Fine-grained causal anomaly inference for complex system fault diagnosis
A computer-implemented method for diagnosing system faults by fine-grained causal anomaly inference is presented. The computer-implemented method includes identifying functional modules impacted by causal anomalies and backtracking causal anomalies in impaired functional modules by a low-rank network diffusion model. An invariant network and a broken network are inputted into the system, the invariant network and the broken network being jointly clustered to learn a degree of broken severities of different clusters as a result of fault propagations.
1. A computer-implemented method executed on a processor for diagnosing system faults by fine-grained causal anomaly inference, the method comprising:
inputting an invariant network and a broken network;
identifying functional modules impacted by causal anomalies; and
backtracking causal anomalies in impaired functional modules by a low-rank network diffusion model;
wherein the invariant and broken networks are processed by a network clustering and broken cluster score learning module and a causal anomaly inference module; and
wherein the network clustering and broken cluster score learning module includes an invariant network clustering module employing stochastic matrix decomposition and a broken cluster learning module for learning a broken score for each cluster in the broken network to indicate a broken degree for each cluster.
2. A computer-implemented method executed on a processor for diagnosing system faults by fine-grained causal anomaly inference, the method comprising:
inputting an invariant network and a broken network;
identifying functional modules impacted by causal anomalies; and
backtracking causal anomalies in impaired functional modules by a low-rank network diffusion model;
wherein the invariant and broken networks are processed by a network clustering and broken cluster score learning module and a causal anomaly inference module; and
wherein the causal anomaly inference model includes a label propagation module, a reconstruction error module, and an optimization module for computing causal anomaly vectors for each cluster in the invariant network.
3. A computer-implemented method executed on a processor for diagnosing system faults by fine-grained causal anomaly inference, the method comprising:
inputting an invariant network and a broken network;
identifying functional modules impacted by causal anomalies; and
backtracking causal anomalies in impaired functional modules by a low-rank network diffusion model;
wherein the invariant and broken networks are processed by a network clustering and broken cluster score learning module and a causal anomaly inference module;
wherein the causal anomaly inference model includes a label propagation module, a reconstruction error module, and an optimization module for computing causal anomaly vectors for each cluster in the invariant network; and
wherein clustering results, broken cluster score results, and anomaly score results are combined into a causal anomaly score vector to reflect multiple fault propagations in different clusters in the invariant network.
4. A system for diagnosing system faults by fine-grained causal anomaly inference, the system comprising:
a memory; and
a processor in communication with the memory, wherein the processor is configured to:
identify functional modules impacted by causal anomalies; and
backtrack causal anomalies in impaired functional modules by a low-rank network diffusion model;
wherein an invariant network and a broken network are inputted into the system;
wherein the invariant and broken networks are processed by a network clustering broken cluster score learning module and a causal anomaly inference module; and
wherein the network clustering and broken cluster score learning module includes an invariant network clustering module employing stochastic matrix decomposition and a broken cluster learning module for learning a broken score for each cluster in the broken network to indicate a broken degree for each cluster.
5. A system for diagnosing system faults by fine-grained causal anomaly inference, the system comprising:
a memory; and
a processor in communication with the memory, wherein the processor is configured to:
identify functional modules impacted by causal anomalies; and
backtrack causal anomalies in impaired functional modules by a low-rank network diffusion model;
wherein an invariant network and a broken network are inputted into the system;
wherein the invariant and broken networks are processed by a network clustering broken cluster score learning module and a causal anomaly inference module; and
wherein the causal anomaly inference module includes a label propagation module, a reconstruction error module, and an optimization module for computing causal anomaly vectors for each cluster in the invariant network.
6. A system for diagnosing system faults by fine-grained causal anomaly inference, the system comprising:
a memory; and
a processor in communication with the memory, wherein the processor is configured to:
identify functional modules impacted by causal anomalies; and
backtrack causal anomalies in impaired functional modules by a low-rank network diffusion model;
wherein an invariant network and a broken network are inputted into the system;
wherein the invariant and broken networks are processed by a network clustering broken cluster score learning module and a causal anomaly inference module;
wherein the causal anomaly inference module includes a label propagation module, a reconstruction error module, and an optimization module for computing causal anomaly vectors for each cluster in the invariant network; and
wherein clustering results, broken cluster score results, and anomaly score results are combined into a causal anomaly score vector to reflect multiple fault propagations in different clusters in the invariant network.
7. A non-transitory computer-readable storage medium comprising a computer-readable program for diagnosing system faults by fine-grained causal inference, wherein the computer-readable program when executed on a computer causes the computer to perform the steps of:
identifying functional modules impacted by causal anomalies; and
backtracking causal anomalies in impaired functional modules by a low-rank network diffusion model;
wherein the network clustering and broken cluster score learning module includes an invariant network clustering module employing stochastic matrix decomposition and a broken cluster learning module for learning a broken score for each cluster in the broken network to indicate a broken degree for each cluster; and
wherein the causal anomaly inference module includes a label propagation module, a reconstruction error module, and an optimization module for computing causal anomaly vectors for each cluster in the invariant network.