Machine Learning Alignment of Optical Parts in Light Source Units
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for aligning optical parts in light source units are operator-dependent and time-consuming, leading to inefficiencies and potential misalignment, especially when dealing with varying part qualities, which can result in reduced production efficiency and unnecessary operations.
Innovation Solution
A machine learning device and manufacturing apparatus that utilize a state observation unit, decision data acquisition unit, and learning unit to calculate rewards and update value functions based on adjustment time and light state measurements, determining optimal movement methods for aligning optical parts using reinforcement learning and deep learning techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If an operator performs alignment based on experience and manual adjustment, then alignment can be performed with current equipment, but the operation time varies greatly and production efficiency is reduced
Solution Approach 1:
The patent replaces the manual mechanical adjustment system with an automated control system that uses a camera to capture images of the optical part, processes these images through image processing algorithms, and automatically determines the optimal position and orientation. This substitution of mechanical manual operation with automated optical and computational systems directly reduces alignment operation time and increases production efficiency.
Solution Approach 2:
The system enables self-service alignment by automatically capturing images, processing them to determine optical part characteristics, and calculating the optimal position and orientation without requiring operator intervention. The automated control unit performs the alignment procedure independently based on pre-stored algorithms, eliminating the variability in operation time caused by operator experience levels.
2Extent of automation
If an operator manually adjusts the optical part position and orientation, then alignment can be achieved, but the process is difficult to automate and depends on operator skill
Solution Approach 1:
The patent replaces manual operator skill-based adjustment with an automated control system that uses image processing algorithms to determine optical part characteristics. The system captures images with a camera, processes them automatically to extract position and orientation information, and controls the adjustment mechanism without requiring operator skill or experience.
Solution Approach 2:
The patent introduces an intermediary image processing system that acts as a mediator between the optical part and the control system. The image processing unit captures visual information and converts it into actionable data about position and orientation, enabling automated control without direct operator intervention or skill-based judgment.
3Adaptability or versatility
If a fixed alignment procedure is used for all parts, then the process is simple to execute, but it cannot adapt to varying part qualities and may include unnecessary operations
Solution Approach 1:
The patent implements a dynamic alignment procedure where the control unit adjusts the alignment steps based on real-time image processing results. The system determines the number and type of adjustment operations needed based on the actual characteristics of each optical part, rather than following a fixed procedure. This dynamic adaptation eliminates unnecessary operations for parts that require minimal adjustment and ensures optimal procedures for parts needing precise alignment.
Solution Approach 2:
The system performs preliminary image capture and processing to assess the current state of the optical part before initiating the alignment procedure. Based on this preliminary analysis, the control unit determines the optimal alignment steps in advance, avoiding unnecessary operations by identifying parts that are already properly aligned or require only minimal adjustment.
Data Source
AI summary
A machine learning device acquires decision data including an adjustment time of a position and an orientation of an optical part and a state of light measured by a light measurement device. The machine learning device includes a learning unit that learns a procedure for adjusting the position and the orientation of the optical part. The learning unit includes a reward calculation unit that calculates a reward based on the alignment adjustment time and the state of light, and a value function updating unit that updates a value function based on the reward. The learning unit includes a decision unit that sets a movement method of the optical part based on the value function.