Modular diagnostics of computer-controlled machine tools and mechatronic systems.

Abstract

As the manufacturing environment becomes more flexible and reconfigurable, machine tool systems have become more modular and complex. Although the probability of failure of a complex machining system is higher, very little effort has been invested in enhancing the diagnostics of these complex systems. Although many machine modules have limited built-in diagnostics, they are not well utilized for system-level diagnostics. The goal of this research is to develop a framework that can be applied to generic diagnostic modeling of mechatronic systems. A new approach based on functional modeling has been developed. The conventional functional modeling approach for diagnostics has limitations in modeling the failures of dynamic systems in existing applications. To overcome this, the proposed approach defines the failure modes using only lower-level component functions, and models the information propagation through higher levels. A methodology is developed to compose the higher-level models from the lower-level models and to analyze the failure behavior of the higher-level model. Using a graph representation of the module structure, this method models (1) the horizontal propagation of states from one module to another module, and (2) the vertical propagation of undiagnosable failures to the higher levels of the system. This methodology allows one to determine the diagnosability of the entire system. The proposed framework has been applied to modeling a real boring machine composed of several mechatronic devices, such as a piezoelectric actuator embedded in the cutting tool. It is shown that the proposed model is effective in modeling modular system diagnostics and evaluation of the diagnosability of different cutting tools and sequential machining processes. In addition to the modeling framework, a failure detection method for a cutting process has been developed, which utilizes a disturbance observer to estimate the cutting force in the process. The failure detection method has been successfully implemented in a real machine tool. The experimental results demonstrate that the method is able to detect tool breakage, the poor surface of a workpiece, and the tool misalignment problem.