Mechanical interface optimization of printed circuit boards using multi-objective genetic algorithm

Abstract

Over the decades, numerous technological advancements have emerged to address the evolving needs of humanity. To satisfy these increasing demands, technological devices have been required to become faster, smaller, and more powerful. These requirements have introduced various engineering challenges across different domains. One of the most significant challenges is the thermal management of high-power hardware within acceptable temperature limits. The aerospace industry represents one of the most demanding fields where this thermal control problem is most severe. Considering the substantial costs associated with launching even the smallest components into space, the need for higher processing capacities within smaller volumes becomes increasingly evident. However, the high processing density required in such confined volumes leads to excessive heat dissipation per volume. In the harsh environment of space, where there is no medium to facilitate heat transfer via convection, the effective dissipation of high heat loads remains a critical problem. Hence, optimizing the thermal performance of any printed circuit board (PCB) designed for space missions is essential. Another formidable challenge for electronic hardware is the high shock and vibration loads encountered during rocket launch and pyroshock inducing space applications. Any structure involved in a space mission requires a launch vehicle to reach outer space. Extremely powerful rockets are employed to overcome the Earth's gravitational force. Due to this immense propulsion power, the payload is subjected to destructive shock and vibration loads. Any plastic deformation or structural failure occurring under such mechanical stress can jeopardize the overall success of the mission. Even if the mechanical structure is designed with sufficient strength to prevent catastrophic failure, even a minute micro-crack in the solder joints of components within the onboard equipment can disrupt signal or power transmission, leading to mission failure. These hazardous structural loads are encountered not only during the launch phase but also within the space environment itself. Pyrotechnic mechanisms used for the deployment of solar panels or the separation from the launch vehicle can also induce similar high-magnitude shock loads. Therfore, similar to thermal performance, a circuit board must also be structurally optimized to withstand these intense mechanical stresses. While it is possible for an analysis and design engineer to achieve these multi-disciplinary optimizations to a sufficient degree, the process is highly time-consuming. Therefore, this multi-objective optimization problem can be addressed by using a computer-based automation algorithm. The optimization required for a printed circuit board within space mission equipment to withstand the thermal and structural challenges imposed by harsh space environments is predominantly dependent on the positioning of the PCB's mechanical interfaces. Strategically designed mounting interfaces facilitate the effective transfer of excess heat dissipated by components to the mechanical chassis without causing significant temperature gradients. Simultaneously, these interfaces can dampen the structural loads transmitted to component solder joints, thereby preventing the formation of micro-cracks. To optimize the spatial distribution of these mounting interfaces on the PCB, genetic algorithms (GA), a widely utilized search and optimization method in this field, have been preferred. Genetic algorithm is an evolutionary computation technique that mimics the principles of biological evolution to solve complex search and optimization problems. By utilizing evolutionary sub-mechanisms such as selection, mutation, and crossover, it aims to derive the most suitable design for a defined objective function from an initial set of stochastic designs. The desired optimization is achieved by encoding the design parameters of the problem into an appropriate genetic representation and iteratively processing them through these aforementioned sub-mechanisms. In this study, the positions of the mechanical interfaces on the circuit board were selected as the design parameters. A population consisting of diverse board designs was generated by randomly assigning a predetermined number of mounting interfaces to the feasible regions on the board. Subsequently, successive generations were derived from this population by evaluating their fitness relative to an objective function aimed at minimizing component temperatures and displacements. In addition to the selection mechanism based on objective function evaluation, genetic diversity necessary for optimization was maintained through mutation and crossover methods during generational evolution. The mutation principle was implemented by applying random, minor movements to the positions of mechanical interfaces on randomly chosen boards. Furthermore, the crossover mechanism was utilized to produce offspring designs by combining an equal number of mounting interfaces from two randomly selected parent board designs. To prevent the accidental loss of high-performing designs in the evolutionary competition during generational transitions, an elitism mechanism was adopted. Through elitism, the most successful design of any given generation is privileged and transferred to the next generation without being subjected to any modification mechanisms. Consequently, the possibility of performance loss between generations was effectively eliminated. Thermal and structural analysis methods, essential for objective function evaluation, have been integrated into the developed genetic algorithm software using the open-source Python programming language. The finite difference (FDM) and Rayleigh-Ritz methods were employed to perform the specified numerical analyses. To verify the accuracy of the conducted numerical method analyses, a sample board design was replicated in the MATLAB PDE Toolbox, a well-established commercial software. The results obtained from the commercial software demonstrate a high level of agreement with the numerical analysis results generated via the custom script developed in the Python programming language. Through this adopted multi-objective optimization approach, a method has been established that significantly enhances the thermal and structural performance of printed circuit boards within a short computational timeframe.