Applied Signal Processing

Cabin-noise VolvoThe development of a fast vibro-acoustic analysis system for 100% production testing of final cars at Volvo. The difficulty was to manage the analysis of sound and vibration components related to the rpm (rotations per minute), creating a non-stationary process, making regular FFT analysis difficult. The classical solution has been to use an rpm sensor and re-sample the data versus rpm, and thus making the data stationary in order domain. The sensor mounting was, however,  too tedious and consequently an rpm sensor could not be used in the implementation. Thomas Lagö had an idea on how to perform rpm based FFT analysis (order analysis) without a tachometer sensor, thus avoiding the largest problem. In principle, the sound contains the rpm information and thus a tachometer sensor is not needed, given that the incoming data is correctly sampled and analyzed using a robust rpm classification method. Not easy, but durable. The system is unique and Volvo was the first company being able to perform 100% testing of the “sound quality according to specification” on-line in production, using a “red light” versus a “green light” for cars failing or passing, respectively.  Signal Control AB in Lund, President Sven Olsson, has been responsible for all installations and maintenance of the production systems. The work has led to a patent and several papers at international conferences.

The development of a fast multi-axis balancing system, designed for Volvo’s AWD cars. Since the power transmission axis are coupled, it took too many “trial and errors” to find proper locations that would balance all degrees of freedom for the axis. By using a unique concept based on SVD (Singular Value Decomposition) and FRF (Frequency Response Function) data, the degrees of freedom could be analyzed and the system coupling decoupled, and thus the number of measurements needed to successfully mount the required weights, became much less than before. A saving of typically a factor of four was found, which amounted to a substantial production saving for Volvo in terms of time and money. Many systems have been installed in the production lines both in Sweden and Belgium. The system is still unique. The work was performed in collaboration with Signal Control in Lund, President Sven Olsson. The work has led to several papers at international conferences.

HAFT Window 1The development of fixed bandwidth frequency filters for FFT analysis using a modified Flattop window approach. This work was performed on request from the National Standards Group in Bor s (SP), Sweden. They needed to make telephone measurements using their HP3562 Dynamic Signal Analyzer, but it was a must to fulfill the telephone standard, demanding a 10 Hz analysis bandwidth. The FFT, together with the Flattop window (P401) gave about 9.3 Hz. Close to 10Hz, but not close enough. In collaboration With Signal Control AB in Lund, this work was successfully performed and implemented in the HP3562 analyzer. The work has led to several papers at international conferences, and the approach is still unique.

Rapid Production Line Testing using Sound and Vibration as Clasification Parameters. The ability to correctly classify a bad component over a good one is not the key challenge in PASS-FAIL testing. The key challenge is to make the yellow zone around the production limit as small as possible – the “faulty zone.” If the system indicates a PASS when it should have indicated a FAIL, a faulty product is being shipped out to a customer. If the system indicates a FAIL when the product should have indicated a PASS, a good product is probably wasted. Usually, the latter is better than the first but that rather depends on the consequences for each “mistake.” In the left figure below, the classical system approach using a system that is not able to fully locate the key system parameters for a valid GO-NOGO testing is depicted.

The common thread in the above work has been applied signal processing, with applications in vibro-acoustics. Within these topics, areas like applied signal processing, control system theory, frequency analysis, filter theory (both analog and digital), optimization theory and sampling techniques has become essential. Thomas Lagö has successfully combined these multi-discipline applied signal processing areas with the vibro-acoustic applications.