As the number of electrical devices in modern vehicles increases, the battery becomes more critical component for the operation of vehicles. To ensure the startability of the vehicle, battery conditions such as state of charge and state of health should be properly monitored and maintained. To reduce walk-home incidents due to no-start situation, appropriate warning should be issued to the driver to advise necessary actions such as replacing or re-charging the battery. For the last couple of years, General Motors has studied and developed several battery health monitoring methods based on different battery health signatures. Yet, it is found that relying on a single method may lead to false alarm or misdetection due to lack of information or uncertainty. This paper develops the algorithm for more robust and reliable battery health monitoring and prognosis, by applying Evidence Theory to fuse different battery health signatures. The algorithm is evaluated by using battery cranking data from the set of batteries collected from field. The test results show reliable and robust performance.
