In-Memory Deep Learning Accelerator for Edge Factory Hazard Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing factory safety monitoring systems face challenges in efficiently processing large volumes of data from various sensors due to limited computational resources and high energy consumption, necessitating improved data processing and analysis methods.
Innovation Solution
Implementing a Deep Learning Accelerator (DLA) integrated with random access memory for in-memory neuromorphic computing, which includes a set of programmable hardware logic units optimized for parallel vector and matrix calculations, allowing for efficient distribution and local processing of Artificial Neural Network computations near sensor locations, reducing energy consumption and computation time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by stationary object
If data is processed centrally in factory safety monitoring systems, then system coordination is simplified, but energy consumption increases and computation time increases
Solution Approach 1:
The patent divides the factory safety monitoring system into distributed edge computing nodes that process data locally rather than centralizing all processing. Each node segments the overall computational task, performing local anomaly detection and feature extraction independently, which reduces energy consumption and computation time while maintaining system coordination through standardized communication protocols.
2Productivity
If computational resources are increased for processing sensor data, then data processing speed improves, but energy consumption increases
Solution Approach 1:
The patent implements local quality by deploying specialized edge computing nodes with optimized neural network accelerators at strategic locations within the factory. Each node is tailored to process specific types of sensor data locally, providing high processing speed for critical real-time monitoring tasks while consuming less energy than a centralized high-performance computing system would require.
3Measurement precision
If more computational power is allocated to anomaly detection, then detection accuracy improves, but system resource consumption increases
Solution Approach 1:
The patent applies partial action by implementing a two-stage processing approach: first, lightweight filtering and feature extraction are performed at the edge nodes to identify potential anomalies, then only suspicious cases are escalated to more computationally intensive centralized analysis. This provides high detection accuracy for critical anomalies while conserving system resources by avoiding full computational power for all data points.
Data Source
AI summary
Systems, devices, and methods related to safety monitoring in a factory using an artificial neural network are described. For example, the system can use a plurality of sensors installed at different locations of a factory to generate a plurality of streams of sensor data. At least one memory device can be configured in the system to perform matrix computations of the artificial neural network according to the plurality of streams of sensor data written into at least one memory device. Based on an output of the artificial neural network responsive to the plurality of streams of sensor data, the system generates an event identification representative of a hazard or anomaly in the factory and activates safety control or notification responsive to the event identification.