Cutting Process Evaluation Using Sensor Fusion and CNN Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cutting process evaluation methods struggle to accurately assess the quality of processed products due to reliance on trial and error adjustments and limited ability to detect and diagnose abnormalities in cutting processes, especially in high-speed fracture mechanics where data sampling is challenging, leading to inconsistent product quality.
Innovation Solution
A learning device with an input processor and learning processor that acquires physical quantities during the cutting process, converts them into state variables, and updates an evaluation model using a convolutional neural network to predict abnormalities and their causes, enabling precise evaluation of cutting process quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If trial and error adjustment of mold position and shape is used, then the cutting process can be performed with simple equipment, but the manufacturing precision and reliability of processed products cannot be ensured
Solution Approach 1:
The patent replaces manual trial-and-error adjustment with an automated evaluation system that uses sensors to measure physical quantities (load, sound, vibration, temperature) and an evaluation model to automatically determine optimal mold settings. This substitutes mechanical adjustment processes with automated measurement and evaluation, resolving the contradiction between simple equipment operation and high manufacturing precision.
Solution Approach 2:
The patent implements a feedback mechanism where the evaluation model uses measured physical quantities from the cutting process to automatically adjust and optimize mold position and shape settings. This closed-loop feedback system ensures consistent manufacturing precision without requiring complex manual adjustments, as the system learns from actual process data to improve settings automatically.
2Manufacturing precision
If physical quantity measurement and threshold comparison is used for quality evaluation, then the manufacturing precision can be improved, but the reliability of abnormality diagnosis is insufficient
Solution Approach 1:
The patent transforms multiple physical quantity parameters (load, sound, vibration, temperature) into a comprehensive evaluation by changing from simple threshold comparison to a multi-parameter analysis system. The evaluation model processes these parameters together to determine both quality and abnormality causes, improving reliability by considering the interrelationships between different physical parameters rather than evaluating them in isolation.
Solution Approach 2:
The patent creates a composite evaluation model that integrates multiple types of sensor data and evaluation criteria into a unified diagnostic system. This composite approach combines information from load sensors, sound sensors, vibration sensors, and temperature sensors to provide comprehensive abnormality diagnosis, enhancing reliability by using the combined information from multiple sources rather than relying on single-parameter thresholds.
3Productivity
If internal information of processing apparatus is measured for quality determination, then the productivity can be improved through automated evaluation, but the measurement precision of cutting process abnormalities is insufficient
Solution Approach 1:
The patent creates a multi-functional evaluation system that simultaneously performs multiple tasks: measuring physical quantities, evaluating processing quality, diagnosing abnormality causes, and optimizing process parameters. This universal system uses the same sensor infrastructure and evaluation model for multiple purposes, improving measurement precision by coordinating all measurement functions rather than relying on separate specialized systems for each function.
Solution Approach 2:
The patent introduces an evaluation model as an intermediary layer between raw sensor measurements and quality determination. This intermediary processes and integrates data from multiple sensors (load, sound, vibration, temperature) to extract meaningful information about cutting abnormalities, enhancing measurement precision by transforming raw data into diagnostic insights rather than directly using unprocessed sensor readings.
4Device complexity
If conventional evaluation methods are used, then the device complexity remains low, but the ability to detect and diagnose cutting process abnormalities is limited
Solution Approach 1:
The patent segments the evaluation system into distinct functional modules: load measurement, sound measurement, vibration measurement, temperature measurement, and evaluation model processing. This segmentation allows each sensor and processing function to be independently optimized while maintaining overall system simplicity through modular architecture, thereby improving abnormality detection capability without excessively increasing device complexity.
Solution Approach 2:
The evaluation model serves as an intermediary that integrates data from multiple simple sensors to provide comprehensive abnormality detection. This intermediary layer combines information from load, sound, vibration, and temperature measurements to diagnose cutting process abnormalities, enhancing detection capability by coordinating multiple simple measurement functions rather than requiring a single complex detection system.
Data Source
AI summary
A learning device includes an input processor and a learning processor. The input processor acquires a physical quantity related to a cutting process, and inputs a state variable based on the physical quantity to the learning processor, and the learning processor updates, based on a measured cutting result, an evaluation model that outputs an evaluation result of the cutting process based on the state variable.