Optimal and Reliable Design of the Bonding Wire for a Power Module by Using Reliability-based Design Optimization and a Kriging Metamodel

Power electronic modules are widely used in various industries such as aerospace, railway, automotive, alternative energy generation and distribution application for conversion and control of electrical power. An insulated gate bipolar transistor (IGBT) is a power module that is wildly used in various electric products such as power supplies, DC/DC converters and uninterruptible power systems. In automotive applications, new power modules are characterized by smaller sized or more integrated components, finer pitches and new materials. New requirements to comply with environmental friendly directives create challenges in assuring the reliability for higher performance and lower cost. However, for hybrid and electric vehicles, the power modules operate in hostile environments including wide temperature ranges and vibrations. In operation, the generated heat under high power can cause the IGBT chip to overheat and affect the performance of the IGBT. In fact, thermal cycles build up thermo-mechanical stresses in the solder joints connecting electronic components and generate damage. These damage accumulation leads to crack initiation and to solder joint failure, such as bonding wire cracking and bonding wire lift off [1]. The bonding wire failure is one of the most common failures observed on power modules. In automotive power electronics, design optimization of these devices is crucial in order to ensure low cost and reliability requirements. The Reliability-Based Design Optimization approach (RBDO) is developed to balance cost and reliability, where it offers a means to quantify uncertainty propagation and determine a priori the most reliable design that meets performance criteria [2]. In this paper, the RBDO approach combined to the kriging metamodel is developed to search the best optimal and reliable design of the wire bonding. Due to time consumption of the numerical model based on the electro-thermo-mechanical simulation, the kriging metamodeling approximation is adopted in order to surrogate the performance functions. The RBDO approach used in this work combines updating kriging approximation by using the technique of constraint boundary sampling to enhance the prediction of the Kriging model [3]. This methodology is demonstrated in the design of a wire bond structure in a power electronic module with aim to reduce the probability of failure due to the wire bond lift off in power electronic module. The numerical results show the efficiency and the attractiveness of the proposed RBDO methodology to deal with industrial application.