Intra-class Adaptation for Bearing Fault Diagnosis Under Variable Conditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fault diagnosis methods for rolling bearings under variable working conditions face challenges due to differing data distributions between training and real-time data, limiting their universality and generalization, especially in complex industrial environments.
Innovation Solution
An intra-class adaptation fault diagnosis method using a deep convolutional neural network with transfer learning, which involves acquiring and segmenting bearing vibration data under different health states, performing FFT, extracting features with an improved ResNet-50, and optimizing a target function with conditional distribution distance and classification loss to adapt to varying conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional deep learning methods are used for fault diagnosis, then automatic feature extraction and processing speed are improved, but model universality and generalization are worsened due to distribution differences between training and real-time data
Solution Approach 1:
The patent applies domain adaptation techniques that transform the parameter distribution of source domain data to match the target domain distribution. By changing the statistical parameters (mean, variance, etc.) of the input data through adaptation layers, the model maintains high processing speed while improving universality across different working conditions
Solution Approach 2:
The patent introduces an intermediary domain adaptation layer between the convolutional neural network layers. This intermediary component acts as a mediator that aligns feature distributions from different domains, enabling the model to process data from various working conditions effectively while maintaining automatic feature extraction capabilities
2Measurement precision
If manual feature extraction methods are used, then diagnostic accuracy can be improved with expert knowledge, but processing efficiency and automation are worsened
Solution Approach 1:
The patent replaces manual mechanical feature extraction processes with automated deep learning-based feature extraction. The convolutional neural network automatically learns diagnostic features from raw vibration signals, substituting the need for expert manual feature engineering while maintaining or improving diagnostic accuracy through data-driven feature learning
3Measurement precision
If sufficient vibration data for each equipment status under all working conditions is collected, then model training accuracy is improved, but data acquisition complexity and time are worsened
Solution Approach 1:
The patent performs preliminary domain adaptation training on source domain data that is easier to acquire. By pre-training the model on available data and then adapting it to the target domain with minimal target data, the system achieves high training accuracy without requiring extensive collection of data under all possible working conditions
Solution Approach 2:
The patent creates a universal model through domain adaptation that can function effectively across multiple working conditions and equipment states. The adapted model learns transferable features that generalize to different scenarios, reducing the need for condition-specific data collection for each working state
Data Source
AI summary
The invention relates to a fault diagnosis method for a rolling bearing under variable working conditions. Based on a convolutional neural network, a transfer learning algorithm is combined to handle the problem of the reduced universality of deep learning models. Data acquired under different working conditions is segmented to obtain samples. The samples are preprocessed by using FFT. Low-level features of the samples are extracted by using improved ResNet-50, and a multi-scale feature extractor analyzes the low-level features to obtain high-level features as inputs of a classifier. In a training process, high-level features of training samples and test samples are extracted, and a conditional distribution distance between them is calculated as a part of a target function for backpropagation to implement intra-class adaptation, thereby reducing the impact of domain shift, to enable a deep learning model to better carry out fault diagnosis tasks.