Mechanical Press State Separation for Real-Time Failure Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing manufacturing processes fail to detect failures such as cracks, ripples, or micro-cracks in sheets or plates during the pressing process, leading to costly corrections when these failures are discovered later.
Innovation Solution
A computer-implemented method that uses machine learning to indicate failures in manufacturing processes by receiving input signals from sensors monitoring physical quantities, transforming these signals into parameters, deriving latent features, mapping them into clusters representing different states of the manufacturing process, and optionally indicating a failure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If visual inspection is used to detect failures, then detection capability is provided, but detection timing is delayed until after several pressing cycles
Solution Approach 1:
The patent applies preliminary action by performing detection during the pressing process itself rather than after completion. Sensors monitor physical quantities (force, position, acceleration) in real-time during pressing, enabling failure detection at the moment it occurs rather than after several cycles have passed.
Solution Approach 2:
The patent replaces manual visual inspection with automated sensor-based monitoring systems. Mechanical/physical sensing (force sensors, position sensors, acceleration sensors) substitutes for human visual detection, enabling continuous real-time monitoring during the pressing process.
2Reliability
If detection is performed later, then fewer false alarms may occur, but correction costs increase significantly
Solution Approach 1:
The patent implements feedback by continuously monitoring pressing parameters and comparing them against learned patterns from the neural network. When deviations indicating potential failures are detected, the system provides immediate feedback to operators or automated systems, enabling timely intervention before failures escalate to costly defects.
Solution Approach 2:
By detecting failures during the pressing process rather than later, the patent enables preliminary corrective action to be taken while the workpiece is still in the press or immediately after removal, preventing the development of severe defects that would require expensive rework or scrapping.
3Reliability
If multiple sensors and machine learning processing are implemented, then detection reliability is improved, but system complexity increases
Solution Approach 1:
The patent applies universality by using a single neural network processing system that handles multiple sensor input types (force, position, acceleration) and performs multiple functions (pattern recognition, failure detection, classification). This multi-functional approach consolidates what could be separate complex systems into one integrated platform.
Solution Approach 2:
The neural network acts as an intermediary that translates complex multi-sensor data into simple, actionable failure indicators. The MLS processes the complex relationships between multiple sensor signals and provides simplified output (failure/no failure classification), reducing the complexity burden on the overall control system.
4Loss of time
If real-time monitoring during pressing is implemented, then failure detection timing is improved, but processing requirements and complexity increase
Solution Approach 1:
The patent replaces complex real-time mechanical monitoring systems with a computational approach using neural networks. The MLS performs the complex pattern recognition and decision-making that would otherwise require sophisticated real-time control systems, enabling accurate failure detection through software-based analysis of sensor data.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
The present invention is related to a computer-implemented method of, a data processing system for and a computer program product for indicating a failure of a manufacturing process as well as to a corresponding manufacturing machine and further to a computer- implemented method of training a machine learning system (MLS) for indicating states of a manufacturing process. An input signal of a sensor is transformed into a parameter. The parameter is provided to the MLS, which derives latent features. The latent features are mapped into one of several distinct clusters each representing a mode of the manufacturing process. Finally, a failure of the manufacturing process based on the different states of the manufacturing process may be indicated.