Convolutional Neural Network for Real-Time Clamp Force Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting clamp force in bolts using artificial neural networks are complex and difficult to implement in real-time, resulting in low accuracy due to variations in torque coefficients caused by physical and environmental factors.
Innovation Solution
A method utilizing a convolutional neural network that generates a cepstrum image, applies convolution and pooling filters to extract representative features, and uses Adam optimization for improved weight updates, enabling real-time clamp force prediction by comparing synthesized images with predetermined values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional artificial neural network methods are used for clamp force prediction, then prediction capability is provided, but the system complexity increases and real-time prediction becomes difficult
Solution Approach 1:
The patent segments the clamp force prediction task into distinct processing stages: signal acquisition, cepstrum image generation, convolutional feature extraction, and prediction output. This segmentation allows each component to be optimized independently, reducing overall system complexity while maintaining prediction accuracy.
Solution Approach 2:
The patent replaces conventional neural network architectures with a convolutional neural network (CNN)-based image processing system. By transforming audio signals into cepstrum images and applying convolutional operations, the system achieves real-time prediction capability with reduced computational complexity compared to traditional approaches.
2Measurement precision
If conventional neural network processing is used, then prediction is possible, but processing speed is slow and real-time measurement is difficult
Solution Approach 1:
The patent substitutes conventional neural network signal processing with an image-based convolutional neural network approach. By converting audio signals to cepstrum images and applying parallel convolutional operations, the system achieves significantly faster processing speeds suitable for real-time measurement while maintaining prediction accuracy.
Solution Approach 2:
The patent implements periodic processing of audio signals by dividing them into frames and generating cepstrum images at regular intervals. This periodic action enables real-time prediction by processing signals in continuous streams rather than analyzing entire signal sequences at once.
3Ease of operation
If torque control method is used for bolt tightening, then workability is excellent, but clamp force distribution becomes large due to torque coefficient variations
Solution Approach 1:
The patent implements a feedback mechanism by using acoustic signal analysis during the tightening process to predict clamp force in real-time. This feedback allows for monitoring and adjustment of tightening parameters to achieve more uniform clamp force distribution while maintaining the simplicity of torque control operation.
Data Source
AI summary
A method for predicting a clamp force using a convolutional neural network includes: generating a cepstrum image from a signal processing analysis apparatus; extracting a characteristic image by multiplying a predetermined weight value to pixels of the generated cepstrum image through artificial intelligence learning; extracting, as a representative image, the largest pixel from the extracted characteristic image; synthesizing an image by synthesizing the extracted representative image information; and predicting a clamp force by comparing the synthesized image with a predetermined value.