Numerical Controller Machine Learning for Chatter Prevention

Solution Overview

Problem

Existing numerical controllers require significant operator experience and time to adjust machining conditions to prevent chatter and tool wear/breakage, making it difficult to select optimal conditions, especially in situations where trial and error may not yield effective results.

Innovation Solution

A numerical controller equipped with a machine learning device that performs reinforcement learning to adjust spindle revolution number and feed rate based on state data, using reward conditions to optimize machining conditions and prevent adverse effects on the machined surface.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If operator manually adjusts machining conditions through trial and error, then machining conditions can be modified, but it takes a long time and requires significant experience

Engineering Contradiction:

Improvemachining condition adjustmentVSAvoidtime to adjust machining conditions

Core Design Contradiction:

Ease of operationVSLoss of time

Solution Approach 1:

The numerical controller automatically adjusts machining conditions using a machine learning device that performs reinforcement learning. The system self-learns optimal machining parameters by evaluating state data (vibration, sound, tool wear) and receiving reward signals, eliminating the need for operator intervention and trial-and-error adjustments.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system continuously monitors machining state data including vibration acceleration, sound pressure, and tool wear status, compares these against reward conditions, and uses the feedback to automatically adjust machining conditions. The reinforcement learning algorithm updates its policy based on the received rewards, creating a closed-loop feedback system that optimizes machining parameters in real-time.

Inventive Principle:

Principle #23Feedback

2Reliability

If operator adjusts machining conditions to prevent chatter and tool breakage, then machining stability can be improved, but it requires significant experience and ability

Engineering Contradiction:

Improvemachining stabilityVSAvoidmachining condition adjustment

Core Design Contradiction:

ReliabilityVSEase of operation

Solution Approach 1:

The numerical controller automatically adjusts machining conditions using a machine learning device that performs reinforcement learning. The system self-learns optimal machining parameters by evaluating state data (vibration, sound, tool wear) and receiving reward signals, eliminating the need for operator intervention and trial-and-error adjustments.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system dynamically changes machining parameters (spindle speed, feed rate, depth of cut) based on real-time state data and learned policies. The reinforcement learning algorithm continuously optimizes these parameters to maintain machining stability and prevent chatter and tool breakage without requiring operator expertise.

Inventive Principle:

Principle #35Parameter changes

3Adaptability or versatility

If traditional machining condition selection is used, then simple cases can be handled, but complex machining situations cannot be effectively resolved

Engineering Contradiction:

Improvehandling of machining situationsVSAvoideffectiveness in complex situations

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The numerical controller automatically adjusts machining conditions using a machine learning device that performs reinforcement learning. The system self-learns optimal machining parameters by evaluating state data (vibration, sound, tool wear) and receiving reward signals, eliminating the need for operator intervention and trial-and-error adjustments.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system transitions from static, pre-programmed machining parameters to dynamic, real-time parameter adjustment. The reinforcement learning model continuously adapts machining conditions based on current machining state, enabling effective handling of complex and varying machining situations that require flexible, context-dependent decisions.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS9964931B2Numerical controller with machining condition adjustment function which reduces chatter or tool wear/breakage occurrence

Publication Date: 2018.05.08 FANUC LTD

AI summary

A numerical controller includes a machine learning device for performing machine learning of machining condition adjustment of a machine tool. The machine learning device calculates a reward based on acquired machining-state data on a workpiece, and determines an adjustment amount of machining condition based on a result of machine learning and machining-state data, and adjusts machining conditions based on the adjustment amount. Further, the machine learning of machining condition adjustment is performed based on the determined adjustment amount of machining condition, the machining-state data, and the reward.