# US 10,121,107 B2 · Machine learning device and method for optimizing frequency of tool compensation of machine tool, and machine tool having the machine learning device > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 10,121,107 B2 - Document ID: US10121107B2 - Status: Granted - Kind: B2 - Application number: 15/042,839 - Filed: 2016-02-12 - Published: 2017-03-30 - Granted: 2018-11-06 - Earliest claimed: 2015-09-30 - Assignees: Fanuc Corporation - Applicants: FANUC Corporation - Inventors: Fumito Yukawa - Web page: https://patentgrep.com/patent/US10121107B2 ## CPC - G05B 13/029: {using neural networks and expert systems} - G05B 19/404: characterised by control arrangements for compensation, e.g. for backlash, overshoot, tool offset, tool wear, temperature, machine construction errors, load, inertia (G05B19/19, G05B19/41 take precedence) - G06N 3/042: Knowledge-based neural networks; Logical representations of neural networks - G06N 20/00: Machine learning - G05B 2219/33034: Online learning, training - G05B 2219/49166: Compensation for measured deviation of tool path, as function of lenght of path - G05B 2219/49307: Learn, learn operational zone, feed, speed to avoid tool breakage - G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract A machine learning device and a machine learning method for optimizing timing at which a tool is to be compensated in a machine tool, and a machine tool including the machine learning device. The machine learning device includes a state observation unit for observing a time interval for compensating the tool, a processing error amount of a workpiece processed by the machine tool, and a machine working ratio of the machine tool as state variables, and a learning unit for learning an action value with respect to a change of a tool compensation interval based on the tool compensation interval, the processing error amount of a workpiece, and the machine working ratio that are observed by the state observation unit. ## Claims ### Claim 1 A machine learning device for learning an action value with respect to a change of a tool compensation interval that is a time interval for compensating a tool in a machine tool, the device comprising: a state observation unit for observing the tool compensation interval of the machine tool, a processing error amount of a workpiece processed by the machine tool, and a machine working ratio of the machine tool as state variables; a learning unit for learning by associating the tool compensation interval, the processing error amount, and the machine working ratio that are observed by the state observation unit with a change amount of the tool compensation interval; and a decision making unit for determining a change amount of the tool compensation interval based on a result of learning by the leaning unit. ### Claim 2 (depends on 1) The machine learning device according to claim 1, wherein the learning unit includes: a reward calculation unit for calculating a reward based on the tool compensation interval, the processing error amount, and the machine working ratio, and a function update unit for creating or updating an action value table based on the tool compensation interval, the processing error amount, the machine working ratio, and the reward. ### Claim 3 (depends on 2) The machine learning device according to claim 2, wherein the decision making unit determines the change amount of the tool compensation interval based on the action value table. ### Claim 4 (depends on 2) The machine learning device according to claim 2, wherein the reward calculation unit provides a negative reward when the processing error amount increases or the machine working ratio decreases, and provides a positive reward when the processing error amount decreases or the machine working ratio increases. ### Claim 5 (depends on 1) The machine learning device according to claim 1, wherein the learning unit computes state variables observed by the state observation unit using a multilayer structure. ### Claim 6 (depends on 2) The machine learning device according to claim 2, wherein the function update unit updates the action value table thereof using an action value table updated by another learning machine device. ### Claim 7 (depends on 1) A machine tool including the machine learning device according to claim 1. ### Claim 8 (depends on 7) The machine tool according to claim 7, wherein the state observation unit further observes a temperature inside the machine tool, and wherein the learning unit learns by associating the tool compensation interval, the processing error amount, the machine working ratio, and the temperature inside the machine tool that are observed by the state observation unit with a change amount of the tool compensation interval. ### Claim 9 A machine learning method for learning an action value with respect to a change of a tool compensation interval that is a time interval for compensating a tool in a machine tool, the method comprising: observing the tool compensation interval of the machine tool, a processing error amount of a workpiece processed by the machine tool, and a machine working ratio of the machine tool as state variables; learning by associating the tool compensation interval, the processing error amount, and the machine working ratio that are observed by the state observation unit with a change amount of the tool compensation interval; and determining a change amount of the tool compensation interval based on a result of the learning. ### Claim 10 (depends on 9) The machine learning method according to claim 9, wherein the learning includes: calculating a reward based on the tool compensation interval, the processing error amount, and the machine working ratio, and creating or updating an action value table based on the tool compensation interval, the processing error amount, the machine working ratio, and the reward.