Multichannel deconvolution of vibrational signals: A state-space inverse filtering approach

Area of Science:

• Acoustics and Signal Processing
• Vibrational Analysis
• Structural Dynamics

Background:

• Deconvolution of noisy multichannel measurements is a persistent challenge in acoustics.
• Existing methods include Fourier techniques and model-based parametric approaches, with Wiener filters being prominent.
• Model-based state-space techniques offer significant improvements when acoustic physics are incorporated.

Purpose of the Study:

• To develop an advanced deconvolution processor for analyzing the vibrational response of coupled acoustic test objects.
• To extract transient excitation signals from complex vibrational data.
• To enhance the understanding and prediction of operational performance impairments.

Main Methods:

• Development of a multiple input/multiple output (MIMO) structural model for the test objects.
• Application of subspace identification techniques to create an inverse filter during calibration.
• Utilizing model-based state-space methods incorporating underlying process physics.

Main Results:

• The developed inverse filter successfully extracted excitation signals from test objects.
• Feasibility was demonstrated through a mass transport experiment and object calibration tests.
• The processor proved effective in deconvolving noisy, multichannel vibrational data.

Conclusions:

• Model-based state-space techniques, combined with subspace identification, offer a powerful approach to deconvolution in acoustics.
• The developed inverse filter is effective for analyzing vibrational responses of coupled acoustic systems.
• This method enhances the capability to accurately identify transient excitations impacting operational performance.