Sensor-Based Quality Prediction for Manufacturing Process Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity and high cost of quality control checks in manufacturing processes, particularly for products like circuit boards, where imaging methods such as X-ray radiation are often required, lead to increased complexity and resource allocation.
Innovation Solution
A method utilizing a learning-capable algorithm to calculate a quality indicator based on production data from sensors, which can associate the indicator with the product, potentially omitting the need for a quality control check by using a computer program product and apparatus that includes sensors, a computing unit, and a database to determine if the product meets quality requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional quality control checks using imaging methods (e.g., X-ray radiation) are employed, then product quality assessment is achieved, but device complexity and cost increase significantly
Solution Approach 1:
The patent replaces complex imaging-based quality control systems (X-ray, optical imaging) with a computational system that uses sensor data and machine learning algorithms to predict quality indicators. This substitutes physical measurement infrastructure with information processing, thereby reducing device complexity while maintaining quality assessment capability.
Solution Approach 2:
Instead of directly imaging the product to assess quality, the system creates a digital model (quality indicator) that copies or represents the quality state based on sensor measurements and production parameters. This virtual copy eliminates the need for complex physical inspection equipment.
2Reliability
If traditional quality control checks using imaging methods are employed, then product quality assessment is achieved, but production cost increases significantly
Solution Approach 1:
The patent uses inexpensive sensors and computational resources instead of expensive imaging equipment. The quality control function is achieved through low-cost data acquisition and processing, making the system economically viable without sacrificing assessment reliability.
Solution Approach 2:
The patent replaces expensive physical inspection infrastructure (X-ray machines, optical imaging systems) with computational algorithms running on standard hardware. This substitution dramatically reduces capital expenditure and operational costs while maintaining quality assessment capability.
3Reliability
If imaging methods are used for quality control, then internal product quality can be assessed, but radiation exposure and safety concerns arise
Solution Approach 1:
The patent replaces radiation-based imaging methods with non-invasive sensor measurements and computational modeling. This substitution eliminates harmful radiation exposure while maintaining the ability to assess internal product quality through indirect measurement and prediction.
Solution Approach 2:
The system uses sensor data and production parameters as intermediaries to infer internal quality characteristics without directly exposing the product to harmful radiation. The machine learning model acts as a mediator that translates safe, non-invasive measurements into quality assessments.
Data Source
AI summary
The disclosure relates to a method and a device for producing a product and to a computer program product. The product is produced in at least one production step. A quality control check is optionally carried out after at least one of the production steps to determine a quality index of the product in question. To save on the quality control check, a quality indicator of the product in question is determined using production data. The production data are advantageously provided by sensors. The quality indicator of the product in question may be calculated using an adaptive algorithm. The adaptive algorithm may be taught and/or improved using quality indices of a quality control unit and the corresponding production data. The adaptive algorithm may be taught with the aid of a further computing unit, in particular in a cloud.