top of page

coremicro® Structural Health Monitoring and Vibration Analysis Suite (SHM-Suite)
 

RTSHM Kernel.bmp

This software tool is used for the analysis of sensor data logs and allows characterization of events associated with a monitored system. Features correlated to system conditions can be identified and used for processing by custom algorithms (user application software). A key advantage of this product is that it can be integrated with our Optimized Neuro Genetic Fast Estimator (ONGFE), an ANN based software toolset based on both conventional and deep neural networks, for conducting automated diagnostics. For example, data logs containing vibration signals that correspond to structural states can be input for training neural networks with this software. The trained networks can then be deployed for real-time structural health classification. The main screen of the RTSHM-kernel provides the following main features:
 

  • Data Log Feature Extraction Plotting and Recording. In this section, the user has the ability to perform analysis of sensor signals (such as vibration) captured in a data log. After entering the file name and selecting acquire data, the following functions are available:

  1. View the evolution of the time-domain signal (real-time emulation)

  2. Compute, plot, and record features of the time-domain signal. Baseline features include mean, rms, variance, kurtosis, crest factor, and peak amplitude for a selectable window size 

  3. View the evolution of the frequency domain signal with an adjustable FFT window size. Spectrum patterns can be analyzed for identifying and visualizing frequency data patterns

  4. Compute, plot, and record features of the frequency-domain signal. Baseline features include mean, rms, variance, kurtosis, crest factor, and peak amplitude for a selectable window size

  5. Activate multi-resolution analysis (MRA) and plot/record results for four types of wavelets. Four wavelet algorithms are provided: (1) the standard discrete Haar transform (S-DHT); (2) lifting scheme discrete Haar transform (LS-DHT); (3) Linear interpolation wavelets; and (4) Daubechies D4 wavelet. The user can select a desired MRA level for generating a set of averages and coefficients

  • Feature Extraction for Neural Network Diagnostics. The second section is feature extraction for designing pattern recognition algorithms. In the case of methods based on Artificial Neural Networks (ANN), the user can identify, select, and extract parameters which are correlated with the system response (features) for building neural network training and testing files. This process entails the following:

  1. Compute and plot a characteristic frequency response for a representative state (such as normal operation)

  2. Extract features from the representative data log as well as for up to three additional logs.

  3. Generate neural network training and testing files using the extracted features as inputs and user-defined class IDs as outputs.

  4. For deep neural networks (e.g. time series convolutional nerual network), the raw data can be immediately input into the neural network training and testing files without the need for a separate feature extraction step using our library of statistical features

Frequency Domain Features

RTSHM Kernel (Features).bmp

Time and Frequency Domain Features

Wavelet Features Generation

Feature Extraction, Selection, and Neural Network Training FIle Generation

RTSHM Kernel (Feature Extraction).bmp
bottom of page