# US 2026/0065772 A1 · Alarm management systems and methods

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0065772 A1
- Document ID: US20260065772A1
- Status: Published application
- Kind: A1
- Application number: 18/821,796
- Filed: 2024-08-30
- Published: 2026-03-05
- Earliest claimed: 2024-08-30
- Assignees: Hitachi, Ltd.
- Applicants: HITACHI, Ltd.
- Inventors: Pegah MAVAIE; Ravigopal VENNELAKANTI; Malarvizhi SANKARANARAYANASAMY
- Web page: https://patentgrep.com/patent/US20260065772A1

## CPC

- G08B 29/00: Checking or monitoring of signalling or alarm systems; Prevention or correction of operating errors, e.g. preventing unauthorised operation
- G06F 40/14: Tree-structured documents (parsing G06F40/205; validation G06F40/226)
- G06F 40/205: Parsing
- G08B 25/006: {Alarm destination chosen according to type of event, e.g. in case of fire phone the fire service, in case of medical emergency phone the ambulance}
- G08B 29/186: {Fuzzy logic; neural networks}
- G08B: SIGNALLING SYSTEMS, e.g. PERSONAL CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Systems and methods for managing alarms in a process control environment involve receiving, at an LLM agent, input data from various data sources including alarm data and a parsing result from a query parser that generates the parsing result from a user query in a natural language format. To reduce the likelihood of reasoning errors, the parsing result and input data are provided to an LLM that selects an analytical function, which includes domain-specific information. The LLM uses the input data and analytical function to perform an alarm analysis, enhancing the accuracy of the analysis result. The analysis result is then used to generate an alarm summary, which reflects a severity or urgency of an alarm.

## Claims

### Claim 1

A system for managing alarms in a process control environment, the system comprising:
a large language model (LLM) agent configured to perform steps comprising:
receiving input data from various data sources comprising alarm data;
receiving a parsing result from a query parser that is configured to generate the parsing result from a user query in a natural language format; and
reducing a likelihood of a reasoning error by providing the parsing result and the input data to a first large language model (LLM) that is configured to select an analytical function from a set of pre-defined analytical functions that each comprises domain-specific information;
an LLM coupled to the LLM agent, the LLM configured to perform steps comprising:
in response to receiving the input data and the analytical function, performing an alarm analysis to obtain an analysis result, wherein the analytical function enhances an accuracy of the analysis result;
using the analysis result to generate an alarm summary that reflects at least one of a severity or an urgency associated with an alarm; and
communicating the alarm summary to the LLM agent; and
a dashboard organizer that is coupled to the LLM agent, the dashboard organizer configured to cause the alarm summary to be displayed on a user interface.

### Claim 2 (depends on 1)

The system of claim 1, wherein the alarm analysis comprises at least one of a community detection, an alarm co-occurrence analysis, a predictive analysis, or a survival analysis.

### Claim 3 (depends on 2)

The system of claim 2, wherein the alarm co-occurrence analysis comprises performing a time-related analysis and identifying alarms that frequently occur together to identify related issues or systemic problems.

### Claim 4 (depends on 2)

The system of claim 2, wherein the survival analysis comprises evaluating a persistence or resolution of a set of alarms over time to determine system health or to identify a time-related system issue.

### Claim 5 (depends on 1)

The system of claim 1, wherein the LLM agent is configured to automatically integrate a new analytical function or API into a toolset to reduce additional training or fine-tuning.

### Claim 6 (depends on 1)

The system of claim 1, wherein the user interface comprises cascading style sheets (CSS) that enhance at least one of an aesthetic or a usability of a customizable dashboard.

### Claim 7 (depends on 1)

The system of claim 1, wherein at least one of a layout or a style of the user interface is customizable according to user-preference.

### Claim 8 (depends on 1)

The system of claim 1, wherein the LLM agent is further configured to identify a pattern or a cluster within the alarm data to detect a systemic issue.

### Claim 9 (depends on 1)

The system of claim 1, wherein the input data comprises historical alarm data that is used to forecast a future alarm occurrence.

### Claim 10 (depends on 1)

The system of claim 1, wherein the alarm analysis comprises prioritizing alarms based on a user preference.

### Claim 11

A method for managing alarms in a process control environment, the method comprising:
at a large language model (LLM) agent, performing steps comprising:
receiving input data from various data sources comprising alarm data:
receiving a parsing result from a query parser that is configured to generate the parsing result from a user query in a natural language format; and
reducing a likelihood of a reasoning error by providing the parsing result and the input data to an LLM configured to select an analytical function from a set of pre-defined analytical functions that each comprises domain-specific information; and
at an LLM, performing steps comprising:
in response to receiving the input data and the analytical function, performing an alarm analysis to obtain an analysis result, wherein the analytical function enhances an accuracy of the analysis result;
using the analysis result to generate an alarm summary that reflects at least one of a severity or an urgency associated with an alarm; and
communicating the alarm summary to the LLM agent; and
using a dashboard organizer to cause the alarm summary to be displayed on a user interface.

### Claim 12 (depends on 11)

The method of claim 11, wherein the alarm analysis comprises at least one of a community detection, an alarm co-occurrence analysis, a predictive analysis, or a survival analysis.

### Claim 13 (depends on 12)

The method of claim 12, wherein the survival analysis comprises evaluating a persistence or resolution of a set of alarms over time to determine system health or to identify a time-related system issue.

### Claim 14 (depends on 11)

The method of claim 11, wherein the LLM agent is configured to automatically integrate a new analytical function or API into a toolset to reduce additional training or fine-tuning.

### Claim 15 (depends on 11)

The method of claim 11, wherein the user interface comprises cascading style sheets (CSS) that enhance at least one of an aesthetic or a usability of a customizable dashboard.

### Claim 16 (depends on 11)

The method of claim 11, wherein at least one of a layout or a style of the user interface is customizable according to user-preference.

### Claim 17 (depends on 11)

The method of claim 11, wherein the LLM agent is further configured to identify patterns or clusters within the alarm data to detect a systemic issue.

### Claim 18 (depends on 11)

The method of claim 11, wherein the input data comprises historical alarm data that is used to forecast a future alarm occurrence.

### Claim 19 (depends on 11)

The method of claim 11, wherein the alarm analysis comprises prioritizing alarms based on a user preference.

### Claim 20 (depends on 11)

The method of claim 11, wherein the method evaluates a persistence or a resolution of alarms over time to assess an overall health of a system.
