Business Management System

Abstract

To provide a business management system that enables even non-expert operators to obtain answers according to the situation and the questioner's intention, and improves visibility through organic linkage with a dashboard screen. [Solution] The equipment management system 100, which is a business management system, is composed of a vibration analysis and diagnosis system 1 for rotating machinery equipment supported by plain bearings or rolling bearings, a smart dashboard 2 that also displays a chatbot screen that combines a large-scale language model and an information search system, a platform server 3 that handles various types of data, various vibration sensors 4 that are arranged on each piece of equipment to monitor the status of the equipment in factories and plants, and a static data storage unit 5 that also stores user registration data registered by users themselves, such as process data, procedure manuals, and emergency response manuals, and the smart dashboard displays a question and answer screen that displays questions and answers and a dashboard screen side by side.

Description

[Technical Field]

[0001] The present invention relates to a business management system. [Background technology]

[0002] In the management of various types of information such as equipment operation information and manufacturing execution information in manufacturing operations, it is necessary to take appropriate measures according to the situation encountered. For example, in factories or plants, in order to prevent equipment downtime and maintain performance, sensor output data from measuring instruments installed on equipment is collected and analyzed. For example, machine vibration data is analyzed to detect equipment abnormalities.

[0003] The large amount of digital data generated as a result of collecting and analyzing sensor output data is visualized and displayed in easy-to-understand formats such as aggregated values, tables, and graphs on a dashboard screen, which is one of the BI (Business Intelligence) tools. For example, if the target is vibration data, trend graphs that can be used to determine whether the vibration is in a steady state or is deviating from the steady state, as well as spectrum graphs based on frequency analysis, are displayed.

[0004] The dashboard screen displays analysis results at a glance, reducing the need for even non-experts to refer to each piece of data individually. However, this only applies when the equipment is operating normally and safely. When equipment malfunctions, experienced operators review the individual data and determine the necessary measures based on their experience. Even experienced operators must search through large volumes of paper manuals to find the necessary information and take appropriate action when encountering a situation they have never experienced before. This type of work is not limited to preventing equipment downtime. In other words, to ensure efficient productivity, operators must make daily decisions based on large volumes of paper manuals, such as how to operate machines most efficiently based on their status, condition, and production volume, and how to maximize production volume with minimal energy consumption. Summary of the Invention [Problem to be solved by the invention]

[0005] Of course, attempts have been made to turn large volumes of paper manuals into a database and make it possible to access necessary information through a search. However, this ultimately relies on full-text searches using free words, and unless the operator is an experienced user, they may not know what clues to use and may not be able to set appropriate free words for a specific search, resulting in ineffective results. It is also difficult to search image data. Furthermore, the search tool is separated from the information on the dashboard screen, completely separating the work of identifying the problem from the work of searching for or considering a solution, making it an ineffective system.

[0006] The present invention has been made in view of the above, and aims to provide a business management system that enables even non-experienced operators to obtain answers that correspond to the questioner's intentions depending on the situation they encounter, and that can be organically linked with a dashboard screen to further improve the visibility of the screen display. [Means for solving the problem]

[0007] In order to solve such problems, the present invention provides a business management system that outputs information for management based on multiple data parameters to be managed, and includes: a dynamic data analysis means that collects and analyzes the data parameters and outputs visualized information that includes at least one of aggregate values, tables, and graphs; a display means that includes within its display area a dashboard screen that displays a list of the information output by the dynamic data analysis means; a static data storage unit that accumulates user-registered data consisting only of documents, images, table data, and graph data registered by the user; and an answer generation means that outputs an answer sentence generated in response to a question entered by a user using search expansion generation, which is a large-scale language model using the user-registered data.The display means displays a question and answer screen that displays the question sentences received and the answer sentences to be output by the answer generation means, and the dashboard screen side by side, and the search expansion generation is configured to suppress the occurrence of hallucination. [Effects of the Invention]

[0008] According to the present invention, it is possible to provide a business management system that enables even non-experienced operators to obtain answers that correspond to the questioner's intentions depending on the situation encountered, that can improve the visibility of the screen display by organically linking with the dashboard screen, and that can prevent non-experts with excellent fact-checking abilities from making incorrect responses due to incorrect answers. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a system configuration and functional block diagram of a business management system according to a first embodiment of the present invention; [Figure 2] FIG. 2 is a functional block diagram of a static data storage unit in the business management system according to the first embodiment of the present invention. [Figure 3] FIG. 2 is an explanatory diagram illustrating a smart dashboard function in the business management system according to the first embodiment of the present invention. [Figure 4]FIG. 2 is a diagram showing an example of the layout of a smart dashboard screen in the business management system according to the first embodiment of the present invention. [Figure 5] FIG. 2 is a diagram showing an example of the layout of a smart dashboard screen in the business management system according to the first embodiment of the present invention. [Figure 6] FIG. 2 is a diagram showing an example of the layout of a smart dashboard screen in the business management system according to the first embodiment of the present invention. [Figure 7] FIG. 10 is a functional block diagram of a static data storage unit in the business management system according to the second embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. The drawings are created for the purpose of explanation, and for the sake of clarity, some components not necessary for the explanation may not be shown.

[0011] First Embodiment The first embodiment relates to an equipment management system that is an operation management system that outputs information for equipment maintenance based on sensor output data from a plurality of measuring instruments installed in a plurality of equipment devices operating in a factory or plant. The following will explain the system with reference to Figures 1 to 6.

[0012] (Overall configuration of facility management system) FIG. 1 is a system configuration and functional block diagram of an equipment management system, which is an operation management system according to a first embodiment of the present invention. The equipment management system comprises an equipment status analysis and diagnosis system (dynamic analysis means), a smart dashboard, a platform server, multiple measuring instruments, and a static data storage unit. Here, as an example, a system that performs analysis of each vibration type based on the output of multiple vibration sensors installed in multiple pieces of equipment operating in a factory or plant will be described. That is, equipment management system 100, which is an operation management system according to the first embodiment of the present invention, comprises a vibration analysis and diagnosis system 1 (hereinafter simply referred to as "vibration analysis and diagnosis system 1") for rotating machinery equipment supported by plain bearings, rolling bearings, or the like, a smart dashboard 2, a platform server 3, various vibration sensors 4, and a static data storage unit 5.

[0013] Factories and plants are equipped with dispersed equipment such as pumps, fans, drive motors, actuators, and valves. Various vibration sensors 4 are installed on each piece of equipment to monitor its status. For example, an acceleration sensor is attached to the bearing of a drive motor to capture vibration data. The sensor outputs from these various vibration sensors 4 are collected as vibration IFs (various vibration information) by the platform server 3 (described below). Specifically, vibration waveform data measured by wireless or wired sensors attached to the equipment, along with vibration waveform data from the internal circuits of the wireless or wired sensors, are collected by a data collection device. The various vibration data are then subjected to analytical and calculation processes such as Fourier transform and vibration amplitude transformation, and transmitted to a higher-level controller via a communication bus that transmits digital signals. In the case of wireless sensors, all or part of the analytical and calculation processes may be performed by the internal circuits of the wireless sensors. Furthermore, various vibration data from some vibration sensors can also be collected via a cloud-based condition monitoring system.

[0014] Platform Server 3 handles EAI tools provided as middleware that connects different systems. While it is common to have systems that handle sensor output data from vibration sensors to monitor abnormalities in the manufacturing system, and systems that accumulate manufacturing process data and create production plans for production management, data integration between these systems is sometimes insufficient. Platform Server 3 handles vibration data as manufacturing EAI based on vibration IF (various vibration information), process data as manufacturing EAI accumulated in the static data storage unit 5, and vibration data as information EAI collected by a cloud-based condition monitoring system. In addition to process data, Platform Server 3 also handles documents registered by users themselves.

[0015] Because the platform server 3 uses an EAI tool, the business management system according to the first embodiment of the present invention realizes an integrated system that can be used in common by various departments. For example, the facility management system 100 can be effectively used by both the operation department and the maintenance department. This will be described in more detail later in the section on examples of using the smart dashboard.

[0016] In addition to process data, the static data storage unit 5 also stores user registration data registered by users themselves, such as procedure manuals and emergency response manuals. The user registration data is contextualized data that has been created by removing information that is unnecessary for generating answers to user questions. However, the static data storage unit 5 also stores the original data from which unnecessary information has not been removed.

[0017] In addition to the conventional dashboard screen that processes and displays dynamic data obtained from the system, Smart Dashboard 2 also centrally manages static data such as documents, procedure manuals, and other materials, obtains necessary information from various sources, and creates answers to questions from users using a large-scale language model to provide information to users.

[0018] The vibration analysis and diagnosis system 1 has various functions necessary for monitoring the status of facility equipment (detecting abnormalities) and for performing vibration analysis and diagnosis after abnormalities are detected. For example, when monitoring the status of facility equipment, it is possible to detect abnormalities in facility equipment early based on various vibration data obtained from vibration sensors attached to the drive motor, using the overall (OA) absolute value alarm, OA relative value alarm, spectrum alarm, and vector alarm functions in accordance with ISO evaluation standards, etc. Alarm generation information can be output in CSV format. In addition, for equipment where an abnormality has been detected, vibration diagnosis engineers can use 16 types of vibration data analysis functions, including trend analysis, spectrum analysis, orbit & waveform, and polar analysis, to estimate the cause of the abnormality and evaluate its condition (evaluate the extent of damage). For example, when evaluating the extent of damage to rolling bearings, the bearing analysis function can be used to compare and detect the characteristic vibration frequencies that occur when the rolling bearing is damaged with the spectrum data obtained by performing a Fourier transform on the measured vibration waveform data, and the magnitude of the vibration values, as well as use parameters such as the crest factor to determine the cause of the abnormality and evaluate its condition (evaluate the extent of damage). These analysis results can be output in CSV format. Furthermore, the vibration analysis and diagnosis system 1 has a precision diagnosis function that uses analysis data such as rotation speed, vibration waveform data, and spectrum data, as well as facility equipment information, to automatically estimate and display 17 types of abnormal phenomena in facility equipment supported by sliding bearings, and 16 types of abnormal phenomena in rotating machinery equipment supported by rolling bearings, etc. This diagnosis function is scheduled to perform diagnosis periodically or when an abnormality occurs, and outputs the results as numerical data (for example, unbalance = "1"). These analysis or diagnosis results are displayed on the dashboard screen in various formats, such as trend graphs and spectrum graphs. Because the content of these is no different from that of conventional asset management systems, further detailed explanation will be omitted. However, as explained by listing 17 or 16 types of abnormal phenomena, it is easy to understand that a paper manual detailing how to deal with abnormalities would be enormously voluminous. On the other hand, as will be described later, the business management system according to the first embodiment of the present invention does not learn information other than that obtained directly or indirectly from the vibration analysis and diagnosis system 1, such as web information obtained by connecting to the Internet. This is because such information has a greater negative impact on asset management due to its noise nature.

[0019] The distinctive feature of the present invention is the handling of documents, images, and table data registered by users themselves, which is the role of the static data storage unit. In other words, the features of this data include the processing of data before registration, the usage of registered data, and the confirmation of primary data. These points are explained below.

[0020] (Static data storage configuration) 2 is a functional block diagram of the static data storage unit 5 in the business management system according to the first embodiment of the present invention. The directions indicated by the dashed arrows represent the order of execution processes or the flow of data (information) in each functional block.

[0021] The static data storage unit 5 is composed of a registration processing unit 51 and a registration database 52. That is, the static data storage unit 5 is composed of a control means for data registration processing and a storage means for accumulating data. Specifically, the registration processing unit 51 is a means for performing the necessary processing to make the sentences, images, table data, etc. that the user wishes to register suitable for generating answer sentences using a large-scale language model. The registration database 52 is a means for accumulating and storing the processed user registration data and other data.

[0022] The registration processing unit 51 is composed of an input means 511, a contextualization means 512, and a vector data conversion means 513. As shown by the dashed arrows in Figure 2, data input by the input means 511 is sent to the contextualization means 512 and then to the vector data conversion means 513.

[0023] The input means 511 is a means for the user to select and input the document, image, or table data they wish to register. The operator can select a file name on the PC screen to determine the target and register it, or register it using drag-and-drop processing. A wide variety of documents can be registered, including plant system diagrams, operation manuals, emergency operating procedures, and diagnostic reports. Documents are not limited to text and PDF formats; they can also include word processing documents, spreadsheets, PowerPoint documents, and other documents that contain information for display on the screen, such as character size, font type, and display position. In addition to documents, images and table data can also be registered.

[0024] The contextualization means 512 is a means for extracting contextualization information from unstructured data such as text, images, and table data that the user intends to register, and assigning meaning to it. In other words, it is a control processing means for understanding the meaning of the data, assigning that meaning, and constructing relationships between data.

[0025] The vector data conversion means 513 generates vector data by quantifying the meanings assigned by the contextualization means 512 and the relationships between data constructed by the contextualization means 512. The generated vector data is added to the unstructured data and registered as user registration data in the registration database 52.

[0026] The registration database 52 accumulates and stores primary data 522, which is the original data that the user has designated as the registration target, in addition to the user registration data 521 generated by the registration processing unit 51. It also accumulates and stores sensor output history data 523, which is data that has been previously collected and analyzed by the vibration analysis and diagnosis system 1. The sensor output history data 523 includes alarm history, vibration data, process data, etc. Although not shown in the figure, the sensor output history data is also subjected to contextualization processing before being accumulated and stored.

[0027] Large-scale language models are natural language processing models constructed with a huge amount of calculations, data, and parameters, and, as is well known, use information available on the Internet as data for machine learning. Although answers to questions generated by large-scale language models have a certain degree of accuracy, it is also well known that incorrect answers may be generated depending on the Internet information used.

[0028] In this regard, the facility management system 100, which is a business management system according to the first embodiment of the present invention, does not obtain or store information from outside the system in the registration database 52, whether it is user registration data registered by the user himself or data previously collected and analyzed by the vibration analysis and diagnosis system 1. Therefore, even though a large-scale language model is used to generate and output answer statements, the occurrence of hallucination is suppressed. In a facility management system, the most important thing to avoid from a safety perspective is for managers or workers to perform incorrect operations based on erroneous information provided by the search expansion generation system, and thus suppressing the occurrence of hallucination is a major benefit.

[0029] Furthermore, although not shown in the drawings, in the equipment management system 100, which is a business management system according to the first embodiment, the static data storage unit is configured to accept registrations only from limited users. Specifically, two types of users are assumed as system users: non-limited users who can use only questions and answers through primary authentication, and limited users who can register user registration data in addition to questions and answers through secondary authentication, and the system is configured so that only limited users can register data. As a method of authentication, various methods such as MFA authentication (authentic, SMS, email) can be appropriately adopted for pre-registered users. Furthermore, other security enhancements, such as IP restrictions, may be used in combination with remote operations by limited users.

[0030] (Smart Dashboard Functionality Configuration) The Smart Dashboard display screen not only displays the conventional dashboard screen, which allows users to check equipment anomalies on-screen based on data from the vibration analysis and diagnosis system1, but also displays a chatbot screen that utilizes so-called RAG (Retrieval-Augmented Generation), which combines a large-scale language model with an information search system. By loading internal documents, users can obtain answers to questions about equipment anomalies. The size of each screen can also be freely set. This section explains the internal functional configuration that realizes the screen display on this Smart Dashboard.

[0031] FIG. 3 is an explanatory diagram illustrating the smart dashboard function in the business management system according to the first embodiment of the present invention. The large-scale language model generates answers to questions using the plant system diagram, operation manual, emergency operation procedure, and diagnostic report in the user registration data 521, as well as the alarm history, vibration data, and process data in the sensor output history data 523. As described above, the user registration data 521 and the sensor output history data 523 are contextualized and assigned vector data in advance, making it possible to utilize a vector search mechanism. Specifically, the question sentence is also contextualized, assigned vector data, and similar data is searched for by measuring the vector distance. The answer sentence is generated by the large-scale language model based on the similar data thus searched and extracted.

[0032] Searches using conventional databases and search tools involve full-text searching to find parts of all documents in the database that contain the search keyword, and there are two typical search methods: a Grep-type method that searches all documents each time a search is performed, and an index-type method that creates an index from documents and then searches them. The former has the disadvantage of becoming slower as the amount of data increases, while the latter is fast but has the disadvantage of requiring the creation of an index, which must be rebuilt when the materials are updated.

[0033] In contrast, the smart dashboard function in the business management system according to the first embodiment of the present invention employs a vector search mechanism, enabling faster and more relevant results to be returned compared to conventional keyword searches. Furthermore, a large-scale language model is used to create answers tailored to each user's query, based on the returned results, as needed. This makes it possible to provide answers that better reflect the questioner's intent, unlike full-text searches. Furthermore, by defining the search scope as needed, searches can be further accelerated.

[0034] (Example of using the Smart Dashboard) The smart dashboard screen in the business management system according to the first embodiment of the present invention displays a question and answer screen that displays questions received and answers output by the answer generation means, and a dashboard screen that allows users to check equipment anomalies on the screen. The question and answer screen also displays a link button for the user-registered data used to generate the answer. When this link button is operated, the user-registered data used is displayed on the same screen, allowing users to simultaneously check the answer and the materials registered by the user. This screen configuration makes it possible to appropriately address problems encountered in a variety of situations in a variety of departments. An example of a display screen is described below.

[0035] 4 and 5 are diagrams showing an example of the layout of a smart dashboard screen in the business management system according to the first embodiment of the present invention, and are intended to explain that the screens change in the order of Fig. 4 to Fig. 5. Here, an example will be described in which an alarm is generated in a pump, and the operation department, having confirmed abnormal vibrations on the dashboard screen, seeks out an appropriate response.

[0036] The upper left of the dashboard screen shown in Figure 4 displays an illustration of a pump and the fact that an "OA danger alarm" has been issued for the equipment name "C-axis refrigeration P." Additionally, the lower left of the screen displays a trend graph with the tab set to "Trend Graph," which shows that midway through the latter half of the graph's time axis, the velocity vibration suddenly increased from 3.3 mm / sRMS to 8.5 mm / sRMS, exceeding the OA danger alarm threshold of 7.0 mm / sRMS, that the flow rate dropped significantly from 18.3 m3 / h to 14.9 m3 / h at the same time as the velocity vibration increased, and that the pressure increased from 1.2 MPa to 3.9 MPa at the same time as the velocity vibration increased.

[0037] On the right side of the dashboard screen, a chatbot screen is displayed as a question and answer screen. When an operator in the operations department inputs the question, "Please tell me how to switch the C-axis chilled water pump," the chatbot responds, "Information on how to deal with the C-axis chilled water pump is included in the search results, so I will answer based on that. If the flow rate of the C-axis chilled water pump suddenly decreases and the speed vibration exceeds the danger alarm threshold, the recommended way to deal with it is to stop the C-axis chilled water pump and switch to the spare unit." In addition, a hyperlink is displayed, reading "Quote: How to deal with the C-axis chilled water pump.pdf." Clicking this hyperlink changes the display in the lower left corner of the dashboard screen to the screen shown in Figure 5.

[0038] In the lower left corner of the dashboard screen shown in Figure 5, the tab has been switched from "Trend Graph" to "Reference Materials," and the screen showing the original data (primary data) used for the responses is displayed. The original data also allows you to check the pump system diagram.

[0039] After the operation department has responded, the operator of the operation department typically uses a chat system (not shown) included in the facility management system 100 to report to the maintenance department staff that operation has been switched to the backup machine and to request an investigation into the cause of the abnormality. Here, we will continue with the explanation using an example in which the maintenance department, having received the report and request for investigation, conducts the investigation. Note that the screen of the chat system (not shown) has the same display as that of a chatbot, but it may be displayed in full screen to make the dialogue easier.

[0040] FIG. 6 is a diagram showing an example of the layout of a smart dashboard screen in the business management system according to the first embodiment of the present invention, and the situation set here is assumed to be one in which the system is operated by staff from the maintenance department.

[0041] The right side of the dashboard screen displays a chatbot screen for questions and answers. When a maintenance department staff member inputs the question, "Can you provide me with similar analysis reports from the past?", the chatbot responds, "The search results include past analysis reports for the C-axis chilled water pump. Specifically, you can view pump diagnostic results reports, trend graphs, spectrum graphs, and pump maintenance documents. By viewing these documents comprehensively, you can understand the condition of the C-axis chilled water pump and past troubleshooting methods." Multiple hyperlinks to reference documents are also displayed. This example shows an example in which an annotated image of a similar trend graph from the past is displayed, assuming that the trend graph hyperlink has already been clicked. Subsequent screen transitions are omitted, but the maintenance department staff further referenced the multiple documents linked to the hyperlinks and discovered a situation that matched the previous trouble analysis report. They determined that the cause of the problem was excessive impeller passing vibration due to a reduced flow rate caused by deterioration and damage to the pump discharge valve, which changed the static load and bearing load acting on the impeller.

[0042] From the above explanation, it can be seen that different tasks performed by multiple departments were smoothly carried out using a single integrated system. That is, when a pump alarm occurred, an operator in the operations department confirmed the vibration abnormality on the dashboard screen. After checking the document "C-Axis Chilled Water Pump Troubleshooting.pdf" on the chatbot screen, he switched to the backup pump in accordance with the chatbot's response. The operations department operator then used the system's chat function to request an actual inspection of the pump for which the alarm occurred. In response, the maintenance department used the chatbot screen to conduct an actual inspection, identify the cause of the abnormality, perform repairs, and verify its integrity. This is the flow of operations. The maintenance department then reported the cause and countermeasures to the operations department. The created report was added to the user registration data 521 by registering it using the drag-and-drop process described above.

[0043] In the first embodiment, we have described a business management system that outputs information for equipment maintenance based on sensor output data from multiple measuring devices installed on multiple pieces of equipment operating in a factory or plant. However, it should be readily understood that the problem described in the Background Art, in which even non-expert operators must respond appropriately to encountered situations, also applies to tasks other than maintenance. Specifically, unless the following can be done effectively, it is difficult to find an appropriate solution: handling production plan information, production KPI information, and budget-actual management information in production planning; handling equipment operation information, production execution information, and manufacturing KPI information in manufacturing; handling quality inspection information, sampling inspection information, and quality KPI information in quality control; handling inventory information, purchasing information, order information, and purchasing / procurement KPI information in purchasing and procurement; handling back office and various KPI information in general affairs and accounting; and handling quota information, order performance information, and sales KPI information in sales. The technical concept of the present invention can be similarly effectively applied to management or planning systems in these business fields. It is common for multiple systems to be connected using EAI, but the business management system of the present invention, in particular the smart dashboard function, makes it possible to further promote business collaboration among multiple departments in a single, more integrated system. The second embodiment will be briefly described below.

[0044] Second Embodiment The second embodiment relates to a business management system that manages and plans quota information in sales operations. The business management system according to the second embodiment manages so-called quota information (for example, targets or standards that a sales team or individual should achieve within a certain period of time, such as sales, number of appointments, number of leads, etc.) as various indicators for effectively promoting sales activities.

[0045] 7 is a functional block diagram of a static data storage unit 5A in a business management system according to a second embodiment of the present invention. The directions indicated by dashed arrows represent the order of execution processes or the flow of data (information) in each functional block.

[0046] The static data storage unit 5A is composed of a registration processing unit 51A and a registration database 52A. That is, the static data storage unit 5A is composed of a control means for data registration processing and a storage means for accumulating data. Specifically, the registration processing unit 51A is a means for performing the necessary processing to make the sentences, images, table data, etc. that the user wishes to register suitable for generating answer sentences using a large-scale language model. The registration database 52A is a means for accumulating and storing the processed user registration data and other data.

[0047] The registration processing unit 51A is composed of an input means 511A, a contextualization means 512A, and a vector data conversion means 513A. As indicated by the dashed arrows in Fig. 2, data input by the input means 511A is sent to the contextualization means 512A and then to the vector data conversion means 513A. The registration processing unit 51A also has a data cleaning means 514A that is not provided in the business management system of the first embodiment.

[0048] The input means 511A is a means for the user to select and input the documents, images, and table data they wish to register. The operator can select the file name on the PC screen to determine the target and register it, or register it using drag-and-drop processing. A wide variety of documents can be registered, including plant system diagrams, operation manuals, emergency operating procedures, and diagnostic reports. Documents are not limited to text and PDF formats, but also include word processing documents, spreadsheets, PowerPoint documents, and other documents that contain information for display on the screen, such as character size, font type, and display position. In addition to documents, images and table data can also be registered.

[0049] The contextualization means 512A is a means for extracting contextualization information from unstructured data such as text, images, and table data that the user intends to register, and assigning meaning to it. In other words, it is a control processing means for understanding the meaning of data, assigning that meaning, and constructing relationships between data.

[0050] The vector data conversion means 513A generates vector data by quantifying the meanings assigned by the contextualization means 512 and the relationships between data constructed by the contextualization means 512. The generated vector data is added to the unstructured data and registered as user registration data in the registration database 52A.

[0051] The data cleaning means 514A is a means for selecting reliable sources of data collected from outside the system and collecting unbiased data. Specifically, it compares the data with existing knowledge bases and fact-checking databases to check for inconsistencies and selects the data. Data reconciliation and summarization to improve search accuracy are also examples of data cleaning means. From the viewpoint of safety, the facility management system, which is the business management system of the first embodiment, is configured not to obtain information from outside the system and not to store and store it, in order to minimize the occurrence of hallucination. On the other hand, when formulating various indicators to effectively advance sales activities, it is not appropriate to ignore the active use of external knowledge. While a certain degree of hallucination is tolerated, the key point of the second embodiment is to perform sufficient data cleaning while utilizing external data, with the aim of reducing it as much as possible.

[0052] The registration database 52A accumulates and stores the user registration data 521A generated by the registration processing unit 51A, as well as primary data 522A, which is the original data that the user has selected to register. It also accumulates and stores previously collected history data 523A. Although not shown in the figure, the history data is also subjected to contextualization processing before being accumulated and stored.

[0053] To prevent the occurrence of hallucination, not only are measures taken in the static data storage unit but also in search expansion generation, which will now be described. Although not shown in the figures, the large-scale language model in the second embodiment has multiple models prepared, and the search expansion generation is configured to select an appropriate large-scale language model depending on the query.

[0054] Specifically, when a question is written in Japanese and relates to mathematics, a Japanese language model suitable for processing numerical data is selected. While the performance of Japanese language models has improved in recent years, e.g., fewer incorrect answers are obtained when prompts are entered in Japanese rather than English, it has been reported that Japanese language models are not good at answering mathematical questions. For example, it has been reported that setting the prompt to "Which is larger, 9.11 or 9.9?" yields a correct answer, whereas setting the prompt to "Which is larger, 9.11 or 9.9?" results in an incorrect answer. This is because Western European language models can accommodate commas every three digits, and operators commonly use commas, whereas previous Japanese language models do not accommodate the four-digit numeration system unique to Japanese numerical representation. In the second embodiment, a Japanese language model suitable for processing numerical data is selected for mathematical (numerical) questions, e.g., by accommodating the four-digit numeration system.

[0055] Numerical information is essential for the goals and indicators of sales activities, and the search expansion generation in the second embodiment offers significant benefits. In addition, when selecting an appropriate large-scale language model according to a question, it is preferable to prepare models that are suited to answers that are good at handling natural language, images, etc., in addition to those that focus on numerical data, and to select such a model.

[0056] The business management systems according to each embodiment of the present invention have been described above in detail with reference to the drawings, giving specific examples of the vibration analysis and diagnosis system 1 and the business management system for quota information in sales operations. However, the specific configurations are not limited to these examples, and the present invention also includes design changes that do not deviate from the gist of the present invention. For example, in a configuration in which information is not obtained from outside the system and stored, multiple models may be prepared as large-scale language models, and a configuration may be adopted in which the appropriate large-scale language model is selected for search expansion generation depending on the question. Furthermore, depending on the equipment installed in a factory or plant, the measuring instruments installed in the equipment are not limited to vibration sensors and accelerometers, but may include all measuring instruments used to understand and manage the status of equipment, such as flow meters, temperature sensors, and pressure sensors. Furthermore, in an equipment management system that outputs information for equipment maintenance, equipment maintenance should not be understood as being limited to preventing system downtime. As mentioned in the "Background Art" section at the beginning, since systems are also required to maintain performance, it should be understood in a broader sense as maintaining or managing the production volume and quality produced by the equipment. In that sense, the business management system of the present invention encompasses the concept of a production management system and its development into predictive detection and factor analysis using AI such as machine learning. It should be fully understood that the significance of this invention lies in the fact that it enables business collaboration across multiple departments in a single, more integrated system using the smart dashboard function, and that the great significance of the multifaceted measures taken to prevent the occurrence of hallucination is great. [Explanation of symbols]

[0057] 1. Vibration analysis and diagnostic system (dynamic data analysis means) 2. Smart Dashboard (display method) 3 Platform Server 4. Vibration sensors (measuring instruments) 5. Static Data Storage 51 Registration processing section 511 Input Method 512 Contextualization Measures 513 Vector Data Conversion Method 52 Registration Database 521 User Registration Data 522 Primary Data 523 Sensor output history data 5A Static data storage section 51A Registration Processing Unit 511A Input means 512A Contextualization Means 513A Vector data conversion means 514A Data Cleaning Methods 52A Registration Database 521A User Registration Data 522A Primary Data 523A Historical Data

Claims

1. A business management system that outputs information for management based on a plurality of data specifications to be managed, a dynamic data analysis means for collecting and analyzing the data elements and outputting the collected data as visualized information including at least one of a summary value, a table, and a graph; a display means including a dashboard screen in a display area for displaying a list of information output by the dynamic data analysis means; a static data storage unit for storing user registration data consisting only of documents, images, table data, and graph data registered by users themselves; and an answer generation means for generating an answer to a question input by a user using a search expansion generation, which is a large-scale language model using the user registration data, and outputting the generated answer, the display means displays a question and answer screen, which displays a question received by the answer generation means and an answer to be output, and the dashboard screen side by side; The search expansion generation is configured to suppress the occurrence of hallucination. A business management system characterized by:

2. The search expansion generation is configured to prevent the occurrence of hallucination by preventing the static data storage unit from acquiring and storing information from outside the system.

2. The business management system according to claim 1.

3. the static data store has a contextualization means; The contextualization means extracts contextualization information from unstructured data of text, images, table data, and graph data that the user wishes to register, converts the extracted information into vector data, and registers the vector data in the user registration data.

3. The business management system according to claim 2.

4. The static data storage unit also stores, as historical data, information that the dynamic data analysis means has previously collected and analyzed the data specifications and visualized to include at least one of an aggregated value, a table, and a graph; The answer generating means treats the history data as a category of the user registration data and outputs an answer sentence by utilizing the history data.

4. The business management system according to claim 3.

5. a plurality of large-scale language models are prepared, The search expansion generation is configured to select an appropriate model depending on the query, thereby suppressing the occurrence of hallucination.

2. The business management system according to claim 1.

6. If the question is in Japanese and is related to mathematics, a Japanese language model suitable for processing numerical data is set.

6. The business management system according to claim 5.

7. 7. The business management system according to claim 6, wherein the Japanese language model suitable for processing the numerical data corresponds to a four-digit numbering system.

8. The search expansion generation is configured so that the static data storage unit only accepts registrations from authorized users, thereby suppressing the occurrence of hallucination.

2. The business management system according to claim 1.

9. System users are classified into two types: non-restricted users who can only use questions and answers after the first authentication, and restricted users who can register data in addition to questions and answers after the second authentication.

9. The business management system according to claim 8.

10. The data specifications are sensor output data from a plurality of measuring devices installed in a plurality of equipment devices operating in a factory or plant, and the management is management of these equipment devices.

10. The business management system according to claim 1, 2, or 5 to 9.

11. The data specification is quota information in sales operations, and the management is management or planning in sales operations.

10. The business management system according to claim 1, 3 or 9.

Images