Faults diagnosis based on system model in a discrete-part machining system

Root cause identification is one of deciding factors in current manufacturing competitions. For a discrete-part machining system, it is very challenging to identify the faults, since the final product variation caused by faults is an accumulation from all stations. This paper explores a faults diagnosis methodology for the root causes identification of a serial machining system. Firstly, a system model is described to capture the relationship between process faults and product quality. Then based on the model, the maximum likelihood estimation algorithms are built to estimate the key parameters of measurement data, such as the mean value and variance, which followed by a hypothesis testing method to determine the root causes at certain confidence level. A real machining case illustrates the effectiveness of the proposed faults diagnosis methodology.