Production support device

Abstract

A production support device is provided with a first learning data acquisition part for acquiring a plurality of first learning data for use in first machine learning on component type pairs of components to be mounted on a board, for which a reward is obtained by attempting and improving mounting processing by exchanging the component type pairs between a plurality of component mounting machines; a learning data storage part for storing the acquired plurality of first learning data in a classified manner according to a predetermined classification criterion; a sampling part for randomly sampling from each of the first learning data stored in a classified manner in the learning data storage part; and a learned model storage part for storing a learned model generated by performing the first machine learning using the randomly sampled first learning data.

Claims

[1] Production support device with a first learning data acquisition part for acquiring a plurality of first learning data for use in a first machine learning about component type pairs of components to be mounted on a board, in which a reward is obtained by attempting and thus improving a mounting process by exchanging the component type pairs between a plurality of component mounting machines, a learning data storage part for storing the acquired plurality of first learning data in a classified manner according to a predetermined classification criterion, a sampling part for randomly sampling the individual first learning data stored in a classified manner in the learning data storage part, and a learned model storage part for storing a learned model generated by performing the first machine learning using the randomly sampled first learning data. [2] Production support device according to claim 1, comprising a production information acquisition part for acquiring production information containing at least arrangement data for representing the arrangement of the component placement machines and component type data for representing the component types of the components and commanding the production of the board by mounting the new components using the component placement machines, and a reasoning part that uses the arrangement data, component type data of the components and the learned model contained in the production information to reason and output the component type pairs from the new component types distinguished by the component type data as exchange objects. [3] The production support device according to claim 1 or 2, wherein the sampling part randomly samples the first learning data from the individual first learning data stored in the learning data storage part in a classified manner at an arbitrarily adjustable composition ratio. [4] The production support device according to claim 1 or 2, wherein the sampling part randomly samples the first learning data from the individual first learning data stored and collected in a predetermined number or more in the learning data storage part. [5] The production support device according to claim 1 or 2, comprising a learned model generating part for generating the learned model by repeating the first machine learning using the randomly extracted first learning data. [6] The production support device according to claim 5, wherein the first learning data is randomly extracted in a state where the individual first learning data in a predetermined number or more are stored and collected in the learning data storage part. [7] A production support device according to claim 1 or 2, wherein the first learning data is formed by combining arrangement data representing the arrangement of the component mounting machines, component type data representing the component types of the components, component type pair data representing the component type pairs of the component types distinguished from each other by the component type data, and result data representing a result obtained when the components of the exchanged component types are mounted by the plurality of component mounting machines. [8] The production support device according to claim 7, wherein the learning data storage part classifies and stores the obtained plurality of first learning data according to the predetermined classification criterion for the result data. [9] The production support device according to claim 8, wherein the result data includes a cycle time required for mounting the components, wherein the classification criterion is a criterion for classifying the first learning data according to the cycle time. [10] A production support device according to claim 8, wherein the result data includes a case where the swapping of the component type pairs is impossible, wherein the classification criterion is a criterion for classifying the first learning data depending on whether the swapping of the component type pairs is possible or not. [11] The production support device according to claim 2, further comprising a second learning data obtaining part for obtaining second learning data including inferred component type pair data representing component type pairs inferred by the inferring part, wherein the learned model storing part stores the learned model generated by performing either the first machine learning using the first learning data or a second machine learning on the component type pairs in which a reward is obtained by trying and improving the mounting processing by exchanging the component type pairs inferred using the second learning data. [12] The production support apparatus according to claim 11, comprising a learned model generating part for generating the learned model by repeating either the first machine learning or the second machine learning. [13] The production support device according to claim 12, wherein the learned model generating part selects and executes either the first machine learning or the second machine learning according to an exploration ratio for determining the exploration rate according to the epsilon greedy method, thereby generating the learned model. [14] A production support device according to claim 1 or 2, wherein the reward is given when the cycle time required for mounting the components is shortened in the simulation after the pairs of component types have been exchanged. [15] The production support device according to claim 14, wherein the reward becomes greater the greater the reduction in the cycle time after exchanging the component type pairs compared to the cycle time before exchanging the component type pairs. [16] A production support device according to claim 1 or 2, wherein the reward is given when, in the simulation, after exchanging the pairs of component types, the components are mounted on the board in ascending order of size. [17] A production support device according to claim 16, wherein the reward is given when, in the simulation, after exchanging the pairs of component types, the components are mounted on the board in ascending order of height from the surface of the board. [18] A production support device according to claim 2, wherein the plurality of components are each accommodated in a carrier tape wound on a reel, each reel being inserted into a feeder which feeds the components accommodated in the carrier tape to the component placement machine. [19] The production support apparatus according to claim 18, wherein the inference part infers and outputs feeder pairs representing the feeders that feed the components to the component mounting machine as the component type pairs. [20] A production support device according to claim 2, which comprises a first step of determining component mounting machine pairs for representing the component mounting machines among a plurality of component mounting machines based on the arrangement data, and a second step of inferring and outputting the component type pairs at the component mounting machine pairs. [21] The production support device according to claim 20, wherein the inference part infers and outputs the component mounting machine pairs in the first process step using the production information and the learned model.

Images