Machining Diagnosis Model Using Simulation-to-Actual Data Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machining state diagnosis devices have large errors between estimated and actual machining state information, leading to unreliable abnormality determination in machining states.
Innovation Solution
A training device acquires simulation and actual machining data to generate a trained model, which predicts actual machining data using simulation data, enabling accurate diagnosis of machining abnormalities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a machining state diagnosis device learns the relationship between estimated machining state information and actual machining state information associated with multiple machining processes, then the device can determine whether the machining state has an abnormality, but the error between machining state information and actual machining state information becomes large, making the diagnosis result unreliable
Solution Approach 1:
The patent segments the machining processes into individual process-specific models rather than using a single unified model for multiple processes. Each machining process (e.g., turning, milling, drilling) has its own trained model that learns the relationship between simulation and actual machining state information separately. This segmentation reduces the error in each specific process and improves overall diagnosis reliability.
Solution Approach 2:
The patent changes the approach from using a single unified model to using multiple process-specific models with different parameters optimized for each machining process. By training separate models for different machining processes, the system adapts parameters specific to each process type, reducing generalization errors and improving measurement precision for each specific process.
2Measurement precision
If simulation data and actual machining data are used to train a prediction model, then accurate prediction of actual machining data is achieved, but the complexity of the system increases due to the need for data acquisition and model training infrastructure
Solution Approach 1:
The patent creates a virtual copy of the machining process through simulation data that mirrors the actual machining process. By training a prediction model on simulated data that replicates real machining conditions, the system achieves accurate predictions without requiring complex real-time sensing infrastructure. The simulation acts as a digital twin or copy of the physical machining process.
Solution Approach 2:
The patent performs preliminary training of the prediction model using simulation data before actual machining operations. The model is pre-trained offline on simulated machining state information, so that during actual machining, the model can directly predict machining state information without requiring complex real-time data acquisition and processing infrastructure.
Data Source
AI summary
A training device includes a simulation data acquirer that acquires first simulation data through machining simulation with a first machining program, an actual machining data acquirer that acquires first actual machining data through actual machining with the first machining program, and a model generator that learns a relationship between the first simulation data and the first actual machining data and generates a trained model for predicting, based on second simulation data acquired through machining simulation with a second machining program, second actual machining data acquired through actual machining with the second machining program.