Spectrum-based modal parameters identification with Particle Swarm Optimization

• Description of spectrum-based modal parameters identification with Particle Swarm Optimization (smiPSO) algorithm.
• Results compared to other known modal identification methods for simulated and real signals.
• Good identification results in noisy conditions.

The paper presents the new method of the natural frequencies and damping identification based on the Artificial Intelligence (AI) Particle Swarm Optimization (PSO) algorithm. The identification is performed in the frequency domain. The algorithm performs two PSO-based steps and introduces some modifications in order to achieve quick convergence and low estimation error of the identified parameters’ values for multi-mode systems. The first stage of the algorithm concentrates on the natural frequencies estimation. Using the information about the natural frequencies, measurement data are filtered and corrected dampings as well as amplitudes are calculated for each preliminary identified mode. This allows regrouping particles to the area around proper parameters values. Particle regrouping is based on the physical properties of modally tested structures. This differs the algorithm from other PSO based algorithms with particles regrouping. In the second stage of the algorithm parameters of all modes are tuned together in order to adjust estimates. The procedure of identification, as well as the appropriate algorithm, is presented and some SISO examples are provided. Results are compared with the results obtained for the selected, already developed modal identification methods. The paper presents practical application of AI method for mechanical systems identification.