In-Storage Neural Network Acceleration for Vehicle Failure Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automotive maintenance scheduling is based on predetermined milestones, which can lead to inconvenient and potentially unsafe breakdowns of vehicle components, as they often require immediate attention without prior indication of impending failure.
Innovation Solution
Implementing a system that uses artificial neural networks, such as spiking neural networks, to analyze sensor data from vehicles to predict component failures, allowing for proactive maintenance scheduling and reducing the likelihood of breakdowns by identifying deviations from normal operating patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional predetermined milestone-based maintenance scheduling is used, then maintenance services can be scheduled regularly, but vehicle components may break down unexpectedly without prior indication of failure
Solution Approach 1:
The system performs preliminary analysis of sensor data using neural networks to predict component failures before they occur. By analyzing patterns in sensor data and identifying deviations from normal operating conditions, the system can schedule maintenance services in advance, preventing unexpected breakdowns while maintaining manageable system complexity through automated prediction algorithms
2Speed
If neural network computation is performed using conventional processors, then the system can operate with standard computing hardware, but computation speed is insufficient for real-time predictive maintenance
Solution Approach 1:
The patent replaces conventional processor-based neural network computation with a neural network accelerator that uses specialized hardware circuits to perform computations. This substitution of general-purpose processing with dedicated hardware acceleration achieves real-time computation speeds required for predictive maintenance while managing hardware complexity through purpose-built acceleration architecture
3Measurement precision
If more sensor data is collected for analysis, then prediction accuracy improves, but data processing time and computational load increase
Solution Approach 1:
The system continuously collects and pre-processes sensor data in the background, maintaining updated neural network models ready for immediate prediction. This preliminary data preparation and continuous monitoring enable rapid real-time predictions without increasing processing delays, as the computational work is performed incrementally before critical decisions are needed
Solution Approach 2:
The neural network accelerator uses specialized hardware circuits that can process large volumes of sensor data in parallel at high speeds. This hardware acceleration reduces data processing time significantly compared to conventional processors, enabling the system to handle increased data collection for improved prediction accuracy without suffering from processing bottlenecks
Data Source
AI summary
Systems, methods and apparatus of accelerating neural network computations of predictive maintenance of vehicles. For example, a data storage device of a vehicle includes: a host interface configured to receive a sensor data stream from at least one sensor configured on the vehicle; at least one storage media component having a non-volatile memory to store at least a portion of the sensor data stream; a controller; and a neural network accelerator coupled to the controller. The neural network accelerator is configured to perform at least a portion of computations based on an artificial neural network and the sensor data stream to predict a maintenance service of the vehicle.