Circuit Breaker Arc Fault Detection Using ML Accumulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing arc fault circuit interrupters (AFCIs) rely on internal electronics to analyze analog signals for arc fault detection, but they do not effectively utilize machine learning classifiers to accumulate inferences over time, leading to potential unwanted tripping of circuit breakers in residential products.
Innovation Solution
A circuit breaker with a microcontroller that samples RSSI, voltage, and current signals, performs pre-processing, and inputs the data into a machine learning classifier to determine the probability of an arc fault, incrementing or decrementing an accumulator based on the classifier's output, and trips the circuit if the accumulator exceeds a threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional analog signal analysis is used in AFCI, then arc fault detection capability is provided, but unwanted tripping on residential products occurs
Solution Approach 1:
The system dynamically adjusts the accumulator threshold and accumulation rate based on learned patterns from training data representing different load conditions. The machine learning classifier adapts the detection criteria dynamically rather than using fixed thresholds, allowing the system to distinguish between normal residential appliance operation and actual arc faults, thereby reducing unwanted tripping while maintaining detection accuracy
Solution Approach 2:
The patent transforms the detection approach by changing from fixed analog threshold analysis to probabilistic machine learning classification. The system uses trained models that output probability scores, which are then accumulated and compared against adaptive thresholds. This parameter transformation enables more nuanced discrimination between hazardous arcs and benign signals from residential products
2Measurement precision
If machine learning classifier is implemented, then arc fault detection accuracy is improved, but device complexity increases
Solution Approach 1:
The detection system is segmented into distinct functional modules: signal acquisition, pre-processing, machine learning classification, accumulator management, and trip decision. This modular architecture allows the complex machine learning functionality to be implemented as a separate, manageable component that can be trained and updated independently without redesigning the entire system
Solution Approach 2:
The accumulator acts as an intermediary between the machine learning classifier and the trip circuit. It accumulates probability scores over time and compares against thresholds, providing a buffer that smooths out classifier variations and simplifies the final trip decision logic. This intermediary layer reduces the complexity of directly translating classifier output into trip decisions
3Speed
If accumulator threshold is set low, then arc fault detection speed is improved, but false tripping increases
Solution Approach 1:
The system uses periodic sampling of the analog signals and periodic updates to the accumulator based on classifier probability scores. This periodic accumulation allows the system to build confidence in arc fault detection over multiple cycles, enabling faster response when arcs are confirmed while filtering out transient false indications through the cumulative nature of the approach
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A circuit breaker with arc fault detection by accumulation of machine learning classifications is provided. The circuit breaker comprises a microcontroller including a processor, a memory and computer-readable software code which, when executed by the processor, causes the microcontroller to: sample analog signals representing one or more of the following: a RSSI signal, a voltage signal, and a current signal, perform multiple pre-processing steps on the analog signals to derive a data set, and input the data set into a machine learning classifier such that an output of the machine learning classifier is a value between 0 and 1 which represents a percent chance that the data set is from an electrical arc. Based on the value of the percent chance an accumulator value is either incremented or decremented and if the accumulator value passes an upper threshold level, the microcontroller sends a signal to trip open the circuit breaker.