High pressure waterjet cutting, with or without abrasive additive, is an unconventional machining process that is being used in nearly every manufacturing sector today, from aerospace to the food and textile industry.
The process involves the conversion of water pressure energy into kinetic energy to allow cutting almost every kind of material. For such a task, an Ultra High Pressure (UHP) pump is used to achieve nominal pressures in the range between 20 and 50 MPa, or even higher.
Machine tool components, due to those challenging pressure conditions, are strongly affected by faults and performances deterioration, with consequent impacts on quality and productivity. In-process analysis of sensor signals may support the end-user in assessing the health state of the system and detecting possible malfunctions.
This study investigates the usage of the water pressure signal as a potential source of information for both health condition monitoring and fault diagnosis. Due to the nature of the signal, a functional data analysis approach is proposed, aimed at assessing the stability over time of the functional pattern that characterizes the in-control behavior of the system, in a Multivariate Statistical Process Control framework.
A Fourier basis is selected to convert sampled signals into functional form, and time-warping functions are applied to cope with curve misalignment. Fourier parameters and warping function coefficients are merged into a multivariate vector that is monitored over time for early detection of possible malfunctions.
Functional data classification is then used to identify the root cause of the observed pattern deviation in case of an actual fault affecting one of the machine tool components.
Real data acquired under both in-control working conditions and in the presence of actual faulty UHP components are analyzed. Finally a comparison of the proposed classification technique with other methods already presented in the literature is provided and discussed.