Natural language dialogue type risk management system based on artificial intelligence

Abstract

The invention relates to the technical field of risk management systems, and discloses an artificial intelligence-based natural language dialogue type risk management system, which comprises a data layer, a model layer, a control layer, an application layer and an interaction layer, and is characterized in that the data layer is in signal connection with the model layer, the model layer is in signal connection with the control layer, the control layer is in signal connection with the application layer, and the interaction layer is in signal connection with the application layer. And the application layer is in signal connection with the interaction layer. According to the natural language dialogue type risk management system based on artificial intelligence, by integrating data of multiple aspects such as building fireproof monitoring, unit fire safety management, fire-fighting equipment and equipment state monitoring, law enforcement historical records and third-party maintenance, detection and evaluation, the system can comprehensively master fire safety related conditions; compared with a traditional mode of evaluating depending on a single or partial data source, the method can avoid evaluation deviation caused by information loss or one-sided information, enables a safety evaluation result to be more suitable for the actual situation, and more accurately reflects potential safety hazards.

Description

Technical Field

[0001] The present invention relates to the technical field of risk management systems, and in particular to a natural language conversational risk management system based on artificial intelligence. Background Art

[0002] The building fire risk management system is a comprehensive system for comprehensive management, evaluation, and prevention of fire-related risks in buildings. It can use cutting-edge technologies such as artificial intelligence and big data to more accurately identify risks and predict fire trends.

[0003] At present, the building fire risk management system used can only evaluate a single data during use. The missing or one-sided information during data analysis leads to data evaluation deviation. At the same time, the current building fire risk management system has low working efficiency. Therefore, it is necessary to invent a natural language conversational risk management system based on artificial intelligence to solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to provide a natural language conversational risk management system based on artificial intelligence to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a natural language conversational risk management system based on artificial intelligence, comprising a data layer, a model layer, a control layer, an application layer and an interaction layer, wherein the data layer is signal-connected to the model layer, the model layer is signal-connected to the control layer, the control layer is signal-connected to the application layer, and the application layer is signal-connected to the interaction layer.

[0006] The data layer includes a data collection module and a data storage module, and the data collection module and the data storage module are connected by signals.

[0007] The model layer includes a large language model module and a model deployment module, and the large language model module and the model deployment module are signal-connected.

[0008] The control layer includes an input processing module and an output processing module, and the input processing module and the output processing module are signal-connected.

[0009] The application layer includes an identity security management module and a function module, and the identity security management module and the function module are connected by signals.

[0010] The interaction layer includes a hardware interaction module and a software interaction module.

[0011] Preferably, the data collection module includes building fire monitoring data collection, unit fire safety management data collection, fire protection facility and equipment status data collection, law enforcement history data collection, and third-party maintenance data, detection data, and evaluation data collection.

[0012] The building fire monitoring data collection uses smoke detectors and temperature sensors to collect building fire data for real-time monitoring of fire hazards in buildings.

[0013] The unit safety management data collection includes the unit's fire protection facilities, fire personnel configuration and training records, and fire emergency plans, which are used to evaluate the unit's own fire safety management level and its response capabilities when a fire occurs.

[0014] The fire-fighting facilities and equipment status data collection includes water pressure monitoring of fire hydrants, pressure and validity period of fire extinguishers, and operating status data of automatic sprinkler fire extinguishing systems, which are used to ensure the normal operation of fire-fighting facilities.

[0015] The enforcement history data collection includes data recorded by the fire enforcement department during the fire inspection and law enforcement process of the unit, including inspection results, problems found, rectification requirements and penalties, which are used to reflect the fire safety compliance of the unit.

[0016] The collection of law enforcement historical data and third-party maintenance data, inspection data, and evaluation data includes the maintenance, inspection, and evaluation of fire protection facilities by third-party professional organizations, which are used to understand the actual status of fire protection facilities and the fire safety status of the unit.

[0017] Preferably, the large language model module uses the ChatGLM2 large language model and the LORA model, the ChatGLM2 large language model is used to process the building fire monitoring data and unit fire safety management data from the data layer, and the LORA model is used to fine-tune the ChatGLM2 large language model.

[0018] Preferably, the LORA model is represented by a feedforward neural network of the original pre-trained model, and the formula is:

[0019] Among them, x is the input vector, W is the weight matrix, b is the bias vector, and h is the output vector. The modification of LORA fine-tuning is performed by adding two low-rank decomposition matrices. The modified calculation becomes:

[0020]

[0021] Among them, B and A are two low-rank matrices, usually the dimension of B is r×k, the dimension of A is k×n, assuming that the dimension of the original weight matrix W is r×n, k≪r and k≪n.

[0022] Preferably, the input processing module includes receiving input data and preprocessing the data format. The received input data is responsible for receiving input from the application layer, including Prompt prompts, History historical records, and Argument model parameters. The Prompt prompts are used to guide the large language model to analyze and answer in the direction of a specific security assessment. The History historical records can provide past related situations as a reference for the current assessment. The Argument model parameters involve the configuration parameters required for the specific operation of the model, unify the encoding format of the received input data, and adjust the text length and structure.

[0023] Preferably, the output processing module includes result post-processing and feedback to the application layer, the result post-processing is used to generate corresponding evaluation results according to the input data, and the feedback to the application layer is used to accurately feed back the processed valid results to the application layer.

[0024] Preferably, the identity security management module includes identity authentication, permission control and operation auditing. The identity authentication is used to authenticate the user who logs into the system through biometric technology. The permission control is used to assign corresponding system operation permissions to different users based on their roles and responsibilities. The operation auditing is used to record in detail every operation of the user in the system, including the time, content and data involved in the operation.

[0025] Preferably, the functional modules include intelligent risk assessment, daily self-inspection assessment and safety assessment report generation. The intelligent risk assessment utilizes the processed data obtained from the control layer, combined with preset risk assessment rules and algorithms, to conduct intelligent analysis of fire safety related situations and check for various potential safety risks. The daily self-inspection assessment is used to provide corresponding self-inspection process guidance and data entry interface. The safety assessment report generation is based on the comprehensive safety assessment data collected by the system, including various monitoring data, inspection data and model analysis results, and automatically generates a safety assessment report in accordance with standardized report templates and format requirements.

[0026] Preferably, the hardware interaction module includes tablet interaction and AR glasses interaction, wherein the tablet interaction is used for system support to connect with a dedicated tablet, by developing an adaptive driver and application program interface, and the AR glasses interaction is used for inspectors to view the actual scene while superimposing relevant safety assessment information in the field of view through the system.

[0027] Preferably, the software interaction module includes mobile application interaction and PC application interaction, the mobile application interaction is used to interact with the application layer of the system through a network communication protocol for data exchange, and the PC application interaction is used to meet the needs of large-scale data import and export, generate detailed security assessment reports, and perform in-depth configuration functions of the system.

[0028] Compared with the prior art, the present invention provides a natural language conversational risk management system based on artificial intelligence, which has the following beneficial effects: This natural language conversational risk management system based on artificial intelligence can comprehensively grasp the fire safety-related situations by integrating data from building fire monitoring, unit fire safety management, firefighting facilities and equipment status monitoring, law enforcement history records, and third-party maintenance, testing and evaluation. Compared with the traditional evaluation method that relies on a single or partial data source, it can avoid evaluation bias caused by missing or one-sided information, making the safety assessment results more in line with the actual situation and more accurately reflecting potential safety hazards.

[0029] This artificial intelligence-based natural language conversational risk management system uses ChatGLM2, a powerful large language model, and fine-tunes it for safety assessment tasks through the LORA model. The model can deeply understand and analyze various complex fire data, explore deep-seated correlations and risk factors, and accurately identify small fault hazard trends in fire protection facilities and coordination issues between different fire management measures, thereby providing more accurate and detailed analysis results for safety assessments and further improving the accuracy of the overall assessment.

[0030] This natural language conversational risk management system based on artificial intelligence has realized automated functions such as intelligent risk assessment, daily self-inspection and assessment, and security assessment report generation at the application layer. In the past, manual risk assessment and report writing were often time-consuming and labor-intensive, and prone to human errors. Now the system can complete these processes quickly and automatically based on the input data, shortening working hours and enabling security assessment-related work to be carried out more efficiently. Units can complete multiple comprehensive security self-inspections and assessments in a shorter time.

[0031] This natural language conversational risk management system based on artificial intelligence supports multiple interaction methods such as dedicated tablets, AR glasses, mobile applications and PC applications at the interactive layer. Both on-site inspectors and managers in the office can easily interact with the system. Inspectors can obtain and enter information in real time on site through dedicated tablets or AR glasses, without the need for cumbersome paper records and subsequent repeated entry work, making the entire safety assessment work smoother, reducing unnecessary waste of time and improving overall work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative labor: Figure 1 Schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0035] See also Figure 1 The present invention provides a technical solution: a natural language conversational risk management system based on artificial intelligence, including a data layer, a model layer, a control layer, an application layer and an interaction layer, the data layer is signal-connected with the model layer, the model layer is signal-connected with the control layer, the control layer is signal-connected with the application layer, and the application layer is signal-connected with the interaction layer.

[0036] The data layer includes a data collection module and a data storage module, and the data collection module and the data storage module are connected by signals.

[0037] The model layer includes a large language model module and a model deployment module, and there is a signal connection between the large language model module and the model deployment module.

[0038] The control layer includes an input processing module and an output processing module, and the input processing module and the output processing module are signal connected.

[0039] The application layer includes an identity security management module and a functional module, and there is a signal connection between the identity security management module and the functional module.

[0040] The interaction layer includes hardware interaction module and software interaction module.

[0041] Furthermore, the data collection module includes building fire monitoring data collection, unit fire safety management data collection, fire protection facility and equipment status data collection, law enforcement history data collection, and third-party maintenance data, detection data, and evaluation data collection.

[0042] Building fire protection monitoring data collection uses smoke detectors and temperature sensors to collect building fire protection data for real-time monitoring of fire hazards in buildings.

[0043] Unit safety management data collection includes the unit's fire protection facilities, fire personnel configuration and training records, and fire emergency plans, which are used to evaluate the unit's own fire safety management level and its response capabilities when a fire occurs.

[0044] The status data collection of fire-fighting facilities and equipment includes water pressure monitoring of fire hydrants, pressure and validity period of fire extinguishers, and operating status data of automatic sprinkler fire-fighting systems, which are used to ensure the normal operation of fire-fighting facilities.

[0045] The collection of historical law enforcement data includes data recorded by the fire law enforcement department during the fire inspection and law enforcement process of the unit, including inspection results, problems found, rectification requirements and penalties, which are used to reflect the fire safety compliance of the unit.

[0046] The collection of historical law enforcement data and third-party maintenance data, inspection data, and evaluation data includes the maintenance, inspection, and evaluation of fire protection facilities by third-party professional organizations, which are used to understand the actual status of fire protection facilities and the fire safety status of the unit.

[0047] Furthermore, the large language model module uses the ChatGLM2 large language model and the LORA model. The ChatGLM2 large language model is used to process the building fire monitoring data and unit fire safety management data from the data layer, and the LORA model is used to fine-tune the ChatGLM2 large language model.

[0048] Furthermore, the LORA model is represented by a feedforward neural network using the original pre-trained model, and its formula is:

[0049] Among them, x is the input vector, W is the weight matrix, b is the bias vector, and h is the output vector. The modification of LORA fine-tuning is performed by adding two low-rank decomposition matrices. The modified calculation becomes:

[0050]

[0051] Among them, B and A are two low-rank matrices, usually the dimension of B is r×k, the dimension of A is k×n, assuming that the dimension of the original weight matrix W is r×n, k≪r and k≪n.

[0052] Furthermore, the input processing module includes receiving input data and preprocessing the data format. The input data is used to receive input from the application layer, including Prompt words, History records and Argument model parameters. Prompt words are used to guide the large language model to analyze and answer in the direction of specific security assessment. History records can provide past relevant situations as a reference for the current assessment. Argument model parameters involve the configuration parameters required for the specific operation of the model, which unify the encoding format of the received input data and adjust the text length and structure.

[0053] Furthermore, the output processing module includes result post-processing and feedback to the application layer. The result post-processing is used to generate corresponding evaluation results according to the input data, and the feedback to the application layer is used to accurately feed back the processed valid results to the application layer.

[0054] Furthermore, the identity security management module includes identity authentication, permission control and operation auditing. Identity authentication is used to authenticate the user who logs into the system through biometric technology. Permission control is used to assign corresponding system operation permissions to different users based on their roles and responsibilities. Operation auditing is used to record every operation of the user in the system in detail, including the time, content and data involved in the operation.

[0055] Furthermore, the functional modules include intelligent risk assessment, daily self-inspection assessment and safety assessment report generation. Intelligent risk assessment uses the processed data obtained from the control layer, combined with preset risk assessment rules and algorithms, to conduct intelligent analysis of fire safety related situations and check for various potential safety risks. Daily self-inspection assessment is used to provide corresponding self-inspection process guidance and data entry interface. Safety assessment report generation is based on the comprehensive safety assessment data collected by the system, including various monitoring data, inspection data and model analysis results, and automatically generates a safety assessment report in accordance with standardized report templates and format requirements.

[0056] Furthermore, the hardware interaction module includes tablet interaction and AR glasses interaction, wherein the tablet interaction is used for system support to connect with a dedicated tablet, by developing adaptive drivers and application programming interfaces, and the AR glasses interaction is used for inspectors to view the actual scene while superimposing relevant safety assessment information in the field of view through the system.

[0057] Furthermore, the software interaction module includes mobile application interaction and PC application interaction. The mobile application interaction is used to interact with the application layer of the system through the network communication protocol for data exchange, and the PC application interaction is used to meet the needs of large-scale data import and export, generate detailed security assessment reports, and perform in-depth configuration of the system.

[0058] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

Claims

1. A natural language conversational risk management system based on artificial intelligence, comprising a data layer, a model layer, a control layer, an application layer and an interaction layer, characterized in that: The data layer is connected to the model layer by signal, the model layer is connected to the control layer by signal, the control layer is connected to the application layer by signal, and the application layer is connected to the interaction layer by signal; The data layer includes a data collection module and a data storage module, and the data collection module and the data storage module are connected by signals; The model layer includes a large language model module and a model deployment module, and the large language model module and the model deployment module are connected by signals; The control layer includes an input processing module and an output processing module, and the input processing module and the output processing module are connected by signals; The application layer includes an identity security management module and a functional module, and the identity security management module and the functional module are connected by signals; The interaction layer includes a hardware interaction module and a software interaction module.

2. The natural language conversational risk management system based on artificial intelligence according to claim 1, characterized in that: The data collection module includes building fire monitoring data collection, unit fire safety management data collection, fire protection facilities and equipment status data collection, law enforcement history data collection, and third-party maintenance data, detection data, and evaluation data collection; The building fire monitoring data collection uses smoke detectors and temperature sensors to collect building fire data for real-time monitoring of fire hazards in buildings; The unit safety management data collection includes the unit's fire protection facilities, fire personnel configuration and training records, and fire emergency plans, which are used to evaluate the unit's own fire safety management level and its response capabilities when a fire occurs; The fire-fighting facilities and equipment status data collection includes water pressure monitoring of fire hydrants, pressure and validity period of fire extinguishers, and operation status data of automatic sprinkler fire extinguishing systems, which are used to ensure the normal operation of fire-fighting facilities; The enforcement history data collection includes data recorded by the fire enforcement department during the fire inspection and enforcement process of the unit, including inspection results, problems found, rectification requirements and penalties, which are used to reflect the fire safety compliance of the unit; The collection of law enforcement historical data and third-party maintenance data, inspection data, and evaluation data includes the maintenance, inspection, and evaluation of fire protection facilities by third-party professional organizations, which are used to understand the actual status of fire protection facilities and the fire safety status of the unit.

3. The natural language conversational risk management system based on artificial intelligence according to claim 1, characterized in that: The large language model module uses the ChatGLM2 large language model and the LORA model. The ChatGLM2 large language model is used to process the building fire monitoring data and unit fire safety management data from the data layer, and the LORA model is used to fine-tune the ChatGLM2 large language model.

4. The natural language conversational risk management system based on artificial intelligence according to claim 3 is characterized by: The LORA model is represented by a feedforward neural network using the original pre-trained model, and its formula is: ； Among them, x is the input vector, W is the weight matrix, b is the bias vector, and h is the output vector. The modification of LORA fine-tuning is performed by adding two low-rank decomposition matrices. The modified calculation becomes: ； ； Among them, B and A are two low-rank matrices, usually the dimension of B is r×k, the dimension of A is k×n, assuming that the dimension of the original weight matrix W is r×n, k≪r and k≪n.

5. The natural language conversational risk management system based on artificial intelligence according to claim 1, characterized in that: The input processing module includes receiving input data and preprocessing the data format. The received input data is responsible for receiving input from the application layer, including Prompt prompts, History historical records and Argument model parameters. The Prompt prompts are used to guide the large language model to analyze and answer in the direction of a specific security assessment. The History historical records can provide past related situations as a reference for the current assessment. The Argument model parameters involve the configuration parameters required for the specific operation of the model, unify the encoding format of the received input data, and adjust the text length and structure.

6. The natural language conversational risk management system based on artificial intelligence according to claim 1, characterized in that: The output processing module includes result post-processing and feedback to the application layer. The result post-processing is used to generate corresponding evaluation results according to input data, and the feedback to the application layer is used to accurately feed back the processed valid results to the application layer.

7. The natural language conversational risk management system based on artificial intelligence according to claim 1, characterized in that: The identity security management module includes identity authentication, authority control and operation auditing. The identity authentication is used to authenticate the identity of users logging into the system through biometric technology. The authority control is used to assign corresponding system operation permissions to different users based on their roles and responsibilities. The operation auditing is used to record in detail every operation of the user in the system, including the time, content and data involved in the operation.

8. The natural language conversational risk management system based on artificial intelligence according to claim 1, characterized in that: The functional modules include intelligent risk assessment, daily self-inspection assessment and safety assessment report generation. The intelligent risk assessment uses the processed data obtained from the control layer, combined with preset risk assessment rules and algorithms, to conduct intelligent analysis of fire safety related situations and check various potential safety risks. The daily self-inspection assessment is used to provide corresponding self-inspection process guidance and data entry interface. The safety assessment report generation is based on the comprehensive safety assessment data collected by the system, including various monitoring data, inspection data and model analysis results, and automatically generates a safety assessment report in accordance with standardized report templates and format requirements.

9. The natural language conversational risk management system based on artificial intelligence according to claim 1, characterized in that: The hardware interaction module includes tablet interaction and AR glasses interaction, wherein the tablet interaction is used for system support to connect with a dedicated tablet, by developing an adaptive driver and application program interface, and the AR glasses interaction is used for inspectors to view the actual scene while superimposing relevant safety assessment information in the field of view through the system.

10. The natural language conversational risk management system based on artificial intelligence according to claim 1, characterized in that: The software interaction module includes mobile application interaction and PC application interaction. The mobile application interaction is used to interact with the application layer of the system through a network communication protocol for data exchange, and the PC application interaction is used to meet the needs of large-scale data import and export, generate detailed security assessment reports, and perform in-depth configuration functions of the system.

Images