Mechanical Joining Monitoring Using Fourier Fault Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current process monitoring methods for mechanical joining processes, such as riveting, struggle to provide precise evaluations and assign measured values to specific fault patterns, often resulting in unclear classifications of correct or incorrect processes and limited ability to identify specific errors.
Innovation Solution
The method involves receiving and analyzing multiple measured values of process variables, performing a Fourier transformation to determine Fourier coefficients, reducing these coefficients to a 2-dimensional subspace for comparison with reference points, and using this comparison to determine a monitoring result that can assign deviations to specific error types and improve reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional process monitoring methods (envelope technique, window technique) are used to monitor mechanical joining processes, then the monitoring can be implemented with simple comparison of process variables to reference curves, but the evaluation clarity is insufficient and specific error assignment is not possible
Solution Approach 1:
The patent transforms the monitoring approach by moving from 1D process variable comparison (force vs. displacement) to 2D spectral analysis (frequency spectrum). The Fourier transformation converts time-domain or path-domain data into frequency-domain data, creating a new dimension for comparison where characteristic frequencies and their amplitudes form a spectral fingerprint of the joining process. This dimensional transformation enables clear evaluation of process quality and specific error identification.
2Adaptability or versatility
If individual envelope curves or comparison windows are defined for each joining process to account for different joining tasks and equipment, then the monitoring can be adapted to specific processes, but the divergent data records make consistent evaluation difficult
Solution Approach 1:
The patent creates a universal monitoring framework based on spectral analysis that can handle diverse joining processes and equipment types. By transforming all process data into the frequency domain, the method establishes a common evaluation language that works across different joining tasks, sensors, and equipment configurations. The spectral representation with characteristic frequencies and amplitudes provides a standardized format that enables consistent comparison and evaluation across the entire production line, while still being adaptable to specific process requirements through reference spectral curves.
3Measurement precision
If the sampling rate of sensors is increased to capture more process variables, then the measurement resolution is improved, but the number of value pairs increases resulting in more complex data processing
Solution Approach 1:
The patent extracts the essential information from large volumes of high-resolution process data by applying Fourier transformation. Instead of processing all individual data points, the transformation identifies and extracts characteristic frequencies and their corresponding amplitudes that define the process quality. This extraction process reduces the complex spectral data to key parameters (dominant frequencies, amplitude ratios) that can be compared against reference values, dramatically simplifying the evaluation while maintaining high measurement resolution benefits.
Data Source
Figure 1A~1B
Figure 2
Figure 3
AI summary
The invention relates to a method for process monitoring in a mechanical joining process, the method comprising receiving (S100) a plurality of measured values fi(xi) during the performance of the mechanical joining process, wherein the measured values each relate at least two process variables fi, xi, where i = 1... n, to one another. The method further comprises performing (S110) a Fourier transformation, wherein a number of Fourier coefficients (φ1, φ2,... φn, A1, A2,... An) are determined for the measured values fi(xi) and reducing (S120) the number of Fourier coefficients to an at least 2-dimensional subspace, thereby obtaining a measuring point (φk, φl,....), (φk, Al) or (Ak, Al) (214, 216). The method moreover comprises comparing (S130) the measuring point (φk, φl,....), (φk, Al) or (Ak, Al) (214, 216) with at least one reference point (Rk, Rl,...), (Rk, Sl,...) or (Sk, Sl,...) (219) in the at least 2-dimensional subspace and determining (S140) a monitoring outcome on the basis of the comparison with the reference point (219).