Edge ML Model Scheduling for Selective Sensor Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing quality check systems in manufacturing fail to detect errors in real-time due to the high cost of installing sensors and edge gateways, and lack consideration of apriori probability of process errors, leading to delayed detection and unnecessary propagation of defective parts.
Innovation Solution
A system and method that intelligently selects a subset of sensors for observation based on estimated error probabilities, using machine learning models to detect anomalies and dynamically adjust sensor selection based on detected errors, optimizing resource usage on edge devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensors are installed after each process to detect errors early, then error detection capability is improved, but system cost and complexity increase significantly
Solution Approach 1:
The patent extracts and focuses monitoring resources only on the most critical process steps where errors are most likely to occur and have the greatest impact. Instead of deploying sensors after every process, the system identifies and monitors only the subset of processes with highest error probability, thereby reducing system complexity while maintaining effective error detection capability.
Solution Approach 2:
The system performs partial monitoring by selecting and observing only certain process steps rather than all processes. The patent implements a selective monitoring approach where resources are allocated to monitor a fraction of processes that are deemed most important based on error probability analysis, achieving effective quality control with reduced resource expenditure.
2Reliability
If all sensors observe all process steps continuously, then comprehensive error detection is achieved, but resource consumption increases
Solution Approach 1:
The system implements partial observation by having sensors monitor only specific process steps rather than all processes continuously. The patent dynamically selects which process steps require observation based on real-time error probability assessments, enabling comprehensive error detection for critical steps while conserving computational and energy resources by not monitoring all processes at all times.
Solution Approach 2:
The monitoring system is dynamic rather than static. The patent employs machine learning models that continuously assess error probabilities and dynamically adjust which process steps require sensor observation. This dynamic allocation of monitoring resources allows the system to adapt to changing conditions, maintaining comprehensive detection capability when needed while reducing resource consumption during normal operation.
3Ease of manufacture
If fixed image-based inspection is used at final process, then implementation simplicity is maintained, but error detection timing is delayed
Solution Approach 1:
The system performs preliminary error detection by monitoring critical process steps before the final assembly is complete. The patent uses machine learning models to predict potential errors early in the manufacturing process, allowing corrective actions to be taken before defects propagate through subsequent processes, thereby reducing rework time and improving overall manufacturing efficiency while maintaining implementation feasibility.
4Productivity
If sensors monitor only high-probability error processes, then resource efficiency is improved, but detection coverage may be reduced
Solution Approach 1:
The system implements continuous feedback loops where sensor data from monitored processes feeds into machine learning models that update error probability assessments. The patent uses this feedback to dynamically adjust which processes require monitoring, ensuring that resources are efficiently allocated to high-risk areas while maintaining adequate detection coverage. The feedback mechanism allows the system to learn from observed errors and adapt its monitoring strategy to maintain comprehensive coverage with optimized resource usage.
Data Source
AI summary
Systems and methods described herein can involve management of a system having a plurality of sensors, the plurality of sensors observing a plurality of process steps, which can involve selecting a subset of the plurality of sensors for observation; executing anomaly detection from data provided from the subset of the plurality of sensors; for a detection of an anomaly from a sensor from the subset of sensors, selecting ones of the plurality of process steps based on the detected anomaly; estimating a probability of anomaly occurrence for the selected ones of the plurality of process steps; and for the estimated probability of anomaly occurrence meeting a predetermined criteria, selecting ones of the plurality of sensors associated with the selected ones of the plurality of process steps for observation.