Machine Learning for Early Defect Detection in Automotive Semiconductors

In the automotive sector, semiconductors utilized in modern vehicles undergo rigorous testing prior to being supplied to end-users. This testing phase represent an important phase to guarantee that the semiconductors meet the high standards required in this domain, as well as to avert potential safety concerns that could arise from faulty semiconductors. This phase must be well design and without possible escapes. In fact, the industry defines specific standards on the numbers of defects per million. However, the testing phase accounts for a substantial portion of the total cost of a microcontroller and can create production bottlenecks. The testing phase is generally divided into two stages: pre-packaging (front end) and post-packaging (back end), with the latter being more time-consuming and expensive. To address these challenges, the goal of this master's thesis project is to identify non-compliant semiconductors at an early stage, prior to packaging, in the front end. By doing so, the study aims to minimize the cost of testing in the production process by reducing the number of defective devices that arrive in the back-end phase. Additionally, other potential applications can lead to making the testing process more efficient, such as early termination of testing for a device if it is predicted to be faulty. To achieve this objective, this project focuses on applying well-established machine learning techniques using data from the pre-packaging phase. This involves utilizing well-known data analysis tools and algorithms to identify patterns in the data that can predict, within a certain range, whether a semiconductor is likely to meet the standards or not. One of the most critical aspects of this project is striking a balance between accurate predictions and false predictions, while also considering economic metrics. Achieving this balance will ensure that the testing process is as efficient as possible and that production costs are not exacerbated by false positives or negatives. Overall, this master's project has the potential to significantly impact the automotive industry by enhancing the production process while ensuring that high-quality semiconductors are delivered to customers.