Early Fault Warning of Spindle Based on the Adaptive Weighted Fuzzy Petri Net

Abstract

The spindle system of machine tool produces a huge amounts of process data when it works. These data directly reflect the running state of spindle system but are seldom used to perform early fault warning. This paper proposes a novel early fault warning method adaptive weighted fuzzy Petri-net. Firstly, the long short-term memory (LSTM) is put forward to predict the time-series of future state for spindle system. Then, in order to design a reasoning framework for dynamic knowledge which can adapt to changes in the area of knowledge, an adaptive weighted fuzzy petri-net (AWFPN) is brought up to perform fault diagnosis. Finally, the effectiveness and feasibility of proposed method are verified by simulations and experiments. Results show that the proposed early fault warning method could effectively help to find potential fault information in the manufacturing process and provide the useful advice for maintenance.