Memristor In-Memory Profiling for Electronic Component RUL Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power electronics systems lack efficient methods to evaluate and predict the remaining useful life (RUL) of components, leading to concerns about reliability and longevity, especially in applications like renewable energy and solid-state transformers.
Innovation Solution
An electronic device profiling network utilizing an in-memory computation system with a memristor array and control circuitry to profile electronic devices by performing computations directly in memory, reducing data transfer overhead and enhancing energy efficiency, computing parallelism, and predicting RUL based on component states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If separate CPU and memory units are used for computation, then device complexity is reduced, but energy overhead increases exponentially due to data transfer
Solution Approach 1:
The patent merges the CPU and memory units into a single integrated system where computation is performed directly within the memory array. The memristor array serves dual purposes as both storage and computation unit, eliminating the need for separate CPU and memory components. This consolidation resolves the contradiction by reducing energy overhead from data transfer while maintaining manageable device complexity through unified architecture.
Solution Approach 2:
The patent introduces control circuitry as an intermediary between the sensing system and the memristor array. This control circuitry manages the computation process, applies input signals, and reads output signals, thereby simplifying the overall system architecture while enabling efficient in-memory computation without requiring complex separate processing units.
2Device complexity
If data is transferred between separate CPU and memory units, then computation and storage are separated, but latency increases
Solution Approach 1:
By merging storage and computation functions into the same memory array, the patent eliminates data transfer between separate CPU and memory units. The memristor array performs matrix multiplication computations directly on stored data, completely removing the latency associated with data movement while maintaining a manageable system architecture through functional integration.
3Device complexity
If traditional von Neumann architecture is used, then device complexity is low, but energy efficiency decreases due to data transfer overhead
Solution Approach 1:
The patent replaces the traditional von Neumann architecture with a memcomputing system that uses memristor-based in-memory computation. This substitution eliminates the mechanical/electrical data transfer process between separate CPU and memory units, performing computations directly where data is stored. The result is dramatically improved energy efficiency while the system architecture remains manageable through unified design.
4Measurement precision
If more computing resources are allocated for RUL prediction, then prediction accuracy improves, but power consumption increases
Solution Approach 1:
The patent merges storage and computation into a single memristor array system, eliminating the energy-consuming data transfer between separate processing and storage units. This integration enables accurate RUL prediction through direct in-memory matrix operations while significantly reducing overall power consumption compared to traditional architectures that would require more computing resources for the same predictive accuracy.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system effectively reduces power consumption, computing resource utilization, and latency while increasing energy efficiency and computing parallelism, enabling accurate prediction of RUL and preventing unsafe operations by notifying performance degradation.
Implementation Method 1
Applying the input signals may trigger change in electrical attributes across the memristor cells in accordance with certain electrical rules such as Ohm's law. The changed electrical attributes may result in third electrical attribute magnitudes corresponding to third electrical attributes (e.g., current or voltage).
Data Source
AI summary
A profiling system includes an in-memory computation system and an electronic device profiling system. The in-memory computation system includes computation circuitry and memristor control circuitry, the in-memory computation system profiling one or more electronic components that constitute an electronic device by computing component states of the electronic components based on parameters of the electronic components and a function defining a transformation of the parameters to the component states. The parameters are indicative of operating characteristics. The component states are indicative of states of health of the electronic components. The electronic device profiling system comprising device profiling circuitry, based on the component states at different time instances, predicts future component states of the electronic components, based on the predicted future component states, predicts one or more future device states of the electronic device, and outputs information of the predicted one or more future device states.