Article Conveyance Control Using Non-Production Learning Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In actual production with deep learning-controlled apparatuses, the inflexibility of control parameters hinders the acquisition of sufficient learning data, leading to potential yield deterioration.
Innovation Solution
An article conveyance apparatus and method that generate a learning model using data on article weight, state, and control parameters, allowing selective operation between production and non-production modes to collect and store learning data without affecting yield.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If control parameters are flexibly changed to acquire sufficient learning data, then the quantity of learning data is improved, but yield deteriorates
Solution Approach 1:
The system performs preliminary data collection during non-production periods when the apparatus is not actively manufacturing. Learning data is accumulated in advance during these idle periods, so that sufficient training data is available before actual production begins, eliminating the need to change parameters during production and thus protecting yield.
2Reliability
If control parameters are kept fixed during actual production to maintain yield, then yield is preserved, but the quantity of learning data deteriorates
Solution Approach 1:
The system alternates between production periods and non-production periods. During non-production periods, the control parameters are freely changed to collect diverse learning data. During production periods, fixed parameters ensure stable yield. This periodic switching allows both objectives to be achieved at different times.
3Quantity of substance
If learning data is collected during production mode, then the quantity of learning data is improved, but manufacturing precision deteriorates
Solution Approach 1:
The system performs the data collection action preliminarily during non-production mode before actual manufacturing begins. This ensures that all parameter variations and data gathering are completed in advance, so that during production mode, the system operates with fixed, optimized parameters to maintain high manufacturing precision.
Data Source
AI summary
A method of controlling an article conveyance apparatus (1) includes steps of: (A) generating, with, as learning data, information regarding an amount of charge indicating a weight value of articles that a conveyer (20) conveys to a member disposed on a downstream side, information indicating a state of the articles on the conveyer, and a control parameter for the conveyer, a learning model that estimates the control parameter to be set for conveyance of the articles having a targeted weight; (B) performing conveyance control of the articles, based on the learning model; (C) performing selective switching between a production mode involved in actual production and a non-production mode not involved in the actual production, and causing the article conveyance apparatus to operate; and (D) collecting and storing, when the article conveyance apparatus operates in the non-production mode, as the learning data, the information regarding the amount of charge actually acquired and the control parameter.