Powerplant Component Inspection Using Neural Networks for Defect Variability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing non-destructive inspection methods for internal defects in components are limited by component-to-component variability and often fail to accurately distinguish between normal variations and defects.
Innovation Solution
A neural network-based inspection system using a transducer to transmit and sense signals, combined with a self-attention mechanism, processes response signals to estimate frequency peak offsets and predict defect presence or absence, providing enhanced accuracy and adaptability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional non-destructive inspection methods are used, then the inspection process is simple, but the accuracy of defect detection is limited due to component-to-component variability
Solution Approach 1:
The patent replaces traditional mechanical inspection methods with a neural network-based system that uses signal processing and machine learning. The neural network analyzes response signals from transducers to detect defects, substituting human judgment and simple threshold-based methods with an intelligent system that can distinguish between normal variability and actual defects.
Solution Approach 2:
The patent transforms the inspection approach by changing from direct defect detection to indirect detection through response signal analysis. The neural network processes complex signal parameters including amplitude, frequency, and phase information to infer defect presence, thereby improving detection accuracy beyond what traditional methods can achieve.
2Reliability
If traditional inspection methods are used, then the system is easy to operate, but the ability to distinguish normal variations from defects is insufficient
Solution Approach 1:
The neural network system performs self-learning and self-adjustment during operation. It automatically adapts to different component types and variations without requiring manual recalibration or expert intervention, thereby improving reliability while maintaining ease of operation. The system serves itself by continuously improving its defect detection capabilities through the self-attention mechanism.
Solution Approach 2:
The patent implements feedback through the neural network's ability to process and learn from inspection results. The self-attention mechanism provides feedback on which signal features are most important for defect detection, allowing the system to continuously improve its reliability by learning from previous inspections and adjusting its analysis priorities.
3Measurement precision
If a neural network with self-attention mechanism is used, then the defect detection accuracy is improved, but the processing complexity increases
Solution Approach 1:
The patent segments the complex inspection task into distinct processing stages: signal acquisition, response signal processing, neural network analysis, and defect determination. The self-attention mechanism is applied specifically to the frequency peak offset estimation stage, breaking down the complex processing into manageable segments that can be handled by specialized computational modules.
Solution Approach 2:
The patent introduces an intermediary layer between the raw signal data and the final defect determination. The self-attention mechanism acts as an intermediary that processes and transforms the response signals into meaningful features, bridging the gap between complex signal data and simple defect yes/no decisions, thereby managing processing complexity.
4Adaptability or versatility
If training data from pairs of training components is used, then the neural network adapts to component variability, but the data processing requirements increase
Solution Approach 1:
The patent merges data from multiple training components into a unified training dataset. By combining response signals from pairs of training components, the neural network learns to recognize patterns across different components, improving adaptability to component variability. The self-attention mechanism efficiently processes this merged data by identifying and focusing on the most relevant features across all training examples.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system offers improved defect detection accuracy by mitigating component variability and predicting future defects, enabling efficient inspection with minimal downtime and cost.
Implementation Method 1
transmitting a first signal into the subject component and sensing the subject component for a second signal produced as a result of the first signal being transmitted into the subject component
Implementation Method 2
sensing the subject component for a second signal produced as a result of the first signal being transmitted into the subject component
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A method of inspecting a component (36) for the presence or absence of a defect is provided that includes: using a transducer (48) to inspect a subject component (36) comprising a solid metallic material by transmitting a first signal into the subject component (36) and sensing the subject component (36) for a second signal produced as a result of the first signal being transmitted into the subject component (36), and producing a subject component response signal representative of the second signal; processing the subject component response signal received from the transducer (48), the processing using a neural network trained on response signal training data from pairs of training components, and the processing including producing a neural network output value; and producing an indication of a presence or an absence of a defect in the subject component (36) based on the neural network output value.