minimizing the system's energy and achieving equilibrium, in line with the Hamiltonian-inspired behavior of the Ising model. As an application of the combinatorial optimization inspired by quantum processes, we include a combinatorial optimization example with the knapsack problem, which is intended to be solved on a real quantum computer. To approach this, we map the knapsack Hamiltonian to the Ising model and then convert it into a Quadratic Unconstrained Binary Optimization (QUBO) format, utilizing the QuadraticProgramToQubo converter from Qiskit. This approach will help to explore possible solutions that minimize the device's weight with minimal performance loss.