Challenges of the Press-Fitting Procedure

A plastic element is press-fitted into a metal frame. A fully automated monitoring for this process is required. It has to be checked that no cracks in the plastic element emerge during the operation.

This is because any cracks in the plastic element can lead to fatigue fracture and functional failure of the component under normal load.

A complete visual inspection of the plastic element is not possible. As QASS’ structure-borne sound technology provides a view of what is going on inside the component even when it is not possible to look inside, it is applied to monitor the assembly procedure.

Schematic representation of press-fitting of a plastic element into a metal frame.

Schematic representation of the press-fitting process for joining a plastic element and a metal frame.

QASS' Structure-Borne Sound Analysis Method

QASS’ measuring system Optimizer4D works on the basis of structure-borne sound. Crack formations cause characteristic acoustic impulses, which our measuring system can detect virtually at the moment they occur.

The QASS structure-borne sound sensor is mounted as close as possible to the component. Our sensor captures the acoustic vibrations of the joining process and simultaneously converts the sensor signals (time-amplitude signals) into a measurement signal that the Optimizer4D samples at a very high frequency.

The Optimizer4D digitizes all measurement signals and transforms their data in real-time by using the Fast-Fourier Transform in order to obtain the frequency information (pitches) hidden in the signal. The self-developed method we use for spectral analysis is called high-frequency pulse measurement (HFIM).

Through HFIM it is possible to reliably detect crack formations and to differentiate them from normal process events. In fact, on the basis of energy levels from two-dimensional time-amplitude signal data, crack formations and process events often cannot be properly distinguished from another. Through additional frequency information the differentiability improves considerably. Within the three-dimensional time-amplitude-frequency signal data, crack formations produce very specific signal patterns with amplitudes up to the high-frequency range, which differentiate considerably from those of usual process events.

All measurement data processed with the HFIM are directly available to the QASS software Analyzer4D.

QASS' measuring system Optimizer4D with measuring chain (including structure-borne sound sensor, preamplfier, sensor cable and preamplifier cable).

QASS’ measuring system Optimizer4D.

Threedimensional spectrogram of a NOK press-fitting process with crack formation before use of a spectral filter.

3D-spectrogram of an incorrect press-fitting process with crack occurence in the plastic element before application of a frequency mask to remove typical machine and interference noises.

Pattern Recognition Algorithms

With smart tools such as pattern recognition and automatic decision routines, the Analyzer4D represents an innovative and powerful program for real-time signal data analysis. All data is clearly visualized for the user in a 3D-spectrogram.

In contrast to crack signals, machine and interference noises produce low-frequency signals, which do not show any high-frequency components. A spectral filter (which we call frequency mask) can be employed to cancel out typical machine and interference noises from the signal data. In this way, the signal-to-noise ratio is improved.

With pattern recognition, it is possible to detect crack formations in the plastic element that occur during press-fitting. In a first step, signal patterns of crack formations are referenced and stored in a pattern library. Then, an automated comparison algorithm scans the incoming filtered signal data for the stored patterns and calculates similarities to the observed patterns. If during assembly a crack forms inside the plastic element, the algorithm detects a high similarity score. Then the Optimizer4D produces a NOK-signal and triggers the ejection of the component.

Visualization of the effectiveness of a frequency mask through a comparison of an unfiltered and a filtered 3D-spectrogram.

Comparison between a 3D-spectrogram of a NOK press-fitting process with crack formation without activated spectral filter (right) and a 3D-spectrogram of the same incorrect process after use of a frequency mask to cancel out typical machine and interference noises (left).

Threedimensional spectrogram of a NOK press-fitting process with crack formation after use of a spectral filter.

3D-spectrogram of the incorrect press-fitting process with crack occurence in the plastic element after application of a frequency mask to cancel out typical machine and interference noises.