Rotating Machine Health Classification Using Vibration Spectra
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting rotating machine failures, such as those in manufacturing plants or IoT devices, are not very accurate and often mislead due to varying loads, leading to unplanned downtimes and potential severe damage.
Innovation Solution
A computer-implemented method using a neural network to predict machine failure types and severity by transforming input signals from sensors, such as vibrations, into features, which are then analyzed to provide early and reliable automatic indications of machine failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simple KPIs like RMS or crest factor are used to determine failure type, then the analysis is simple and fast, but the accuracy is low and results are misleading due to varying loads
Solution Approach 1:
The patent transforms vibration signals from time domain to frequency domain using Fast Fourier Transform, changing the representation parameters from simple amplitude values (RMS, crest factor) to spectral features (frequency components, harmonics, sidebands). This parameter transformation enables accurate identification of failure types by analyzing frequency characteristics that remain consistent across different load conditions.
Solution Approach 2:
The patent replaces manual expert analysis of vibration spectra with an automated neural network system. The neural network automatically identifies patterns in frequency domain data corresponding to different failure types (unbalance, misalignment, bearing defects), eliminating the need for expert interpretation while maintaining high accuracy across varying operational conditions.
2Measurement precision
If manual or semi-automatic analysis by experts is performed, then specific failure types can be identified accurately, but the process is time-consuming and not scalable to large amounts of machines
Solution Approach 1:
The patent replaces manual expert analysis with an automated neural network system that processes vibration spectra. The neural network has been trained on labeled data to recognize patterns corresponding to specific failure types, enabling it to automatically classify failures without human intervention. This substitution maintains the accuracy of expert analysis while eliminating time consumption and enabling scalability to monitor thousands of machines simultaneously.
Solution Approach 2:
The neural network system performs self-learning and automatic classification of failure types without requiring continuous expert intervention. Once trained, the system independently analyzes vibration spectra, identifies failure patterns, and provides diagnostic results, enabling autonomous operation across large machine fleets.
3Adaptability or versatility
If many rotating machines are monitored simultaneously, then comprehensive coverage is achieved, but the complexity of analyzing and managing data from all machines increases significantly
Solution Approach 1:
The patent implements a universal neural network model that can analyze vibration data from different types of rotating machines (pumps, motors, compressors, fans) using the same processing pipeline. The system monitors multiple machines simultaneously by applying the identical spectral analysis and classification algorithm to each machine's vibration data, enabling scalable deployment across diverse equipment without increasing system complexity.
Data Source
AI summary
The present disclosure relates to a computer-implemented method of, a data processing system for, and a computer program product for indicating machine failures as well as to a corresponding machine and a computer-implemented method of training a neural network for indicating machine failures. At least one input signal based on at least one physical quantity of at least one machine part is transformed into at least one feature. A neural network predicts a class and/or a severity of at least one machine failure based on the at least one feature.