Development and Validation of a Quality Control System in Automotive Gearbox Assembly Lines Using Adaptive Filtering

Abstract

Traditional signal analysis techniques, together with the monitoring of vibrations arising from the operation of automotive gearboxes, can be used to define objective values ​​for acceptance or rejection in a quality program. However, due to the complexity of the dynamics of such mechanical systems and the noise inherent to a production line, there is great difficulty in defining control symptoms and determining limit values. This work presents a procedure using adaptive filtering, via the least squares algorithm (LMS), as a tool to aid in quality control procedures via vibration monitoring. The methodology is based on the use of signals, measured in a standardized test, as the primary and reference signal in the filtering procedure. It is intended, through its use, to remove from the signal to be filtered, only that portion that is correlated with the mesh, with the residue being a portion of the signal coming from manufacturing or assembly failures. To evaluate the capacity of the proposed methodology, initially computationally generated signals were used, simulating real gearing signals. Subsequently, vibration signals were collected from transmissions approved in a quality control procedure and common manufacturing and assembly non-conformities were inserted. From the filtering results (filtered signal and residue), the procedure was validated.