Machining Chatter Prediction Using Simulated and Experimental Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for predicting machining operation status, particularly chatter occurrence in machine tools, face accuracy and reliability issues due to uncertainties in model parameters and the need for large amounts of experimental data, limiting their applicability in industrial environments.

Innovation Solution

A method involving a two-phase neural network training process, where a pre-trained neural network is first trained with simulated data to learn general dependencies and then fine-tuned with experimental data to enhance prediction accuracy, reducing the reliance on extensive experimental data collection and accounting for uncertainties in variable parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If physical models are used to predict chatter occurrence, then prediction can be performed with limited experimental data, but prediction accuracy is insufficient due to uncertainties in model parameters

Engineering Contradiction:
Improveprediction accuracyVSAvoidreliability of prediction
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent introduces a neural network as an intermediary between the physical model and the prediction output. The neural network learns to correct the uncertainties and errors in the physical model parameters by training on experimental data, while the physical model provides the structural framework. This intermediary layer allows the system to benefit from both the physical understanding in the model and the accuracy of experimental measurements without requiring complete trust in either approach alone.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent creates a composite prediction approach by combining physical models and machine learning models into a hybrid system. The physical model provides the theoretical foundation and structural constraints, while the machine learning component adapts to real-world variations and uncertainties. This composite approach leverages the strengths of both methodologies to achieve superior prediction accuracy and reliability compared to using either approach alone.

Inventive Principle:
Principle #40Composite materials

2Measurement precision

If machine learning is used to perform prediction, then precise prediction can be obtained, but large amounts of experimental data are required which is challenging to collect in industrial fields

Engineering Contradiction:
Improveprediction accuracyVSAvoidamount of experimental data
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent performs preliminary action by pre-training the neural network on simulated data generated from the physical model before fine-tuning with experimental data. This pre-training phase allows the model to learn the fundamental relationships and patterns from abundant simulated data, reducing the amount of experimental data needed for subsequent fine-tuning. The simulated data serves as a preliminary training resource that prepares the model for more accurate learning from limited experimental measurements.

Inventive Principle:
Principle #10Preliminary action

3Productivity

If stability lobe diagrams are used to represent prediction results, then optimal process parameters can be selected for stable machining, but the experimental stability limits differ from theoretical ones due to model inaccuracies and parameter uncertainties

Engineering Contradiction:
Improvemaximal productivityVSAvoidaccuracy of stability limits
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent implements feedback by training the neural network on experimental measurement data that captures the actual stability limits observed in practice. The model learns from the discrepancies between theoretical predictions and experimental results, using this feedback to adjust its predictions and align them with real-world behavior. This feedback mechanism allows the system to continuously improve its accuracy by incorporating actual performance data.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent applies parameter changes by allowing the neural network to learn and adapt the effective parameters that govern chatter stability, rather than relying on fixed theoretical parameters. The model adjusts its internal parameters based on training data, enabling it to account for variations in material properties, tool conditions, and machine dynamics that are not captured by theoretical models. This parameter adaptation bridges the gap between theory and practice.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3893064A1Method for predicting status of machining operation
Publication Date: 2021.10.13 UNITED MACHINING MILL AG
  • EP3893064A1 patent drawingFigure 1~2
  • EP3893064A1 patent drawingFigure 3
  • EP3893064A1 patent drawingFigure 4~6

AI summary

The present invention is directed to a method for predicting status of machining operation, in particular chatter occurrence comprising the following steps: training a neural network having an input layer, at least one hidden layer, an output layer and a plurality of weights in a pre-training phase and a final-training phase, wherein in the pre-training phase a pre-training data set is provided to the neural network to obtain a pre-trained neural network and in the final-training phase a final-training data set is fed to the pre-trained neural network to obtain a final-trained neural network, wherein the pre-training data set comprises simulated data and the final-training data set comprises experimental data; and performing prediction by utilizing the final-trained neural network to derive prediction data.