Sewage Process Fault Monitoring With Incremental Fuzzy Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neural network-based fault monitoring methods for sewage treatment processes are time-consuming and resource-intensive due to their complex structures and large number of hyperparameters, making them unsuitable for fast-paced industrial applications.
Innovation Solution
A fault monitoring method using a fuzzy width adaptive learning model with a set of first-order TS fuzzy subsystems and enhanced node layers, allowing for incremental model reconstruction without retraining the entire network, enabling online fault monitoring in sewage treatment processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional deep neural networks are used for fault monitoring, then measurement precision is improved, but training time and resource consumption increase significantly
Solution Approach 1:
The patent divides the deep neural network into multiple shallow sub-networks, each responsible for processing specific features or aspects of the sewage treatment data. This segmentation allows each sub-network to be trained independently and more quickly, while collectively achieving the fault detection accuracy of a full deep network.
Solution Approach 2:
The patent implements a dynamic model where the network structure and parameters can be automatically adjusted during operation based on incoming data characteristics. This dynamic adaptation enables the system to maintain high detection accuracy without requiring extensive retraining, as the model evolves continuously with minimal computational overhead.
2Measurement precision
If traditional deep neural networks are used for fault monitoring, then measurement precision is improved, but hardware resource requirements increase
Solution Approach 1:
By segmenting the complex deep network into simpler shallow sub-networks, the patent reduces the computational burden on hardware resources. Each sub-network requires fewer processing units and memory, making the overall system more suitable for deployment in industrial environments with limited hardware capabilities.
Solution Approach 2:
The patent uses multiple simplified sub-networks that replicate core processing functions rather than relying on a single complex deep network. This copying approach distributes the computational workload across simpler units, reducing peak resource requirements while maintaining detection accuracy through ensemble processing.
3Measurement precision
If traditional deep neural networks are used for fault monitoring, then measurement precision is improved, but adaptability to changing processes decreases
Solution Approach 1:
The patent implements a dynamically adaptable model where each shallow sub-network can independently adjust its parameters in response to changing sewage treatment processes. This dynamic capability allows the system to quickly adapt to process variations without requiring complete retraining, maintaining both accuracy and adaptability.
Solution Approach 2:
The patent incorporates preliminary adaptation mechanisms where the model is pre-configured with multiple sub-networks that can be selectively activated or adjusted based on anticipated process changes. This preliminary preparation enables faster response to actual process variations compared to traditional models that require full retraining.
Data Source
AI summary
The invention discloses a sewage treatment process fault monitoring method based on fuzzy width adaptive learning model. Including “offline modeling” and “online monitoring” two stages. “Offline modeling” first uses a batch of normal data and 4 batches of fault data as training samples to train the network offline and label the data. After the network training is completed, the weight parameters are obtained for online monitoring. “Online monitoring” includes: using newly collected data as test data, using the same steps as offline training networks for online monitoring. The output result of online monitoring adopts one-hot encoding to realize zero-one discrimination of the output result of online monitoring, so as to realize fault monitoring. The present invention only needs to increase the number of enhanced nodes, reconstruct in an incremental manner, and does not need to retrain the entire network from the beginning. It can complete the network training in a short time and realize the rapid fault monitoring in time, which has high practical application value.