Systems, methods, and devices for reducing systemic risks

Patent metadata

Field	Value
Publication number	US-12164367-B2
Application number	US-202017126148-A
Country	US
Kind code	B2
Filing date	Dec 18, 2020
Priority date	Dec 18, 2020
Publication date	Dec 10, 2024
Grant date	Dec 10, 2024

Abstract

A computer-implemented method may include obtaining, from a system using a middleware component of the system, run-time evidence of the system; applying the obtained run-time evidence to a Directed Acyclic Graph (DAG) Bayesian network to determine marginal probabilities for one or more nodes of the DAG Bayesian network, wherein the DAG Bayesian network comprises a plurality of nodes each representing states and faults of the system, wherein each node includes a parameterized conditional probability distribution, and wherein one or more of the nodes of the plurality of nodes specify a list of one or more safety goals and a safety value; determining which nodes representing faults have probabilities exceeding their specified safety value; and determining one or more risk mitigation techniques to activate for the determined nodes representing faults with probabilities exceeding their respective safety value.

First claim

What is claimed is:

1. A computer-implemented method comprising:
obtaining, from a system using a middleware component of the system, run-time evidence of the system;
applying the obtained run-time evidence to a Directed Acyclic Graph (DAG) Bayesian network to determine marginal probabilities for one or more nodes of the DAG Bayesian network;
determining which of the nodes of the DAG Bayesian network representing faults have probabilities exceeding their specified safety value; and
determining one or more risk mitigation techniques to activate for the determined nodes,
wherein the DAG Bayesian network comprises a plurality of nodes each representing states and faults of the system, wherein each node includes a parameterized conditional probability distribution, and wherein one or more of the nodes of the plurality of nodes specify a list of one or more safety goals and a safety value,
wherein the parameterized conditional probability distribution indicates an a-priori probability of the node representing a set of possible values indicating a presence or absence of a particular fault or system state,
wherein the parameterized conditional probability distribution for each respective node of the DAG Bayesian Network further indicates a conditional probability distribution of affected nodes including all possible value combinations comprising values from mitigation techniques identified from a look-up-table (LUT) data structure indicating a mapping of one or more risk mitigation techniques to one or more of the nodes.

2. The computer implemented method of claim 1 , wherein the parameterized conditional probability distribution is conditioned by a combination of possible values of all incoming dependencies.

3. The computer implemented method of claim 1 , wherein determining the one or more risk mitigation techniques to activate comprises:
identifying likely fault sources for the nodes representing a fault and determined to have a probability exceeding its safety value; and
determining risk mitigation techniques for the identified likely fault sources from the LUT data structure.

4. The computer implemented method of claim 3 , wherein identifying the likely fault sources for the nodes representing a fault and determined to have a probability exceeding its safety value comprises:
determining a joint probability distribution of all incoming dependencies for each node; and
identifying for each node, using the determined joint probability distribution, one or more most probable fault sources for the node representing a fault exceeding its specified safety value.

5. The computer implemented method of claim 4 , wherein determining a joint probability distribution for each node representing faults exceeding its specified safety value comprises using the parameterized conditional probability distribution and the determined probability of each of its parent nodes.

6. The computer implemented method of claim 1 , wherein determining the one or more risk mitigation techniques to activate, for each respective node representing a fault determined to exceed its specified safety value comprises performing for one or more iterations:
selecting a set of candidate mitigation techniques from the LUT data structure based on the identified fault sources for the respective node;
conditioning the parameterized conditional probability distribution based on the selected set of candidate mitigation techniques;
determining a new marginal probability distribution using the conditioned parameterized conditional probability distribution; and
determining whether the probability of the respective node is less than the safety value based on the newly determined marginal probability distribution.

7. The computer implemented method of claim 6 further comprising:
determining that marginal probability of the respective node is less than the safety value based on the newly determined marginal probability distribution, and
activating the selected set of candidate mitigation techniques associated with the newly determined marginal probability distribution.

8. A diagnostic system comprising: a Directed Acyclic Graph (DAG) Bayesian Network stored in an accessible storage device corresponding to a fault model of a system; a middleware component configured to obtain run-time evidence; at least one processor operably coupled to the middleware component and the DAG Bayesian Network, the at least one processor configured to: obtain, from the middleware component, the obtained run-time evidence; determine marginal probabilities for one or more nodes of the DAG Bayesian Network; determine which nodes of the DAG Bayesian Network representing faults have probabilities exceeding their specified safety value; and determine one or more risk mitigation techniques to activate for the determined nodes representing faults with probabilities exceeding their respective safety value, wherein the DAG Bayesian network comprises a plurality of nodes each representing states and faults of the system, wherein each node includes a parameterized conditional probability distribution, and wherein one or more the nodes of the plurality of nodes specify a list of safety goals with a safety value, and wherein to determine marginal probabilities for one or more nodes of the DAG Bayesian Network comprises to apply the obtained run-time-evidence to the DAG Bayesian network wherein to determine the one or more risk mitigation techniques to activate, for each respective node representing a fault determined to exceed its specified safety value comprises the at least one processor to perform the following for one or more iterations: select a set of candidate mitigation techniques from a LUT data structure based on identified fault sources for the respective node; condition the parameterized conditional probability distribution based on the selected set of candidate mitigation techniques; determine a new marginal probability distribution using the conditioned parameterized conditional probability distribution; and determine whether probability of the respective node is less than the safety value based on the newly determined marginal probability distribution.

9. The diagnostic system of claim 8 , wherein the at least one processor configured to determine marginal probabilities for one or more nodes of the DAG Bayesian Network comprises the at least one processor to apply the obtained run-time-evidence to the DAG Bayesian network.

10. The diagnostic system of claim 9 , wherein the parameterized conditional probability distribution indicates an a-priori probability of the node representing a set of possible values indicating a presence or absence of a particular fault or system state.

11. The diagnostic system of claim 10 , wherein the parameterized conditional probability distribution is conditioned by a combination of possible values of all incoming dependencies.

12. The diagnostic system of claim 10 , further comprising: a look-up-table (LUT) data structure stored in a storage device including data indicating a mapping one or more risk mitigation techniques to one or more of the nodes, the at least one processor further operably coupled to the LUT data structure, wherein the parameterized conditional probability distribution for each respective node of the DAG Bayesian Network further indicates a conditional probability distribution of affected nodes including all possible value combinations comprising values from mitigation techniques identified from the LUT data structure, wherein the DAG Bayesian network comprises a plurality of nodes each representing states and faults of the system, wherein each node includes a parameterized conditional probability d

Assignees

• Intel Corp

Inventors

• Rosales Rafael
• Paulitsch Michael
• González Aguirre David Israel
• Geissler Florian
• Graefe Ralf

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06F11/004

  Error avoidance (G06F11/07 and subgroups take precedence) · CPC title

• G06F11/0709

  in a distributed system consisting of a plurality of standalone computer nodes, e.g. clusters, client-server systems · CPC title

• G06F16/9024

  Graphs; Linked lists (G06F16/9027 takes precedence) · CPC title

• G06F11/0751

  Error or fault detection not based on redundancy (power supply failures G06F1/30; network fault management H04L41/06) · CPC title