Search Results

In the current study, an automatic and interpretable predictive maintenance system is proposed. The system provides a fully automatic training process for predictive maintenance models without human intervention. On the other hand, as failure reasons are critical for product development. The proposed pipeline also demonstrates the interpretation on automatic trained model to present insights for engineers to acquire mechanism of interested events. To study the system, four automatic machine learning methods and two interpretation modules are evaluated for the pipeline with Isuzu’ real vehicle data correspondingly. The overall performance of the automatic and interpretable system is demonstrated as well. Key words: predictive maintenance, AutoML, interpretation

Field Issue (Malfunction) incidents are costly for the manufacturer’s service department. Especially for commercial truck providers, downtime can be the biggest concern for our customers. To reduce warranty cost and improve customer confidence in our products, preventive maintenance provides the benefit of fixing the problem when it is small and reducing downtime of scheduled targeted service time. However, a normal telematics system has difficulty in capturing useful information even with pre-set triggers. Some malfunction issue takes weeks to find the root cause due to the difficulty of repeating the error in a different vehicle and engineers must analyze large amounts of data. In order to solve these challenges, a machine-learning-based predictive software/hardware system has been implemented.