The status of fault patterns on part surfaces can provide valuable information about the condition of the manufacturing system. In this work, we aim to develop a reliable fault detection and diagnosis tool in order to assure the automated production of high-quality parts. Such a tool provides a means of integrating the manufacturing and design phases. Accurate detection of the part surface condition in manufacturing ensures the fault-free design of the manufacturing parameters and machine components.
This paper introduces a mathematical transform that has the potential to detect faults in manufacturing machines. Specifically, the paper focuses on the decomposition of complex signals to allow the detection of faults. The Karhunen-Loève transform is investigated by means of numerically-generated signals. Numerical signals are studied to decompose a variety of signals, including deterministic, stochastic, stationary, and nonstationary signals. Finally, the potential utility of the proposed technique is discussed in the context of a newly-maturing manufacturing process.