Fault detection method for refrigeration units based on improved deep learning model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deep learning models for refrigeration unit fault diagnosis face issues such as network degradation, overfitting, and data imbalance, leading to inaccurate fault detection, particularly in refrigeration units with complex and highly time-varying data.
Innovation Solution
A fault detection method combining ResNet with CBAM and utilizing the LOF-ADASYN algorithm to remove outliers and balance data, incorporating dropout layers to mitigate overfitting and enhance generalization, and multiscale convolution to focus on key features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of network layers is increased to deeply mine data features, then feature extraction capability is improved, but network training efficiency deteriorates and overfitting occurs
Solution Approach 1:
The patent divides the deep network into multiple residual blocks, where each block contains skip connections that segment the gradient flow path. This allows training of deeper networks by breaking them into manageable segments with independent optimization, resolving the contradiction between depth for feature extraction and training efficiency.
Solution Approach 2:
The patent introduces batch normalization layers as intermediaries between convolutional layers. These normalization layers act as mediators that stabilize the training process, enable faster convergence, and allow the network to go deeper without sacrificing training efficiency, thus resolving the contradiction between network depth and training speed.
2Measurement precision
If the number of network layers is increased to deeply mine data features, then feature extraction capability is improved, but network degradation occurs
Solution Approach 1:
The patent uses residual blocks that segment the network into functional units with skip connections. Each residual block learns residual features rather than complete features, which prevents degradation and maintains reliability as network depth increases.
Solution Approach 2:
The patent applies batch normalization before activation functions in each residual block as a preventive measure. This normalization cushions the network against internal covariate shift that would otherwise cause degradation in deeper layers, maintaining performance stability.
3Device complexity
If fixed weights are used to process features of each element in the sequence, then model complexity is reduced, but key feature detection capability deteriorates
Solution Approach 1:
The patent replaces fixed weights with dynamic attention mechanisms that adaptively weight different features based on their importance. The attention scores are computed dynamically during inference, allowing the model to focus on key fault features while maintaining reasonable complexity through efficient attention computation.
4Ease of manufacture
If data imbalance is not addressed, then data processing simplicity is maintained, but fault detection accuracy deteriorates
Solution Approach 1:
The patent applies data augmentation and resampling techniques as preliminary actions before model training. By balancing the dataset in advance, the model can be trained on balanced data without modifying the training process itself, maintaining simplicity while improving fault detection accuracy.
Data Source
AI summary
A fault detection method for refrigeration units based on an improved deep learning model is provided, including the following steps: S1: obtaining operating parameters of a refrigeration unit in a normal operating state and in states with different fault types as data sets; S2: detecting local outliers in the data set by using a local outlier factor algorithm and removing the local outliers, and then expanding the data set by using adaptive synthetic sampling; S3: normalizing the data set; S4: constructing a fault detection model; and S5: inputting the parameters of the tested refrigeration unit into the fault detection model, and judging whether the tested refrigeration unit has a fault and the type of the fault.