Intelligent government self-service robot based on data analysis

By introducing data analysis and self-intelligent training technologies into government self-service robots, the problems that existing robots are difficult to answer accurately and monitor in real time are solved, and efficient question-and-answer service and operation status management are realized, ensuring the stability and timely elimination of the robot.

CN117332058BActive Publication Date: 2024-11-08JIANGSU WISOFT INFORMATION TECH LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202311259106.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-09-27
Publication Date
2024-11-08
Estimated Expiration
2043-09-27

AI Technical Summary

Technical Problem

It is difficult for existing government self-service robots to provide accurate answers and cannot detect and evaluate their own operating conditions in real time, which makes it difficult for managers to take timely improvement measures and cannot accurately determine the life status of the robot, affecting its stable and efficient operation and timely elimination.

Method used

An intelligent government self-service robot based on data analysis was designed, including a Q&A service robot unit, a government service knowledge base unit, an LLM service unit, a government service Q&A recovery unit and a government service Q&A training unit, to improve the accuracy of Q&A service through self-intelligent training. and monitor and evaluate the robot's operating status in real time through the robot monitoring and diagnosis unit, the operation comprehensive evaluation unit and the intelligent early warning unit.

Benefits of technology

It realizes robot self-intelligence training, improves the accuracy and user experience of Q&A services, promptly monitors and evaluates the operating status, ensures the stable operation of the robot, and accurately determines the life status when the evaluation fails, and supports timely elimination.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117332058B_ABST
    Figure CN117332058B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of service robots, and specifically is an intelligent government affairs self-service robot based on data analysis, including a question-and-answer service robot unit, a government affairs service knowledge base unit, an LLM service unit, a government affairs service question-and-answer recovery unit, and a government affairs service question-and-answer training unit; after receiving question-and-answer information raised by a user, the present invention first retrieves an answer from the government affairs service knowledge base unit, and if no hit answer is found in the government affairs service knowledge base unit, the answer is directly retrieved from the LLM service unit, and the government affairs service question-and-answer recovery unit screens the generated log audit records, and the government affairs service question-and-answer training unit trains the received corresponding question-and-answer information to realize self-intelligent training of the intelligent government affairs self-service robot, and is able to perform real-time detection and analysis of its own operating status and conduct periodic comprehensive operating evaluation, and accurately determine its life status when the comprehensive evaluation is unqualified, with a high degree of intelligence.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of service robots, and in particular to an intelligent government self-service robot based on data analysis. Background Art

[0002] At present, the application of service robots is becoming more and more common in various fields. Government self-service robots are a type of service robots. However, most of the existing government self-service robots do not have the ability to learn and optimize independently. They can only answer according to pre-set rules and templates and cannot understand the user's intention well. This makes the robots often unable to give accurate answers to various questions raised by users, and even deviate from the original intention of the questions.

[0003] In addition, existing government self-service robots are unable to effectively monitor their own operating conditions and conduct real-time detection and analysis, as well as conduct periodic comprehensive operational evaluations, which is not conducive to managers making timely and targeted improvement measures to ensure the stable and efficient operation of government self-service robots. When the comprehensive evaluation fails, its life status cannot be accurately determined, which is not conducive to its timely elimination. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent government self-service robot based on data analysis, which solves the problems that existing government self-service robots are often difficult to give accurate answers, and are unable to perform real-time detection and analysis of their own operating conditions and conduct periodic comprehensive operation evaluations, which is not conducive to managers to make corresponding improvement measures in a timely and targeted manner to ensure the stable and efficient operation of the government self-service robot, and the problem that their life status cannot be accurately determined when the comprehensive evaluation fails.

[0005] The technical solution adopted by the present invention to solve the above technical problems is:

[0006] Intelligent government self-service robots based on data analysis, including question-answering service robot units, government service knowledge base units, LLM service units, government service question-answering recycling units, and government service question-answering training units;

[0007] The question-and-answer service robot unit is used to receive question-and-answer information raised by users, so as to directly interact with front-end users and answer government affairs questions raised by front-end users, wherein the question-and-answer information includes business questions and non-business questions of government affairs; the government affairs service knowledge base unit stores government affairs knowledge with clear answers, and is used to provide government affairs business standard knowledge; the LLM service unit stores a large amount of general knowledge, and performs free question-and-answer based on the LLM large language model, and is used to provide knowledge other than government affairs standard knowledge;

[0008] The government service question and answer recovery unit is used to record question and answer information for which there is no precise answer in the government service knowledge base unit and the LLM service unit provides free answers. The corresponding question and answer information is recorded in a log and reviewed and screened. The government service question and answer training unit is used to optimize the question and answer information that has been reviewed and screened in the government service question and answer recovery unit, so as to realize the self-intelligent training of the government service robot, reduce the pressure on the server when users ask questions concurrently, improve the accuracy of the question and answer service, and enhance the user's question and answer service experience.

[0009] Furthermore, the working method of the intelligent government self-service robot includes the following steps:

[0010] Step 1: The question-answering service robot unit receives question-answering information raised by the user;

[0011] Step 2: For government affairs business questions, the question-answering service robot unit first retrieves answers from the government affairs service knowledge base unit; if a matching answer is found in the government affairs service knowledge base unit, it will return an accurate answer to the user and provide the most standard answer; if a matching answer is not found in the government affairs service knowledge base unit, the question-answering service robot unit will directly retrieve the answer in the LLM service unit to provide a corresponding answer;

[0012] Step 3: The answers returned by the government service knowledge base unit are first polished by the LLM service unit to ensure that the language expression of the answers is more accurate and natural, and then the optimized answers are returned to the user through the question-answering service robot unit;

[0013] Step 4: The question-answering service robot unit returns the answer to the user, completing the answering service;

[0014] Step 5: The question-answering service robot unit pushes the questions whose answers are generated by the LLM service unit to the government service question-answering recycling unit, and the government service question-answering recycling unit receives the questions and answers and forms a log;

[0015] Step 6: The government service question and answer recovery unit screens the generated log audit records, selects useful question and answer logs and pushes them to the government service question and answer training unit for subsequent training of the government service question and answer training unit;

[0016] Step 7. The government service question and answer training unit will train the corresponding question and answer information received, and divide the useful question and answer log into two parts of data. One part of the data will be stored in the government service knowledge base service unit to become a new government service precise question and answer, providing users with more accurate answers; the other part of the data will be stored in the LLM service unit to improve the general question and answer capabilities, so that the system can better understand the user's questions and provide corresponding answers.

[0017] Furthermore, it also includes a robot monitoring and diagnostic unit, a robot operation comprehensive evaluation unit and an intelligent early warning unit, and the intelligent early warning unit is communicatively connected to the robot monitoring and diagnostic unit and the robot operation comprehensive evaluation unit; the robot monitoring and diagnostic unit monitors the operation status of the intelligent government affairs self-service robot, and diagnoses and analyzes its abnormal operation status, thereby generating a diagnosis qualification signal or a diagnosis failure signal, and sending the diagnosis failure signal to the intelligent early warning unit; the robot operation comprehensive evaluation unit conducts a comprehensive evaluation of the operation of the intelligent government affairs self-service robot, thereby generating a comprehensive evaluation qualification signal or a comprehensive evaluation failure signal, and sending the comprehensive evaluation failure signal to the intelligent early warning unit.

[0018] Furthermore, the specific operation process of the robot monitoring and diagnosis unit includes:

[0019] The number of network interruptions and network delay data within a unit time are collected, and the number of network interruptions and network delay data are numerically compared with a preset network interruption number threshold and a preset network delay data threshold, respectively. If the number of network interruptions or the network delay data exceeds the corresponding preset threshold, it is determined that the network state is abnormal and a diagnosis failure signal is generated;

[0020] If the number of network interruption times and network delay data do not exceed the corresponding preset thresholds, the average upload speed value, average download speed value, upload speed fluctuation amplitude value and download speed fluctuation amplitude value of the robot in unit time are collected, and the average upload speed value, average download speed value, upload speed fluctuation amplitude value and download speed fluctuation amplitude value are numerically calculated to obtain a network comprehensive analysis value; the network comprehensive analysis value is numerically compared with the preset network comprehensive analysis threshold value, and if the network comprehensive analysis value does not exceed the preset network comprehensive analysis threshold value, the network state is judged to be abnormal and a diagnosis failure signal is generated;

[0021] If the network comprehensive analysis value exceeds the preset network comprehensive analysis threshold, the network upload speed curve and the network download speed curve within a unit time are obtained, and a number of detection points are randomly arranged in the network upload speed curve. The network upload speed value of the detection point is compared with the preset network upload speed range. If the network upload speed value is not within the preset network upload speed range, the corresponding detection point is marked as a fluctuation point, and the number of fluctuation points in the network upload speed curve is marked as the network upload fluctuation frequency; the network download fluctuation frequency is obtained in the same way;

[0022] The network upload fluctuation frequency and the network download fluctuation frequency are weighted and calculated to obtain the network wave frequency analysis value, and the network wave frequency analysis value is numerically compared with the preset network wave frequency analysis threshold. If the network wave frequency analysis value exceeds the preset network wave frequency analysis threshold, the network status is judged to be abnormal and a diagnosis failure signal is generated.

[0023] Furthermore, if the network wave frequency analysis value does not exceed the preset network wave frequency analysis threshold, the time when the question-answering service robot unit receives the question-answering information raised by the user is collected and marked as the request time, and the time when the answer is fed back to the user is collected and marked as the answer time, and the time difference between the corresponding answer time and the request time is calculated to obtain the single question-answering efficiency value;

[0024] The single question and answer efficiency value is numerically compared with the preset single question and answer efficiency threshold. If the single question and answer efficiency value exceeds the preset single question and answer efficiency threshold, the corresponding service process is judged to be slow and marked as a slow response service; the number of slow response services per unit time is collected and marked as the slow response frequency, and the ratio of the slow response frequency to the total number of services per unit time is calculated to obtain the slow response ratio; the slow response frequency and the slow response ratio are numerically calculated to obtain the slow response coefficient, and the slow response coefficient is numerically compared with the preset slow response coefficient threshold; if the slow response coefficient exceeds the preset slow response coefficient threshold, a diagnosis failure signal is generated; if the slow response coefficient does not exceed the preset slow response coefficient threshold, a diagnosis qualification signal is generated.

[0025] Furthermore, the specific operation process of the robot operation comprehensive evaluation unit includes:

[0026] A comprehensive evaluation period is set, and the number of times a qualified diagnosis signal and the number of times a unqualified diagnosis signal are generated within the comprehensive evaluation period are collected, and the ratio of the number of times a unqualified diagnosis signal is generated to the number of times a qualified diagnosis signal is generated is calculated to obtain a diagnosis coefficient; and the comprehensive evaluation period is divided into a number of detection time periods, and each group of detection time periods has the same duration, and the number of times a unqualified diagnosis signal is generated in the corresponding detection time period is marked as a time period diagnosis value, and the time period diagnosis value is numerically compared with a preset time period diagnosis threshold value. If the time period diagnosis value exceeds the preset time period diagnosis threshold value, the corresponding detection time period is marked as a diagnosis period, and the number of diagnosis periods is ratio-calculated to the number of detection periods to obtain a diagnosis coefficient;

[0027] By analyzing and obtaining the hardware abnormality data and software abnormality data of the intelligent government self-service robot during the comprehensive evaluation period, the diagnosis coefficient, diagnosis coefficient, hardware abnormality data and software abnormality data are numerically calculated to obtain the operation comprehensive evaluation coefficient, and the operation comprehensive evaluation coefficient is numerically compared with the preset operation comprehensive evaluation coefficient threshold. If the operation comprehensive evaluation coefficient exceeds the preset operation comprehensive evaluation coefficient threshold, a comprehensive evaluation failure signal is generated; if the operation comprehensive evaluation coefficient does not exceed the preset operation comprehensive evaluation coefficient threshold, a comprehensive evaluation qualified signal is generated.

[0028] Furthermore, the analysis and acquisition method of hardware abnormality data and software abnormality data is as follows:

[0029] The hardware anomaly frequency and the corresponding hardware anomaly repair time of the intelligent government self-service robot during the comprehensive evaluation period are collected, and the number of hardware anomalies exceeding the preset repair time threshold is marked as unconventional hard anomaly frequency, and all repair times are summed up to obtain the total hardware repair time; similarly, the software anomaly frequency, unconventional software anomaly frequency and total software repair time are obtained; the hardware anomaly frequency, unconventional hard anomaly frequency and total hardware repair time are numerically calculated to obtain the hardware anomaly data, and the software anomaly frequency, unconventional software anomaly frequency and total software repair time are numerically calculated to obtain the software anomaly data.

[0030] Further, the robot operation comprehensive evaluation unit sends a comprehensive evaluation unqualified signal to the life state detection and analysis unit. After receiving the comprehensive evaluation unqualified signal, the life state detection and analysis unit performs life state detection and analysis on the intelligent government self-service robot, thereby generating a life state unqualified signal or a life state qualified signal, and sending the life state unqualified signal to the intelligent early warning unit. The specific analysis process of the life state detection and analysis is as follows:

[0031] The production interval time and the total use time of the intelligent government self-service robot are collected, and the production interval time and the total use time are numerically compared with the preset production interval time threshold and the total use time threshold, respectively. If the production interval time or the total use time exceeds the corresponding preset threshold, a life status unqualified signal is generated;

[0032] If both the production interval duration and the total operation duration do not exceed the corresponding preset thresholds, the number of times the equipment is impacted during its historical use is collected and marked as the impact damage frequency, and the operation environment detection coefficient is obtained through analysis; the operation environment detection coefficient, the impact damage frequency, the production interval duration and the total operation duration are numerically calculated to obtain the life status detection value, and the life status detection value is numerically compared with the preset life status detection threshold; if the life status detection value exceeds the preset life status detection threshold, a life status unqualified signal is generated; if the life status detection value does not exceed the preset life status detection threshold, a life status qualified signal is generated.

[0033] Furthermore, the specific analysis and acquisition method of the operation cycle detection coefficient is as follows:

[0034] The operating environment information of the intelligent government self-service robot at the corresponding moment is collected, and the operating environment information includes ambient temperature deviation data, ambient humidity deviation data, dust concentration data, light intensity data and electromagnetic interference data. The ambient temperature deviation data, ambient humidity deviation data, dust concentration data, light intensity data and electromagnetic interference data are normalized and calculated to obtain a real-time operating environment value; the real-time operating environment value is numerically compared with a preset real-time operating environment threshold value. If the real-time operating environment value exceeds the preset real-time operating environment threshold value, it is determined that the intelligent government self-service robot is in an abnormal impact state at the corresponding moment;

[0035] The total duration of the intelligent government self-service robot in the abnormal impact state during its historical operation is collected and marked as the total abnormal impact time value, and the duration of each abnormal impact state is marked as the single abnormal impact duration, and the number of single abnormal impact durations that exceed the preset single abnormal impact duration threshold is marked as the super impact frequency; the super impact frequency and the total abnormal impact time value are numerically calculated to obtain the operation cycle detection coefficient.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. In the present invention, the question and answer information raised by the user is received through the question and answer service robot unit, and the answer is first retrieved from the government service knowledge base unit. If a matching answer is found in the government service knowledge base unit, an accurate answer will be returned to the user. If a matching answer is not found in the government service knowledge base unit, the answer will be directly retrieved from the LLM service unit; the question and answer service robot unit pushes the question whose answer is generated by the LLM service unit to the government service question and answer recovery unit, and the government service question and answer recovery unit screens the generated log audit records. The government service question and answer training unit trains the corresponding question and answer information received to realize the self-intelligent training of the intelligent government self-service robot, which is helpful to improve the accuracy of the question and answer service, as well as improve the user's question and answer service experience and user satisfaction;

[0038] 2. In the present invention, the operating status of the intelligent government self-service robot is monitored, and its abnormal operating status is diagnosed and analyzed, thereby generating a qualified diagnosis signal or a failed diagnosis signal, so that the management personnel can grasp its abnormal operating status in real time and take corresponding improvement measures in time to ensure its normal operation; and the operation of the intelligent government self-service robot is comprehensively evaluated, thereby generating a qualified comprehensive evaluation signal or a failed comprehensive evaluation signal, so that the management personnel can promptly investigate and trace the cause, and take targeted improvement measures to ensure service quality; and when a failed comprehensive evaluation signal is generated, the life status of the intelligent government self-service robot is detected and analyzed, so that the management personnel can grasp its life status in detail and eliminate it in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a system block diagram of Embodiment 1 of the present invention;

[0040] Figure 2 This is a system block diagram of Embodiment 2, Embodiment 3 and Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0041] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0042] Embodiment 1: Figure 1 As shown, the intelligent government self-service robot based on data analysis proposed by the present invention has the advantages of strong adaptability, easy expansion, easy maintenance, etc. It can be customized and developed according to actual needs to meet the needs of different users. At the same time, it can also help users quickly find the required information, and it can not only improve the intelligent service level of the system, but also optimize the performance and efficiency of the system. Through continuous training and optimization, the system can better meet the needs of users and improve user satisfaction.

[0043] The intelligent government self-service robot includes a question-and-answer service robot unit, a government service knowledge base unit, an LLM service unit, a government service question-and-answer recovery unit, and a government service question-and-answer training unit; the question-and-answer service robot unit is used to receive question-and-answer information raised by users, so as to directly interact with front-end users and answer government affairs questions raised by front-end users, wherein the question-and-answer information includes business questions and non-business questions of government affairs; the government service knowledge base unit stores government affairs knowledge with clear answers, is used to provide government affairs business standard knowledge, and has high-performance knowledge retrieval, can quickly and accurately answer questions, can help users manage and share knowledge, and improve service efficiency and quality;

[0044] The LLM service unit stores a large amount of general knowledge and conducts free question-answering based on the LLM large language model, which is used to provide knowledge beyond government standard knowledge. By storing a large amount of general knowledge, the unit can quickly obtain the information required for user questions and answers, and freely answer the questions raised by users. It should be noted that the large language model (LLM) refers to a deep learning model trained with a large amount of text data, which can generate natural language text or understand the meaning of language text. LLM is usually based on deep learning architectures such as transformers, which helps them achieve impressive performance on various NLP tasks. LLM is characterized by its large scale and billions of parameters, which help them learn complex patterns in language data. This large language model can handle a variety of natural language tasks, such as text classification, question-answering, dialogue, etc., and is an important path to artificial intelligence.

[0045] The government service question and answer recovery unit is used to record question and answer information for which accurate answers cannot be found in the government service knowledge base unit and free answers are provided by the LLM service unit. It records the corresponding question and answer information into logs and conducts review and screening. It is an important unit for recording and managing user questions and answers; the government service question and answer training unit is used to optimize and train the question and answer information that has been reviewed and screened in the government service question and answer recovery unit, so as to realize the self-intelligent training of the government service robot, reduce the pressure on the server when users ask questions concurrently, improve the accuracy of the question and answer service, and enhance the user's question and answer service experience.

[0046] Furthermore, the working method of the intelligent government self-service robot includes the following steps:

[0047] Step 1: The question-answering service robot unit receives question-answering information raised by the user;

[0048] Step 2: For government affairs business questions, the question-answering service robot unit first retrieves answers from the government affairs service knowledge base unit; if a matching answer is found in the government affairs service knowledge base unit, it will return an accurate answer to the user and provide the most standard answer; if a matching answer is not found in the government affairs service knowledge base unit, the question-answering service robot unit will directly retrieve the answer in the LLM service unit to provide a corresponding answer;

[0049] Step 3: The answers returned by the government service knowledge base unit are first polished by the LLM service unit to ensure that the language expression of the answers is more accurate and natural, and then the optimized answers are returned to the user through the question-answering service robot unit;

[0050] Step 4: The question-answering service robot unit returns the answer to the user, completing the answering service;

[0051] Step 5: The question-answering service robot unit pushes the questions whose answers are generated by the LLM service unit to the government service question-answering recycling unit, and the government service question-answering recycling unit receives the questions and answers and forms a log;

[0052] Step 6: The government service question and answer recovery unit screens the generated log audit records, selects useful question and answer logs and pushes them to the government service question and answer training unit for subsequent training of the government service question and answer training unit;

[0053] Step 7. The government service question and answer training unit will train the corresponding question and answer information received to realize the self-intelligent training of the intelligent government self-service robot, which will help improve the accuracy of the question and answer service and enhance the user's question and answer service experience; the useful question and answer log will be divided into two parts of data, one part of the data will be stored in the government service knowledge base service unit to become a new government service precise question and answer, providing users with more accurate answers; the other part of the data will be stored in the LLM service unit to improve the general question and answer capabilities, so that the system can better understand the user's questions and provide corresponding answers.

[0054] This intelligent government self-service robot has the ability to learn and optimize independently, and can understand the user's intention very well, so that the robot can give accurate answers when answering various questions raised by users, providing users with more accurate and appropriate answers, which helps to solve situations that deviate from the original intention of the question; and the language expression is also more natural and fluent, which can better interact with users and provide a better user experience.

[0055] Embodiment 2: Figure 2 As shown, the difference between this embodiment and embodiment 1 is that the intelligent government self-service robot also includes a robot monitoring and diagnosis unit and an intelligent early warning unit, and the intelligent early warning unit is communicatively connected with the robot monitoring and diagnosis unit; the robot monitoring and diagnosis unit monitors the operation status of the intelligent government self-service robot, and diagnoses and analyzes its abnormal operation status, thereby generating a qualified diagnosis signal or a unqualified diagnosis signal, and sending the unqualified diagnosis signal to the intelligent early warning unit, and the intelligent early warning unit issues a corresponding early warning, and generates a corresponding early warning text message and sends it to the intelligent terminal of the manager, so that the manager can grasp its abnormal operation status in real time and take corresponding improvement measures in time to ensure its normal operation; the specific operation process of the robot monitoring and diagnosis unit is as follows:

[0056] The number of network interruptions and network delay data of the intelligent government self-service robot during operation in unit time are collected. The network delay data is a data quantity value indicating the degree of network delay status. The larger the value of the network delay data, the worse the network status is. The number of network interruptions and network delay data are numerically compared with the preset network interruption number threshold and the preset network delay data threshold, respectively. If the number of network interruptions or the network delay data exceeds the corresponding preset threshold, the network status is judged to be abnormal and a diagnosis failure signal is generated.

[0057] If the number of network interruptions and network delay data do not exceed the corresponding preset thresholds, the average upload speed value, average download speed value, upload speed fluctuation amplitude value and download speed fluctuation amplitude value of the robot in unit time are collected, among which the upload speed fluctuation amplitude value and download speed fluctuation amplitude value are data volume values ​​representing the size of upload speed fluctuation and the size of download speed fluctuation. The larger the values ​​of the two are, the more unstable the running network is; through the formula The average upload speed value ST, the average download speed value XT, the upload speed fluctuation amplitude value SB and the download speed fluctuation amplitude value XB are numerically calculated to obtain the network comprehensive analysis value WZ;

[0058] Among them, eq1, eq2, eq3, eq4 are preset proportional coefficients, and the values ​​of eq1, eq2, eq3, eq4 are all greater than zero; and the larger the value of the network comprehensive analysis value WZ is, the better the current network condition is, and the more conducive to its normal operation; the network comprehensive analysis value WZ is numerically compared with the preset network comprehensive analysis threshold value, and if the network comprehensive analysis value WZ does not exceed the preset network comprehensive analysis threshold value, the network state is judged to be abnormal and a diagnosis failure signal is generated;

[0059] If the network comprehensive analysis value WZ exceeds the preset network comprehensive analysis threshold, the network upload speed curve and the network download speed curve within a unit time are obtained, and a number of detection points are randomly arranged in the network upload speed curve. The network upload speed value of the detection point is compared with the preset network upload speed range. If the network upload speed value is not within the preset network upload speed range, the corresponding detection point is marked as a fluctuation point, and the number of fluctuation points in the network upload speed curve is marked as the network upload fluctuation frequency; the network download fluctuation frequency is obtained in the same way;

[0060] The network upload fluctuation frequency QB and the network download fluctuation frequency WB are weighted and calculated by the formula BP=ef1*QB+ef2*WB to obtain the network wave frequency analysis value BP, wherein ef1 and ef2 are preset weight coefficients, and the values ​​of ef1 and ef2 are both greater than zero; and the larger the value of the network wave frequency analysis value BP is, the worse the network condition is; the network wave frequency analysis value BP is numerically compared with the preset network wave frequency analysis threshold value, and if the network wave frequency analysis value BP exceeds the preset network wave frequency analysis threshold value, the network condition is judged to be abnormal and a diagnosis failure signal is generated;

[0061] If the network wave frequency analysis value BP does not exceed the preset network wave frequency analysis threshold, the time when the question-answering service robot unit receives the question-answering information raised by the user is collected and marked as the request time, and the time when the answer is fed back to the user is collected and marked as the answer time, and the time difference between the corresponding answer time and the request time is calculated to obtain the single question-answering efficiency value; it should be noted that the larger the value of the single question-answering efficiency value, the lower the question-answering efficiency, and the less smooth the corresponding service process;

[0062] The single question-and-answer efficiency value is numerically compared with the preset single question-and-answer efficiency threshold. If the single question-and-answer efficiency value exceeds the preset single question-and-answer efficiency threshold, the corresponding service process is judged to be slow and marked as a slow response service; the number of slow response services per unit time is collected and marked as a slow response frequency, and the ratio of the slow response frequency to the total number of services per unit time is calculated to obtain a slow response ratio;

[0063] The slow reaction coefficient MF is obtained by numerically calculating the slow reaction frequency MP and the slow reaction ratio MB through the formula MF=ef3*MP+ef4*MB, wherein ef3 and ef4 are preset weight coefficients, and ef4>ef3>0; and, the larger the value of the slow reaction coefficient MF is, the lower the operating efficiency and the worse the operating condition of the intelligent government self-service robot are; the slow reaction coefficient MF is numerically compared with the preset slow reaction coefficient threshold; if the slow reaction coefficient exceeds the preset slow reaction coefficient threshold, a diagnosis failure signal is generated; if the slow reaction coefficient MF does not exceed the preset slow reaction coefficient threshold, a diagnosis qualification signal is generated.

[0064] Embodiment 3: Figure 2 As shown, the difference between this embodiment and Embodiment 1 and Embodiment 2 is that the intelligent government self-service robot further includes a robot operation comprehensive evaluation unit, and the intelligent early warning unit is communicatively connected with the robot operation comprehensive evaluation unit. The robot operation comprehensive evaluation unit performs a comprehensive evaluation on the operation of the intelligent government self-service robot, thereby generating a comprehensive evaluation qualified signal or a comprehensive evaluation unqualified signal, and sends the comprehensive evaluation unqualified signal to the intelligent early warning unit. The intelligent early warning unit issues a corresponding early warning, generates a corresponding early warning text message and sends it to the intelligent terminal of the manager, so that the manager can promptly investigate and trace the cause, and take targeted improvement measures to ensure the service quality. The specific operation process of the robot operation comprehensive evaluation unit is as follows:

[0065] A comprehensive evaluation period is set, preferably, the duration of the comprehensive evaluation period is 30 days; the number of times a qualified diagnosis signal is generated and the number of times an unqualified diagnosis signal is generated within the comprehensive evaluation period are collected, and the ratio of the number of times an unqualified diagnosis signal is generated to the number of times a qualified diagnosis signal is generated is calculated to obtain a diagnosis coefficient; and the comprehensive evaluation period is divided into several detection time periods, each group of detection time periods has the same duration, the number of times an unqualified diagnosis signal is generated in the corresponding detection time period is marked as a time period diagnosis abnormality value, the time period diagnosis abnormality value is numerically compared with a preset time period diagnosis abnormality threshold, if the time period diagnosis abnormality value exceeds the preset time period diagnosis abnormality threshold, indicating that the corresponding detection time period frequently has a diagnosis abnormality, the corresponding detection time period is marked as a diagnosis abnormality time period, and the number of diagnosis abnormality time periods is ratio-calculated to the number of detection time periods to obtain a diagnosis abnormality coefficient;

[0066] The hardware anomaly data and software anomaly data of the intelligent government self-service robot during the comprehensive evaluation period are obtained through analysis, specifically: the hardware anomaly frequency of the intelligent government self-service robot during the comprehensive evaluation period and the corresponding hardware anomaly repair time are collected, and the number of hardware anomalies exceeding the corresponding preset repair time threshold is marked as an unconventional hard anomaly frequency, and all repair times are summed up to calculate the total hardware repair time; similarly, the software anomaly frequency, unconventional soft anomaly frequency and total software repair time are obtained; the hardware anomaly frequency YC, unconventional hard anomaly frequency FY and total hardware repair time YF are numerically calculated through the formula YS=tg1*YC+tg2*FY+tg3*YF to obtain the hardware anomaly data YS;

[0067] Among them, tg1, tg2, and tg3 are preset weight coefficients, and the values ​​of tg1, tg2, and tg3 are all greater than zero, and tg2>tg1>tg3; similarly, the software abnormality frequency, the unconventional software abnormality frequency, and the total software repair time are numerically calculated using the above formula to obtain the software abnormality data RS; it should be noted that the larger the values ​​of the hardware abnormality data YS and the software abnormality data RS, the worse the operating condition of the intelligent government self-service robot during the comprehensive evaluation period;

[0068] By formula The diagnostic coefficient ZD, the differential diagnosis coefficient ZP, the hardware abnormality data YS and the software abnormality data RS are numerically calculated to obtain the operation comprehensive evaluation coefficient ZX, wherein up1, up2, up3, up4 are preset proportional coefficients, and the values ​​of up1, up2, up3, up4 are all greater than zero; and the larger the value of the operation comprehensive evaluation coefficient ZX, the worse the evaluation result of the intelligent government self-service robot in the comprehensive evaluation period as a whole; the operation comprehensive evaluation coefficient ZX is numerically compared with the preset operation comprehensive evaluation coefficient threshold; if the operation comprehensive evaluation coefficient ZX exceeds the preset operation comprehensive evaluation coefficient threshold, a comprehensive evaluation unqualified signal is generated; if the operation comprehensive evaluation coefficient ZX does not exceed the preset operation comprehensive evaluation coefficient threshold, a comprehensive evaluation qualified signal is generated.

[0069] Embodiment 4: Figure 2 As shown, the difference between this embodiment and Embodiment 1, Embodiment 2, and Embodiment 3 is that the robot operation comprehensive evaluation unit sends a comprehensive evaluation unqualified signal to the life state detection and analysis unit. After receiving the comprehensive evaluation unqualified signal, the life state detection and analysis unit performs life state detection and analysis on the intelligent government self-service robot, thereby generating an unqualified life state signal or a qualified life state signal, and sending the unqualified life state signal to the intelligent early warning unit. The intelligent early warning unit issues a corresponding early warning, generates a corresponding early warning text message and sends it to the intelligent terminal of the manager, so that the manager can grasp its life state in detail and eliminate it in time. The specific analysis process of the life state detection and analysis is as follows:

[0070] The production interval time and total use time of the intelligent government self-service robot are collected, wherein the production interval time is a data value indicating the interval time between the current date and its production date, and the total use time is a data value indicating the total use time of the robot; the production interval time and the total use time are numerically compared with the preset production interval time threshold and the total use time threshold, respectively; if the production interval time or the total use time exceeds the corresponding preset threshold, a life status unqualified signal is generated;

[0071] If the production interval and the total operation time do not exceed the corresponding preset thresholds, the number of times it has been impacted during its historical use is collected and marked as the impact damage frequency; and the operation environment detection coefficient is obtained through analysis, specifically: the operating environment information of the intelligent government self-service robot at the corresponding moment is collected, and the operating environment information includes ambient temperature deviation data, ambient humidity deviation data, dust concentration data, light intensity data and electromagnetic interference data, among which the ambient temperature deviation data is a data value indicating the degree of deviation of the ambient temperature from the preset suitable temperature, and the ambient humidity deviation data is similar; the electromagnetic interference data is a data value indicating the intensity of electromagnetic interference in the environment;

[0072] By formula The ambient temperature deviation data HW, ambient humidity deviation data HS, dust concentration data FS, light intensity data GQ and electromagnetic interference data DG are normalized and calculated to obtain the real-time environmental operation value SH; wherein, a1, a2, a3, a4, a5 are preset weight coefficients, and the values ​​of a1, a2, a3, a4, a5 are all greater than zero; and the larger the value of the real-time environmental operation value SH is, the greater the damage caused to the intelligent government self-service robot by the environmental conditions at the corresponding moment; the real-time environmental operation value SH is numerically compared with the preset real-time environmental operation threshold value, and if the real-time environmental operation value SH exceeds the preset real-time environmental operation threshold value, it is judged that the intelligent government self-service robot is in an abnormal impact state at the corresponding moment;

[0073] The total duration of the intelligent government self-service robot in the abnormal impact state during the historical operation process is collected and marked as the total abnormal impact time value, and the duration of each abnormal impact state is marked as the single abnormal impact duration, and the number of single abnormal impact durations exceeding the preset single abnormal impact duration threshold is marked as the super-impact frequency; the super-impact frequency CY and the total abnormal impact time value YP are numerically calculated by the formula YJ=b1*CY+b2*YP to obtain the operation cycle detection coefficient YJ; wherein b1 and b2 are preset weight coefficients, and the values ​​of b1 and b2 are both greater than zero;

[0074] The life state detection value ZY is obtained by numerically calculating the operation environment detection coefficient YJ, the impact damage frequency CP, the production interval time SJ and the total operation time ZS through the formula ZY=cw1*YJ+cw2*CP+(cw3*SJ+cw4*ZS) / 2, wherein cw1, cw2, cw3, cw4 are preset proportional coefficients, and the values ​​of cw1, cw2, cw3, cw4 are all greater than zero; and the larger the value of the life state detection value ZY is, the worse the life state of the intelligent government self-service robot is, and it can be eliminated in time as needed; the life state detection value ZY is numerically compared with the preset life state detection threshold; if the life state detection value ZY exceeds the preset life state detection threshold, a life state unqualified signal is generated; if the life state detection value ZY does not exceed the preset life state detection threshold, a life state qualified signal is generated.

[0075] The working principle of the present invention is as follows: when in use, the question and answer information raised by the user is received through the question and answer service robot unit, and the answer is first retrieved from the government service knowledge base unit. If a matching answer is found in the government service knowledge base unit, an accurate answer will be returned to the user to provide the most standard answer; if a matching answer is not found in the government service knowledge base unit, the question and answer service robot unit directly retrieves the answer in the LLM service unit to provide a corresponding answer; for the answer returned by the government service knowledge base unit, the LLM service unit first performs language polishing to ensure that the language expression of the answer is more accurate and natural, and then the optimized answer is returned to the user through the question and answer service robot unit; the question and answer service robot unit pushes the question whose answer is generated by the LLM service unit to the government service question and answer recovery unit, the government service question and answer recovery unit screens the generated log audit records, and the government service question and answer training unit trains the corresponding question and answer information received to realize the self-intelligent training of the intelligent government self-service robot, which helps to improve the accuracy of the question and answer service, as well as improve the user's question and answer service experience and user satisfaction.

[0076] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0077] Rather, it is to better explain the principles and practical applications of the present invention so that those skilled in the art can better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An intelligent government self-service robot based on data analysis, characterized by: It includes question-answering service robot unit, government service knowledge base unit, LLM service unit, government service question-answering recovery unit, government service question-answering training unit, robot monitoring and diagnosis unit, robot operation comprehensive evaluation unit and intelligent early warning unit; The question-and-answer service robot unit is used to receive question-and-answer information raised by users, so as to directly interact with front-end users and answer government affairs questions raised by front-end users, wherein the question-and-answer information includes business questions and non-business questions of government affairs; the government affairs service knowledge base unit stores government affairs knowledge with clear answers, and is used to provide government affairs business standard knowledge; the LLM service unit stores a large amount of general knowledge, and performs free question-and-answer based on the LLM large language model, and is used to provide knowledge other than government affairs standard knowledge; The government service question and answer recovery unit is used to record the question and answer information that cannot find accurate answers in the government service knowledge base unit, but the LLM service unit provides free answers, and records the corresponding question and answer information into logs for review and screening; the government service question and answer training unit is used to optimize the question and answer information that has been reviewed and screened in the government service question and answer recovery unit, so as to realize the self-intelligent training of the government service robot, reduce the pressure on the server when users ask questions concurrently, improve the accuracy of the question and answer service, and enhance the user's question and answer service experience; The intelligent early warning unit is communicatively connected with the robot monitoring and diagnosis unit and the robot operation comprehensive evaluation unit; the robot monitoring and diagnosis unit monitors the operation status of the intelligent government self-service robot, and diagnoses and analyzes its abnormal operation status, thereby generating a qualified diagnosis signal or a failed diagnosis signal, and sending the failed diagnosis signal to the intelligent early warning unit; the robot operation comprehensive evaluation unit performs a comprehensive evaluation on the operation of the intelligent government self-service robot, thereby generating a qualified comprehensive evaluation signal or a failed comprehensive evaluation signal, and sending the failed comprehensive evaluation signal to the intelligent early warning unit; The specific operation process of the robot monitoring and diagnosis unit includes: The number of network interruptions and network delay data within a unit time are collected. If the number of network interruptions or the network delay data exceed the corresponding preset thresholds, the network state is judged to be abnormal and a diagnosis failure signal is generated; if the number of network interruptions and the network delay data do not exceed the corresponding preset thresholds, the average upload speed value, the average download speed value, the upload speed fluctuation amplitude value and the download speed fluctuation amplitude value of the robot within a unit time are collected, and the average upload speed value, the average download speed value, the upload speed fluctuation amplitude value and the download speed fluctuation amplitude value are numerically calculated to obtain a network comprehensive analysis value; if the network comprehensive analysis value does not exceed the preset network comprehensive analysis threshold, the network state is judged to be abnormal and a diagnosis failure signal is generated; If the network comprehensive analysis value exceeds the preset network comprehensive analysis threshold, a network upload speed curve graph and a network download speed curve graph within a unit time are obtained, and a number of detection points are randomly arranged in the network upload speed curve graph. If the network upload speed value is not within the preset network upload speed range, the corresponding detection point is marked as a fluctuation point, and the number of fluctuation points in the network upload speed curve graph is marked as the network upload fluctuation frequency; the network download fluctuation frequency is obtained in the same way; the network upload fluctuation frequency and the network download fluctuation frequency are weighted and calculated to obtain a network wave frequency analysis value. If the network wave frequency analysis value exceeds the preset network wave frequency analysis threshold, the network state is judged to be abnormal and a diagnosis failure signal is generated; If the network wave frequency analysis value does not exceed the preset network wave frequency analysis threshold, the time when the question-answering service robot unit receives the question-answering information raised by the user is collected and marked as the request time, and the time when the answer is fed back to the user is collected and marked as the answer time, and the time difference between the corresponding answer time and the request time is calculated to obtain the single question-answering efficiency value; If the single question-and-answer efficiency value exceeds the preset single question-and-answer efficiency threshold, the corresponding service process is judged to be slow and marked as a slow response service; the number of slow response services per unit time is collected and marked as the slow response frequency, and the ratio of the slow response frequency to the total number of services per unit time is calculated to obtain the slow response ratio; the slow response frequency and the slow response ratio are numerically calculated to obtain the slow response coefficient, and if the slow response coefficient exceeds the preset slow response coefficient threshold, a diagnosis failure signal is generated; if the slow response coefficient does not exceed the preset slow response coefficient threshold, a diagnosis qualification signal is generated.

2. The intelligent government self-service robot based on data analysis according to claim 1 is characterized in that: The specific operation process of the robot operation comprehensive evaluation unit includes: A comprehensive evaluation period is set, and the number of times a qualified diagnosis signal and the number of times a unqualified diagnosis signal are generated within the comprehensive evaluation period are collected, and the ratio of the number of times a unqualified diagnosis signal is generated to the number of times a qualified diagnosis signal is generated is calculated to obtain a diagnosis coefficient; and the comprehensive evaluation period is divided into a number of detection time periods, and the duration of each group of detection time periods is the same, and the number of times a unqualified diagnosis signal is generated in the corresponding detection time period is marked as a time period diagnosis value. If the time period diagnosis value exceeds the preset time period diagnosis threshold, the corresponding detection time period is marked as a diagnosis error time period, and the number of diagnosis error time periods is calculated to obtain a diagnosis error coefficient by ratio of the number of detection time periods; By analyzing and obtaining the hardware anomaly data and software anomaly data of the intelligent government self-service robot during the comprehensive evaluation period, the diagnosis coefficient, diagnosis coefficient, hardware anomaly data and software anomaly data are numerically calculated to obtain the operation comprehensive evaluation coefficient; if the operation comprehensive evaluation coefficient exceeds the preset operation comprehensive evaluation coefficient threshold, a comprehensive evaluation failure signal is generated; if the operation comprehensive evaluation coefficient does not exceed the preset operation comprehensive evaluation coefficient threshold, a comprehensive evaluation qualified signal is generated.

3. The intelligent government self-service robot based on data analysis according to claim 2 is characterized in that: The analysis and acquisition methods of hardware abnormality data and software abnormality data are as follows: The hardware anomaly frequency and the corresponding hardware anomaly repair time of the intelligent government self-service robot during the comprehensive evaluation period are collected, and the number of hardware anomalies exceeding the preset repair time threshold is marked as unconventional hard anomaly frequency, and all repair times are summed up to obtain the total hardware repair time; similarly, the software anomaly frequency, unconventional software anomaly frequency and total software repair time are obtained; the hardware anomaly frequency, unconventional hard anomaly frequency and total hardware repair time are numerically calculated to obtain the hardware anomaly data, and the software anomaly frequency, unconventional software anomaly frequency and total software repair time are numerically calculated to obtain the software anomaly data.

4. The intelligent government self-service robot based on data analysis according to claim 2 is characterized in that: The robot operation comprehensive evaluation unit sends a comprehensive evaluation failure signal to the life state detection and analysis unit. After receiving the comprehensive evaluation failure signal, the life state detection and analysis unit performs life state detection and analysis on the intelligent government self-service robot, thereby generating a life state failure signal or a life state success signal, and sending the life state failure signal to the intelligent early warning unit. The specific analysis process of the life state detection and analysis is as follows: The production interval time and total use time of the intelligent government self-service robot are collected. If the production interval time or the total use time exceeds the corresponding preset threshold, an unqualified life status signal is generated; if both the production interval time and the total use time do not exceed the corresponding preset threshold, the number of times it is impacted during its historical use is collected and marked as the impact damage frequency, and the operation environment detection coefficient is obtained through analysis; the operation environment detection coefficient, the impact damage frequency, the production interval time and the total use time are numerically calculated to obtain the life status detection value; if the life status detection value exceeds the preset life status detection threshold, an unqualified life status signal is generated; if the life status detection value does not exceed the preset life status detection threshold, a qualified life status signal is generated.

5. The intelligent government self-service robot based on data analysis according to claim 4 is characterized in that: The specific analysis and acquisition method of the operation and ring detection coefficient is as follows: The operating environment information of the intelligent government self-service robot at the corresponding moment is collected. The operating environment information includes ambient temperature deviation data, ambient humidity deviation data, dust concentration data, light intensity data and electromagnetic interference data. The ambient temperature deviation data, ambient humidity deviation data, dust concentration data, light intensity data and electromagnetic interference data are normalized and calculated to obtain the real-time operating environment value; If the real-time operation ring value exceeds the preset real-time operation ring threshold, it is judged that the intelligent government self-service robot is in an abnormal impact state at the corresponding moment; The total duration of the intelligent government self-service robot in the abnormal impact state during its historical operation is collected and marked as the total abnormal impact time value, and the duration of each abnormal impact state is marked as the single abnormal impact duration, and the number of single abnormal impact durations that exceed the preset single abnormal impact duration threshold is marked as the super impact frequency; the super impact frequency and the total abnormal impact time value are numerically calculated to obtain the operation cycle detection coefficient.

Citation Information

Patent Citations

  • Real-time environmental quality monitoring and evaluation system based on mobile phone back clip, mobile phone and application software thereof

    CN106482777A

  • Machine learning based operation and maintenance robot network fault detection system

    CN108683517A

  • Platform and system for digital personalized medicine

    CN110622179A

  • Government affair question and answer intelligent reply method and device based on neural network

    CN114625858A

  • Online customer service intelligent answering method and system applied to financial insurance platform

    CN115599889A