NC Machine Tool Control Using Neural Network Error Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current process monitoring systems for numerically controlled machine tools are not sufficiently fast, precise, or sensitive in detecting issues during machining, leading to potential damage and inefficiencies, and often result in false detections and unnecessary downtimes.
Innovation Solution
A control device equipped with a computer-implemented neural network that monitors the machine tool's operating state by reading input data from the machine control device and outputs indicative data to detect errors in real-time, allowing for immediate reactions such as machine stops or tool changes, and potentially adjusts process parameters to optimize machining.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional collision sensors with vibration measurement are used for process monitoring, then the monitoring system can detect collisions during machining, but the detection speed and precision are insufficient leading to delayed responses and potential machine damage
Solution Approach 1:
The patent replaces traditional mechanical vibration-based collision sensors with a neural network-based monitoring system that processes machine control data. This substitution enables faster and more accurate detection of machining errors by analyzing control data patterns rather than relying on mechanical vibration thresholds, thereby improving both detection speed and reliability simultaneously
Solution Approach 2:
The patent changes the monitoring parameters from vibration amplitude measurements to neural network-processed control data patterns. By transforming the input parameters and using intelligent algorithms, the system achieves superior detection performance in terms of both speed and accuracy compared to traditional vibration-based methods
2Reliability
If traditional monitoring systems with simple threshold comparisons are used, then the system structure remains simple, but false detections occur frequently causing unnecessary downtimes
Solution Approach 1:
The patent replaces simple threshold comparison logic with a neural network-based evaluation system. This substitution eliminates false detections by enabling intelligent pattern recognition in control data, while the neural network can be implemented within existing control device hardware, managing the complexity increase through software-based intelligence rather than additional hardware
3Productivity
If faster and more accurate error detection is implemented, then machine damage and rejects are reduced, but the monitoring system becomes more complex and requires more computational resources
Solution Approach 1:
The patent makes the neural network monitoring system multi-functional by using the same system for various types of error detection (collisions, tool breaks, abnormal wear) and for process optimization. This universal approach improves productivity across multiple functions while avoiding the complexity increase that would result from implementing separate specialized systems for each monitoring task
Solution Approach 2:
The monitoring system uses the machine's existing control data as input, making the system self-sufficient without requiring additional sensors or external monitoring equipment. This self-service approach enables advanced error detection and process optimization while minimizing additional hardware complexity
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a control device 200 for use on a numerically controlled machine tool 100, comprising a machine control unit 230 for controlling actuators of the machine tool for a machining process for a workpiece to be performed on the machine tool 100, in particular on the basis of control data, and a monitoring unit 250 for monitoring an operating state of the machine tool 100. In accordance with the invention, the monitoring unit 250 has a computer-implemented neural network 253 (NN), which in particular is designed to read input data from the machine control unit 230 and to output output data specifying an operating state of the machine tool 100.