Method and device for evaluating electromechanical equipment based on large model agent cooperation

Abstract

The application relates to the technical field of equipment detection and equipment evaluation, and discloses an electromechanical equipment evaluation method and device based on large model intelligent agent cooperation, which comprises the following steps: collecting target equipment data, transmitting the target equipment data to a data buffer zone to store the target equipment data; reading the target equipment data and calling a model calculation intelligent agent to perform a data analysis operation to obtain a data analysis result; generating an equipment early warning result based on the data analysis result; performing a data encryption operation on the data analysis result and the equipment early warning result to obtain a data encryption result, generating an equipment evaluation result based on the data encryption result and the target equipment data; extracting evaluation feature data of the equipment evaluation result and sending the evaluation feature data to a large model cloud platform to generate intelligent agent analysis data, and generating an equipment evaluation report of the target equipment according to the intelligent agent analysis data. It can be seen that the application can improve the accuracy and reliability of electromechanical equipment evaluation and improve the efficiency of safety evaluation of electromechanical equipment.

Description

Technical Field

[0001] This invention relates to the field of equipment testing and equipment evaluation technology, and in particular to a method and apparatus for evaluating electromechanical equipment based on large-scale intelligent agent collaboration. Background Technology

[0002] Electromechanical equipment (generator sets, large wind turbines, etc.) are core assets in industrial production and energy supply. Sudden equipment failures can directly lead to production line shutdowns, safety accidents, and huge economic losses. Therefore, conducting full-process status assessments, fault predictions, and risk warnings based on real-time operational data has become a rigid requirement for intelligent manufacturing and industrial safety. Currently, the mainstream industrial sites use general-purpose edge gateways, industrial RTUs, and general-purpose edge computing terminals as data acquisition and cloud interaction carriers, undertaking on-site data acquisition, network forwarding, and local simple threshold alarm functions. However, the existing general-purpose terminal hardware architecture is only compatible with traditional single-machine algorithms and single-model cloud access scenarios, which leads to technical problems such as low reliability and low assessment response efficiency in the assessment process.

[0003] Therefore, it is particularly important to improve the accuracy and reliability of safety assessments of electromechanical equipment, as well as to increase the efficiency of such assessments. Summary of the Invention

[0004] This invention provides a method and apparatus for evaluating electromechanical equipment based on large-scale intelligent agent collaboration, which can improve the accuracy and reliability of safety assessment of electromechanical equipment, as well as improve the efficiency of safety assessment of electromechanical equipment.

[0005] The first aspect of this invention discloses a method for evaluating electromechanical equipment based on large-scale intelligent agent collaboration, the method comprising: Collect target device data and transfer the target device data to a preset data buffer for storage; The target device data is read from the preset data buffer, and a pre-determined model computing agent is invoked to perform data analysis operations on the target device data to obtain data analysis results; Based on the data analysis results, generate the device early warning result for the target device corresponding to the target device data; The data analysis results and the device warning results are encrypted to obtain encrypted data results. Based on the encrypted data results and the target device data, the device evaluation results of the target device are generated. The evaluation feature data of the device evaluation results is extracted, the evaluation feature data is sent to the large model cloud platform, and intelligent agent analysis data is generated based on the large model cloud platform. Finally, the device evaluation report of the target device is generated based on the intelligent agent analysis data.

[0006] As an optional implementation, in a first aspect of the present invention, the method further includes: Based on the equipment evaluation report, determine the abnormal parameters of the target equipment, and generate equipment alarm information based on the abnormal parameters. Based on the device alarm information and the pre-determined alarm components, alarm control parameters corresponding to the alarm components are generated, and the alarm components are controlled to perform alarm operations that match the alarm control parameters to obtain alarm results. Based on the alarm result, real-time environmental information corresponding to the target device is obtained. Based on the real-time environmental information, risk warning information of the current environment where the target device is located is determined. Based on the risk warning information, an isolated passive switch interlock signal is generated. The isolated passive switch interlock signal is used to control the alarm component or the target device to perform a shutdown operation.

[0007] As an optional implementation, in a first aspect of the present invention, sending the evaluation feature data to a large model cloud platform and generating agent analysis data based on the large model cloud platform includes: Based on the evaluation feature data, an evaluation data threshold matching the evaluation feature data is determined in a pre-determined judgment rule database, and a data comparison operation is performed between the evaluation feature data and the evaluation data threshold to obtain the data comparison result; Based on the data comparison results, risk decision parameters corresponding to the data comparison results are determined, wherein the risk decision parameters include at least the risk level of the target equipment, risk recommendation measures, and the confidence score corresponding to the data comparison results; The risk decision parameters, the data comparison results, and the target device data are sent to the big model cloud platform to extract the full-link data corresponding to the target device through the big model cloud platform, and intelligent agent analysis data is generated based on the full-link data. The end-to-end data includes at least the original time-series data, evaluation feature data, risk decision conclusions, device identifier, and evaluation timestamp corresponding to the target device.

[0008] As an optional implementation, in a first aspect of the present invention, generating alarm control parameters corresponding to the alarm component based on the device alarm information and a pre-determined alarm component includes: Based on the device alarm information and the pre-determined alarm components, the structured instructions corresponding to the device alarm information are parsed into predefined hardware control commands through the application processor's fixed scheduler. The hardware control commands are used to determine the level state of the corresponding pins through the GPIO controller, and alarm control parameters corresponding to the alarm component are generated based on the level state.

[0009] As an optional implementation, in a first aspect of the present invention, the step of transmitting the target device data to a preset data buffer to store the target device data includes: The target device data is transmitted to a preset data buffer through a unique fixed transmission channel. The original data inside the data buffer is retrieved. Based on the original data inside the data buffer, hardware-level encryption and digital signature operations are performed on the target device data to obtain encrypted device data. Data processing operations are performed on the encrypted data of the device to obtain the device stored data, and the device stored data is stored in the data buffer; wherein, the data processing operations include data cleaning operations, data removal operations, and data feature extraction operations.

[0010] As an optional implementation, in a first aspect of the present invention, generating a device evaluation result for the target device based on the data encryption result and the target device data includes: Hardware decryption and signature validity verification are uniformly performed on the encrypted data to obtain the data verification result, and it is determined whether the data verification result meets the preset verification conditions. When it is determined that the data verification result meets the preset verification conditions, the plaintext feature data corresponding to the data encryption result and the target device data is determined, and the plaintext feature data is stored in an independent cache partition; Read all independent storage data stored in the independent cache partition, generate multi-dimensional evaluation results based on the independent storage data, and generate device evaluation results for the target device based on the multi-dimensional evaluation results; The multi-dimensional evaluation results include at least the target device's health score, fault type, fault confidence, remaining lifespan, and fault trend.

[0011] As an optional implementation, in a first aspect of the present invention, sending the risk decision parameters, the data comparison results, and the target device data to a large-scale model cloud platform to extract the end-to-end data corresponding to the target device through the large-scale model cloud platform, and generating agent analysis data based on the end-to-end data, includes: Based on the risk decision parameters, the data comparison results and the target device data, data to be analyzed is generated, the data to be analyzed is sent to the large model cloud platform, and the data message information corresponding to the data to be analyzed is generated through the large model cloud platform. Determine whether the data message information is used to indicate that the data to be analyzed meets preset data analysis conditions; When it is determined that the data message information indicates that the data to be analyzed meets the preset data analysis conditions, the full-link data corresponding to the target device is extracted through the large model cloud platform, and a data packaging operation is performed on the full-link data to obtain the link data packaging result. Based on the link data packaging result, intelligent agent analysis data is generated. When it is determined that the data packet information indicates that the data to be analyzed does not meet the preset data analysis conditions, a judgment log information corresponding to the data packet information is generated based on the data to be analyzed, the data packet information, and the preset data analysis conditions. An exception prompt information is generated based on the judgment log information, and an adjustment operation is performed on the data packet information according to the exception prompt information to obtain the adjusted data packet information. The operation of determining whether the data packet information is used to indicate that the data to be analyzed meets the preset data analysis conditions is then triggered again.

[0012] A second aspect of this invention discloses an electromechanical equipment evaluation device based on large-scale intelligent agent collaboration, the device comprising: The acquisition module is used to acquire data from the target device; The transmission module is used to transmit the target device data to a preset data buffer for storing the target device data; A reading module is used to read data from the target device in a preset data buffer; The analysis module is used to calculate the intelligent agent to perform data analysis operations on the target device data using a pre-determined model, and obtain data analysis results; The generation module is used to generate a device warning result for the target device corresponding to the target device data based on the data analysis results. The encryption module is used to perform data encryption operations on the data analysis results and the device warning results to obtain data encryption results; The generation module is also used to generate a device evaluation result for the target device based on the data encryption result and the target device data; An extraction module is used to extract the evaluation feature data of the equipment evaluation results; The generation module is also used to send the evaluation feature data to the large model cloud platform; generate agent analysis data based on the large model cloud platform; and generate an equipment evaluation report for the target device based on the agent analysis data.

[0013] As an optional implementation, in a second aspect of the invention, the apparatus further includes: The determination module is used to determine the abnormal parameters of the target device based on the device evaluation report. The generation module is further configured to generate device alarm information based on the device abnormal parameters; and generate alarm control parameters corresponding to the alarm components based on the device alarm information and the pre-determined alarm components. The control module is used to control the alarm component to perform an alarm operation that matches the alarm control parameters, and to obtain an alarm result; The acquisition module is also used to obtain real-time environmental information corresponding to the target device based on the alarm result; The determining module is further configured to determine risk warning information of the current environment in which the target device is located based on the real-time environmental information; The generation module is further configured to generate an isolated passive switch interlock signal based on the risk warning information, wherein the isolated passive switch interlock signal is used to control the alarm component or the target device to perform a shutdown operation.

[0014] As an optional implementation, in a second aspect of the invention, the specific method by which the generation module sends the evaluation feature data to a large model cloud platform and generates agent analysis data based on the large model cloud platform includes: Based on the evaluation feature data, an evaluation data threshold matching the evaluation feature data is determined in a pre-determined judgment rule database, and a data comparison operation is performed between the evaluation feature data and the evaluation data threshold to obtain the data comparison result; Based on the data comparison results, risk decision parameters corresponding to the data comparison results are determined, wherein the risk decision parameters include at least the risk level of the target equipment, risk recommendation measures, and the confidence score corresponding to the data comparison results; The risk decision parameters, the data comparison results, and the target device data are sent to the big model cloud platform to extract the full-link data corresponding to the target device through the big model cloud platform, and intelligent agent analysis data is generated based on the full-link data. The end-to-end data includes at least the original time-series data, evaluation feature data, risk decision conclusions, device identifier, and evaluation timestamp corresponding to the target device.

[0015] As an optional implementation, in a second aspect of the present invention, the specific method by which the generation module generates alarm control parameters corresponding to the alarm component based on the device alarm information and the pre-determined alarm component includes: Based on the device alarm information and the pre-determined alarm components, the structured instructions corresponding to the device alarm information are parsed into predefined hardware control commands through the application processor's fixed scheduler. The hardware control commands are used to determine the level state of the corresponding pins through the GPIO controller, and alarm control parameters corresponding to the alarm component are generated based on the level state.

[0016] As an optional implementation, in a second aspect of the present invention, the specific method by which the transmission module transmits the target device data to a preset data buffer for storing the target device data includes: The target device data is transmitted to a preset data buffer through a unique fixed transmission channel. The original data inside the data buffer is retrieved. Based on the original data inside the data buffer, hardware-level encryption and digital signature operations are performed on the target device data to obtain encrypted device data. Data processing operations are performed on the encrypted data of the device to obtain the device stored data, and the device stored data is stored in the data buffer; wherein, the data processing operations include data cleaning operations, data removal operations, and data feature extraction operations.

[0017] As an optional implementation, in a second aspect of the present invention, the specific method by which the generation module generates the device evaluation result of the target device based on the data encryption result and the target device data includes: Hardware decryption and signature validity verification are uniformly performed on the encrypted data to obtain the data verification result, and it is determined whether the data verification result meets the preset verification conditions. When it is determined that the data verification result meets the preset verification conditions, the plaintext feature data corresponding to the data encryption result and the target device data is determined, and the plaintext feature data is stored in an independent cache partition; Read all independent storage data stored in the independent cache partition, generate multi-dimensional evaluation results based on the independent storage data, and generate device evaluation results for the target device based on the multi-dimensional evaluation results; The multi-dimensional evaluation results include at least the target device's health score, fault type, fault confidence, remaining lifespan, and fault trend.

[0018] As an optional implementation, in a second aspect of the invention, the generation module sends the risk decision parameters, the data comparison results, and the target device data to a large model cloud platform to extract the end-to-end data corresponding to the target device through the large model cloud platform. The specific method for generating agent analysis data based on the end-to-end data includes: Based on the risk decision parameters, the data comparison results and the target device data, data to be analyzed is generated, the data to be analyzed is sent to the large model cloud platform, and the data message information corresponding to the data to be analyzed is generated through the large model cloud platform. Determine whether the data message information is used to indicate that the data to be analyzed meets preset data analysis conditions; When it is determined that the data message information indicates that the data to be analyzed meets the preset data analysis conditions, the full-link data corresponding to the target device is extracted through the large model cloud platform, and a data packaging operation is performed on the full-link data to obtain the link data packaging result. Based on the link data packaging result, intelligent agent analysis data is generated. When it is determined that the data packet information indicates that the data to be analyzed does not meet the preset data analysis conditions, a judgment log information corresponding to the data packet information is generated based on the data to be analyzed, the data packet information, and the preset data analysis conditions. An exception prompt information is generated based on the judgment log information, and an adjustment operation is performed on the data packet information according to the exception prompt information to obtain the adjusted data packet information. The operation of determining whether the data packet information is used to indicate that the data to be analyzed meets the preset data analysis conditions is then triggered again.

[0019] A third aspect of this invention discloses another electromechanical equipment evaluation device based on large-scale model intelligent agent collaboration, the device comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in the electromechanical equipment evaluation method based on large model intelligent agent cooperation as described in any of the first aspects of the present invention.

[0020] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the electromechanical equipment evaluation method based on large model intelligent agent collaboration as described in any of the first aspects of the present invention.

[0021] Compared with the prior art, the present invention has the following beneficial effects: In this embodiment of the invention, target device data is collected and transmitted to a data buffer for storage. The target device data is read, and a model-based intelligent agent is invoked to perform data analysis operations to obtain data analysis results. Based on the data analysis results, a device early warning result is generated. Data encryption operations are performed on the data analysis results and the device early warning result to obtain encrypted data results. Based on the encrypted data results and the target device data, a device evaluation result is generated. Evaluation feature data from the device evaluation result is extracted and sent to a large-scale model cloud platform to generate intelligent agent analysis data. Finally, a device evaluation report for the target device is generated based on the intelligent agent analysis data. Therefore, implementing this invention can improve the accuracy and reliability of electromechanical equipment evaluation and increase the efficiency of safety evaluation of electromechanical equipment. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating an electromechanical equipment evaluation method based on large-scale intelligent agent collaboration disclosed in an embodiment of the present invention; Figure 2 This is a flowchart illustrating another electromechanical equipment evaluation method based on large-scale intelligent agent collaboration disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an electromechanical equipment evaluation device based on large-scale intelligent agent collaboration disclosed in an embodiment of the present invention; Figure 4 This is a schematic diagram of another electromechanical equipment evaluation device based on large-scale intelligent agent collaboration disclosed in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of another electromechanical equipment evaluation device based on large model intelligent agent collaboration disclosed in the embodiments of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0026] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0027] This invention discloses a method and apparatus for evaluating electromechanical equipment based on large-scale intelligent agent collaboration, which can improve the accuracy and reliability of safety assessments of electromechanical equipment, and also improve the efficiency of safety assessments of electromechanical equipment. These will be described in detail below.

[0028] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for evaluating electromechanical equipment based on large-scale intelligent agent collaboration, as disclosed in an embodiment of the present invention. Figure 1 The described electromechanical equipment evaluation method based on large-scale intelligent agent collaboration can be applied to electromechanical equipment evaluation devices based on large-scale intelligent agent collaboration, and also to integrated devices for electromechanical equipment evaluation based on large-scale intelligent agent collaboration. The electromechanical equipment evaluation device based on large-scale intelligent agent collaboration can be integrated on a cloud server or a local server; this embodiment of the invention does not impose limitations. Figure 1 As shown, the electromechanical equipment evaluation method based on large-scale intelligent agent collaboration may include the following operations: 101. Collect target device data and transfer the target device data to a preset data buffer for storage.

[0029] In this embodiment of the invention, the target equipment may optionally include large-scale electromechanical equipment, including high-value industrial electromechanical equipment such as wind turbines, generator sets, rolling mills, and tunnel boring machines; the target equipment data may include the equipment's periodic operating condition data, with parameters including vibration amplitude / spectrum, bearing temperature, winding current, oil pressure, rotational speed, operating load, and equipment start-up / shutdown status; the data source is the factory's DCS control system; wherein, the factory DCS control system (Distributed Control System) is an automated system for industrial process control, which distributes control functions to multiple field control stations, while operation and monitoring are centralized in a central control room. Simply put: distributed control, centralized management.

[0030] In this embodiment of the invention, optionally, the preset data buffer may include an isolated storage area independently divided by the terminal, which is divided into multiple partitions according to the business type. The partitions include raw data partitions, analysis result partitions, early warning result partitions, encrypted data partitions, and evaluation result partitions. The buffer scheduling rules are locked by the fixed execution logic, and the upper-layer program cannot modify the partition read and write permissions.

[0031] In optional embodiments of the invention, the collection of target device data may include configuring a fixed collection strategy. After the device is powered on, unmodifiable collection configuration parameters are automatically loaded. These parameters include the collection period, data type, data reporting frequency, and network isolation rules. The fixed logic locks all collection parameters, which cannot be rewritten by the software. Through isolated data collection, the system connects to the factory's DCS system, unidirectionally and periodically reading the target device's real-time operating data. The collection link and the terminal's internal processing link are logically isolated, blocking the reverse data write channel. Preprocessing operations, such as data cleaning and filtering, are performed on the target device data. During the collection phase, invalid data (out-of-range, null values, garbled characters) are removed, retaining only valid target device data. Data is then transmitted unidirectionally through a single transmission path. Only through this system-preset single transmission path, the filtered valid target device data is unidirectionally pushed to the original data-specific partition of the data buffer.

[0032] In this embodiment of the invention, optionally, the target device data can be collected through a data acquisition and protocol conversion module, wherein the data acquisition and protocol conversion module is an industrial communication processor chip U1 (such as one based on the ARM Cortex-M series). Its peripheral circuitry includes: an Ethernet PHY chip U2, connected to U1 via an RMII interface; a network transformer T1 and an RJ45 interface J1, used to connect to the factory DCS network. The UART_TXD1 and RXD1 pins of U1 are connected to the U3 processor of the intelligent agent collaborative scheduling module; this module is responsible for periodically and on demand collecting raw operating data (such as vibration, temperature, current, etc.) of large electromechanical equipment from the factory DCS system. Its hardware connection design ensures that the collected data can only be uploaded to the scheduling module through the designated UART serial port physical path, providing a data source for subsequent intelligent agent collaborative processing.

[0033] 102. Read the target device data from the preset data buffer, and call the pre-determined model computing agent to perform data analysis operations on the target device data to obtain the data analysis results.

[0034] In this embodiment of the invention, optionally, reading target device data from a preset data buffer may include: After the target device data is cached and stored, the reading action is automatically triggered by a fixed serial scheduling logic, without manual or software-based triggering; all target device data collected in this round is completely read from the original data partition of the buffer.

[0035] In this embodiment of the invention, optionally, invoking a pre-determined model computational agent to perform data analysis operations on the target device data to obtain data analysis results may include: After reading the raw device data, the data is first encrypted and uploaded to the cloud. The signal processing agent is then invoked to complete data cleaning, time-domain and / or spectral feature extraction, and a standardized feature dataset is returned. The feature dataset consists of device quantification indicators extracted by the signal processing agent, including vibration time-domain mean, spectral peak value, temperature fluctuation coefficient, and current harmonic components. The feature dataset returned from the cloud is decrypted and verified before being stored in an independent feature partition in the buffer. The scheduling logic automatically packages the feature dataset, forcibly encrypts it, uploads it to the large model cloud platform, and initiates a request to call the model computing agent. The cloud-based model computing agent analyzes device feature data, performs multi-dimensional fault diagnosis calculations, generates data analysis results in the cloud and encrypts and sends them back. The terminal performs dual verification of decryption and signature validity. After successful verification, the data analysis results are stored in a dedicated partition of the buffer to generate the final data analysis results. The signature verification standard is a verification rule used to determine whether the data returned by the cloud is authentic and has not been tampered with.

[0036] In this embodiment of the invention, optionally, the pre-determined model computing agent is a cloud-based model computing agent. The cloud-based model computing agent is a specialized AI agent deployed on a cloud-based large model platform, specifically responsible for in-depth diagnosis of the operating status of large electromechanical equipment. It is the core analysis model in the multi-agent collaborative process solidified in the terminal hardware.

[0037] In this embodiment of the invention, optionally, the data analysis results may include the health score, fault type, confidence level, remaining lifespan, and fault trend of the target device. Furthermore, the data analysis results may be stored in a buffer analysis result partition.

[0038] In this embodiment of the invention, optionally, the data analysis operation may include multi-dimensional equipment state calculation operations performed by the cloud model computing agent, and the calculation content may include vibration spectrum fault matching, bearing loss calculation, winding temperature rise deterioration analysis, load matching degree evaluation, and component loss trend fitting.

[0039] In this embodiment of the invention, optionally, the process of obtaining the data analysis results described above can be implemented through an intelligent agent collaborative scheduling module, wherein the collaborative scheduling module may include the hardware control and scheduling core of the terminal. A high-performance application processor U3 (e.g., i.MX 8M Plus) equipped with a hardware security zone (such as TrustZone) is used. Within the hardware security zone of the U3, or in an external serial flash memory chip F1, an immutable boot and scheduling microprogram is stored. This program has a strict execution order logic fixed at the hardware level. The execution order logic may include reading and receiving raw data from the data acquisition module; formatting the data and requesting a cloud-based signal processing intelligent agent to perform preliminary cleaning and feature extraction; receiving the processed feature data and requesting a cloud-based model calculation intelligent agent to perform state assessment and fault prediction; and receiving the model calculation results, which typically include multi-dimensional information such as device health status score, predictive remaining life, identified potential fault modes (such as imbalance, misalignment, bearing wear, etc.) and their confidence levels, and time trend predictions of fault development.

[0040] In this embodiment of the invention, optionally, U3 in the dedicated data buffer and hardware routing circuit has a dedicated memory area (supported by DDR chip M1) as a temporary data buffer for input requests and output results during the aforementioned intelligent agent invocation process. Its PCIe controller pins are directly connected to the secure communication module (MD1), its SPI controller pins are connected to the secure encryption module (SE1), and its GPIO pin group is connected to the local alarm and display driver module. These physical connections constitute an unavoidable and deterministic hardware-level data scheduling and routing path; the hardware circuits and firmware specifically used in this module are responsible for executing the aforementioned intelligent agent collaborative process in sequence. It controls the entire process of data entering the buffer from the acquisition interface, being encrypted and sent to the cloud, receiving the results returned from the cloud and storing them back in the buffer, and driving local hardware actions based on the conclusions of the warning intelligent agent, ensuring the automation, immutability, and efficient execution of the evaluation process. For example, taking the action of the relay (RL1) as an example, the entire process is as follows: When the decision conclusion returned by the cloud-based early warning agent is "high risk, immediate alarm recommended", this conclusion is received by the secure communication module, decrypted and verified by the secure encryption module, and then transmitted to the application processor of the agent collaborative scheduling module. The scheduling microprogram embedded in U3 parses the conclusion and, according to the predefined hardware action mapping table (such as "high risk" mapping to "trigger hard node output"), directly controls the designated GPIO pin P3 to output a high-level signal (e.g., 3.3V). This level signal, through the physical circuit link of "GPIO pin P3 → LED on the input side of optocoupler (OC1) → phototransistor on the output side of optocoupler (OC1) → base of transistor (Q2) receives bias voltage and saturates and conducts → relay (RL1) coil is energized", finally drives the normally open contact of the relay (RL1) to close. The closed contact outputs a passive dry contact signal through the terminal block (TB1), which can be directly connected to the audible and visual alarm system in the factory's central control room or the emergency stop circuit of the equipment, thereby achieving millisecond-level, deterministic linkage from cloud-based intelligent decision-making to on-site physical equipment control. The entire drive process is implemented entirely by hardware logic and circuits, without relying on real-time scheduling by the upper-level operating system and application software, thus ensuring extremely high reliability and timeliness of response.

[0041] 103. Based on the data analysis results, generate the target device early warning results corresponding to the target device data.

[0042] In this embodiment of the invention, optionally, generating a device warning result for the target device corresponding to the target device data based on the data analysis results may include: Based on the data analysis results, the data is merged and packaged with the original equipment time-series data, encrypted with SE1 hardware, and uploaded to the cloud platform via MD1. The third-level early warning intelligent agent is automatically scheduled, and all diagnostic indicators in the data analysis results are extracted by the third-level early warning intelligent agent. The indicators are compared line by line with the preset evaluation and estimation thresholds in the database to generate data comparison results. Based on the comparison results, standardized risk decision parameters (risk level, maintenance suggestions, confidence score) are output. Based on the data comparison results and standardized risk decision parameters, the data is packaged into a structured early warning data package, and the equipment early warning result corresponding to the target equipment data is generated based on the structured early warning data package.

[0043] In this embodiment of the invention, optionally, after being encapsulated into a structured warning data packet, it can be encrypted and sent to terminal MD1; after SE1 decrypts and verifies, U3 stores the plaintext warning data into the warning result partition of the DDR buffer to complete the generation of the device warning result.

[0044] In this embodiment of the invention, optionally, the device warning result of the target device may include complete risk warning structured data of the target device output by the end warning agent, which may include device number, warning occurrence time, out-of-limit parameters, risk level, fault confidence, and shutdown or maintenance suggestions.

[0045] 104. Perform data encryption on the data analysis results and equipment early warning results to obtain encrypted data results, and generate equipment evaluation results for the target equipment based on the encrypted data results and the target equipment data.

[0046] In this embodiment of the invention, optionally, the above-mentioned data encryption operation on the data analysis results and equipment warning results to obtain the data encryption result may include: The data analysis results and equipment warning results are merged into a unified data packet. The unified data packet is sent to the national cryptographic SE1 security chip through the SPI forced hardware path. SE1 performs batch symmetric encryption and integrity digital signature, outputs ciphertext data packets, and generates data encryption results based on the ciphertext data packets.

[0047] In this embodiment of the invention, optionally, the data encryption operation may include a dual encryption method of batch symmetric encryption and digital signature encryption; the data encryption result may include a ciphertext data packet encrypted with SE1 and with an additional tamper-proof digital signature.

[0048] 105. Extract the evaluation feature data of the equipment evaluation results, send the evaluation feature data to the big model cloud platform, generate intelligent agent analysis data based on the big model cloud platform, and generate the equipment evaluation report of the target equipment based on the intelligent agent analysis data.

[0049] In this embodiment of the invention, optionally, the evaluation feature data extracted from the equipment evaluation results may include: Read the device evaluation results from the DDR buffer, filter and extract core feature fields according to preset fields, remove redundant original sampling points, and generate evaluation feature data.

[0050] In this embodiment of the invention, optionally, the process of sending the evaluation feature data to a large model cloud platform and generating agent analysis data based on the large model cloud platform may include: The evaluation feature data is encrypted to obtain encrypted feature data, which is then sent to the large model cloud platform. The large model cloud platform retrieves the full-link historical data of the target device based on the encrypted feature data. The full-link historical data of the target device includes raw time series data, signal processing features, model diagnostic conclusions, risk warning parameters, device identification, and evaluation timestamp data. Based on the target device's full-link historical data, perform link data analysis operations on the full-link historical data to obtain link data analysis results. The link data analysis operations may include data fusion, indicator correlation analysis, and fault root cause reasoning analysis. Based on the link data analysis results, standardized agent analysis data is output.

[0051] In this embodiment of the invention, optionally, the above-mentioned generation of the device evaluation report for the target device based on the agent analysis data may include: Based on the built-in standardized industrial report templates and intelligent agent analysis data, the system automatically fills in data tables, draws trend charts, generates fault analysis and maintenance suggestions in natural language, and outputs an equipment evaluation report for the target equipment.

[0052] In this embodiment of the invention, optionally, the cloud encrypts and sends the complete evaluation report back to the terminal, SE1 decrypts and verifies it and stores it in the DDR buffer, and U3 pushes it to the local LCD screen for display through the LVDS hardware interface, supporting local access and export.

[0053] In this embodiment of the invention, optionally, for example, the intelligent agent collaborative scheduling module then encrypts the data packet containing the model calculation results again through the security encryption module, and sends it to the cloud-based early warning intelligent agent via the secure communication module, requesting it to perform risk assessment and decision-making. The cloud-based early warning intelligent agent has a pre-configurable assessment rule base (e.g., a risk matrix set based on expert experience or industry standards). Its specific assessment process includes: first, parsing the model calculation results and extracting key indicators (such as fault severity, probability of occurrence, and rate of development trend); second, comparing and performing logical operations between the indicators and thresholds in the rule base (e.g., "IF health score < 60 AND fault confidence > 85% THEN"). (Risk level = High); Finally, a structured risk decision conclusion is generated, which includes at least a clear risk level (such as "Normal", "Caution", "Warning", "High Risk"), recommended measures (such as "Continue to observe", "Planned maintenance", "Immediate shutdown for inspection"), and an optional confidence score, and this conclusion is returned to the terminal; Finally, a cloud report is requested to generate an intelligent agent that integrates the entire process data and conclusions. The intelligent agent collaborative scheduling module will automatically extract the full-link data of this evaluation task from the buffer (M1), including: raw time-series data obtained from the data acquisition and protocol conversion module, feature data returned by the signal processing intelligent agent (such as spectral features, time-domain statistics), status and fault analysis results returned by the model calculation intelligent agent, risk decision conclusions returned by the warning intelligent agent, and metadata such as device identifier and evaluation timestamp. The scheduling module packages this data and requests a cloud report to generate an intelligent agent. The report-generating agent has built-in report templates and natural language generation capabilities. The specific process for generating reports is as follows: it automatically fills in the corresponding sections of the report template (such as "Test Data Summary," "Feature Analysis," "Intelligent Diagnostic Results," and "Risk Warnings and Recommendations") based on the data, converting numerical results and conclusions into descriptive text, and potentially automatically generating trend charts for key indicators. Finally, it generates a complete and directly readable evaluation report (usually in PDF or HTML format) and returns it to the terminal. The agent's collaborative scheduling module stores the report, and it can be accessed and displayed on the LCD screen of the local alarm and display driver module.

[0054] It is evident that implementation Figure 1The described electromechanical equipment evaluation method based on large-scale model intelligent agent collaboration can collect target equipment data, transmit the target equipment data to a data buffer for storage, read the target equipment data and call the model computation intelligent agent to perform data analysis operations to obtain data analysis results, generate equipment early warning results based on the data analysis results, perform data encryption operations on the data analysis results and equipment early warning results to obtain encrypted data results, generate equipment evaluation results based on the encrypted data results and target equipment data, extract evaluation feature data from the equipment evaluation results and send it to the large-scale model cloud platform to generate intelligent agent analysis data, and generate equipment evaluation reports for the target equipment based on the intelligent agent analysis data. This method can solve the risks of industrial equipment data tampering and leakage through end-to-end encrypted storage and transmission. The method also includes data analysis results and equipment early warnings. The unified hardware encryption generates encrypted data, and the generation of evaluation results is only allowed after decryption and verification. This improves the security and reliability of the data used in the evaluation process. Furthermore, through hierarchical collaboration of multi-level large-scale model agents, the accuracy and reliability of electromechanical equipment evaluation are significantly improved. The model computing agent performs multi-dimensional fault analysis based on complete equipment time-series data and outputs standardized data analysis results, providing high-quality evaluation feature data for the cloud agent. This further improves the accuracy of cloud-based root cause analysis and report generation. It can also automatically schedule multi-level cloud agents to complete analysis, early warning, evaluation, and report generation, which helps improve the intelligence and efficiency of evaluation results and report generation, thereby improving the accuracy and reliability of electromechanical equipment evaluation and the efficiency of safety evaluation of electromechanical equipment.

[0055] Example 2 Please see Figure 2 , Figure 2 This is a flowchart illustrating a method for evaluating electromechanical equipment based on large-scale intelligent agent collaboration, as disclosed in an embodiment of the present invention. Figure 2 The described electromechanical equipment evaluation method based on large-scale intelligent agent collaboration can be applied to electromechanical equipment evaluation devices based on large-scale intelligent agent collaboration, and also to integrated devices for electromechanical equipment evaluation based on large-scale intelligent agent collaboration. The electromechanical equipment evaluation device based on large-scale intelligent agent collaboration can be integrated on a cloud server or a local server; this embodiment of the invention does not impose limitations. Figure 2 As shown, the electromechanical equipment evaluation method based on large model intelligent agent collaboration may include the following operations.

[0056] 201. Collect target device data and transfer the target device data to a preset data buffer for storage.

[0057] 202. Read the target device data from the preset data buffer, and call the pre-determined model computing agent to perform data analysis operations on the target device data to obtain the data analysis results.

[0058] 203. Based on the data analysis results, generate the target device early warning results corresponding to the target device data.

[0059] 204. Perform data encryption on the data analysis results and equipment early warning results to obtain encrypted data results, and generate equipment evaluation results for the target equipment based on the encrypted data results and the target equipment data.

[0060] 205. Extract the evaluation feature data of the equipment evaluation results, send the evaluation feature data to the big model cloud platform, generate intelligent agent analysis data based on the big model cloud platform, and generate the equipment evaluation report of the target equipment based on the intelligent agent analysis data.

[0061] In this embodiment of the invention, for other descriptions of steps 201-205, please refer to the detailed description of steps 101-105 in Embodiment 1. These descriptions will not be repeated in this embodiment of the invention.

[0062] 206. Based on the equipment assessment report, determine the abnormal parameters of the target equipment, and generate equipment alarm information based on the abnormal parameters.

[0063] In this embodiment of the invention, optionally, determining the abnormal equipment parameters corresponding to the target equipment based on the equipment evaluation report may include: The equipment evaluation report is decrypted and its digital signature is verified for validity. After verification, the report is stored in plaintext in the dedicated partition for evaluation results in DDR memory. The memory hardware controller directly reads the complete evaluation report within the partition, and the fixed scheduler automatically parses all quantitative evaluation indicators, fault conclusions, and risk level fields in the report; and determines the abnormal parameters of the target device based on all the parsed quantitative evaluation indicators, fault conclusions, and risk level fields.

[0064] In this embodiment of the invention, optionally, the process of generating device alarm information based on device anomaly parameters may include: Based on the abnormal parameters of the equipment, the multi-dimensional indicators of the equipment in the evaluation report are traversed line by line. The multi-dimensional indicators of the equipment may include one or more of the following: equipment health score, failure confidence of each component, remaining life of the component, vibration, temperature, current over-limit amplitude, and failure development trend rate. Compare the multi-dimensional indicators of the equipment with the corresponding thresholds, and filter out all indicators that exceed the normal range or reach the warning / high-risk range, and collect them into a set of abnormal equipment parameters. The selected abnormal equipment parameters are encapsulated in a structured manner by a fixed scheduling program to form a standardized and parsable structured alarm data packet, and equipment alarm information is generated based on the structured alarm data packet.

[0065] In this embodiment of the invention, optionally, alarm information is locally cached in a separate alarm partition of the DDR buffer to permanently retain the original alarm record.

[0066] In this embodiment of the invention, optionally, the equipment evaluation report can be a standardized equipment diagnostic document automatically rendered by a cloud-based report generation agent based on end-to-end data. Further, the equipment evaluation report may include the target equipment's basic information, original operating data summary, spectrum, time-domain feature analysis, model-diagnosed fault type, fault confidence, health score, remaining lifespan, risk level, operation and maintenance suggestions, and trend charts.

[0067] In this embodiment of the invention, optionally, the abnormal equipment parameters can be a set of quantitative indicators that exceed the safety threshold in the target equipment, characterize the equipment for faults, and indicate the risk of deterioration. For example, the abnormal equipment parameters may include bearing temperature exceeding the upper limit, vibration spectrum fault amplitude exceeding the standard, winding current fluctuation too large, remaining service life below the threshold, fault confidence > 85%, and health score < 60 points.

[0068] In this embodiment of the invention, optionally, the device alarm information is a standardized structured instruction data packet that integrates abnormal parameters, device identity, timestamp, and risk level, and serves as the upper-level instruction source for scheduling hardware alarm circuits; furthermore, the device alarm information may include device number, alarm trigger time, all abnormal parameter names and their exceeding values, corresponding risk level (attention / warning / high risk), faulty component location, and suggested handling measures.

[0069] 207. Based on the equipment alarm information and the pre-determined alarm components, generate alarm control parameters corresponding to the alarm components, and control the alarm components to perform alarm operations that match the alarm control parameters to obtain alarm results.

[0070] In this embodiment of the invention, optionally, the pre-determined alarm component may include one or more of LED indicator lights, buzzers, and relays, wherein the alarm component is a physical hardware device used for risk warning in the terminal local alarm and display driver module.

[0071] In this embodiment of the invention, optionally, the alarm control parameters corresponding to the alarm components are the underlying hardware driver control instructions for each alarm component. Specifically, they may include LED control parameters, buzzer control parameters, and relay pre-control parameters. The LED control parameters may include high / low level, flashing frequency, and constant-on duration; the buzzer control parameters may include PWM duty cycle, sound interval, and sound frequency; and the relay pre-control parameters may include level threshold and output duration.

[0072] In this embodiment of the invention, optionally, the alarm operation is a physical sound and light prompt action performed by the alarm component according to the alarm control parameters. Specifically, the alarm operation may include an LED indicator light that is constantly on or flashing, a buzzer that sounds intermittently or continuously, and a relay that is pre-engaged in preparation for interlocking.

[0073] In this embodiment of the invention, optionally, the above-mentioned generation of alarm control parameters corresponding to the alarm component based on the device alarm information and the pre-determined alarm component may include: Based on the device alarm information, the risk level field within the alarm information is extracted, and the mapping table is matched to determine the alarm component that needs to be activated; the structured instructions in the device alarm information are parsed and converted into underlying hardware control commands to generate exclusive alarm control parameters for each type of alarm component.

[0074] In this embodiment of the invention, optionally, the above-mentioned control alarm component performs an alarm operation matching the alarm control parameters to obtain an alarm result, which may include: The alarm control unit executes alarm operations that match the alarm control parameters. The corresponding pin level is modified through the internal GPIO controller. The hardware circuit directly drives the corresponding alarm unit. The hardware circuit has built-in pin status readback logic to generate alarm results. The alarm results may include the alarm component type, trigger level, duration, risk level, and whether the hardware driver is functioning correctly.

[0075] In this embodiment of the invention, optionally, the alarm operation can be implemented through a hard-node alarm unit. The hard-node alarm unit includes a GPIO pin P3 of U3 connected to the positive terminal of the LED of optocoupler OC1. The output side of OC1 controls transistor Q2 to drive the coil of relay RL1. The normally open contact of RL1 is led out to terminal TB1. Furthermore, the hard-node alarm unit can convert the instructions generated by U3 based on the collaborative results of the cloud-based intelligent agent into direct, deterministic hardware actions. Specifically, the application processor's fixed scheduler parses the structured instructions (such as warning conclusions) returned from the cloud into predefined hardware control commands and directly operates the level state of the corresponding pins through its GPIO controller. These hardware actions specifically include: level signal output: controlling a specific GPIO pin to output a high or low level. For example, pin P1 outputs a high level to light up a red LED (D1) for visual warning, and pin P2 outputs a PWM wave to drive a transistor to make a buzzer (LS1) sound for audible warning. Relay Drive: As mentioned above, the opening and closing of the relay (RL1) contacts is controlled by the isolation drive circuit through the level change of pin P3, providing dry contact signals for interlocking control. Display Control: Evaluation reports, real-time status, and other screen information are transmitted to the LCD screen for display via the LVDS interface. The essence of these actions is the lowest-level electrical signal change generated by the application processor (U3) through its hardware peripheral interface. This change directly drives the peripheral actuators through the preset circuit path on the PCB, realizing the conversion from digital instructions to physical effects (light, sound, mechanical contacts). This physical link from GPIO to the relay drive circuit is key to realizing cloud-based intelligent driving of local hardware actions.

[0076] 208. Based on the alarm results, obtain the real-time environmental information corresponding to the target device, determine the risk warning information of the current environment where the target device is located based on the real-time environmental information, and generate an isolated passive switch interlock signal based on the risk warning information.

[0077] In this embodiment of the invention, the isolated passive switch interlock signal is used to control the alarm component or target device to perform a shutdown operation.

[0078] In this embodiment of the invention, optionally, the real-time environmental information corresponding to the target device may include the surrounding working environment parameters collected by the on-site environmental sensors of the target electromechanical equipment and the DCS system. Further, the real-time environmental information corresponding to the target device may include the ambient temperature, air humidity, dust concentration, power supply voltage, environmental vibration value, and fire detector status of the equipment room.

[0079] In this embodiment of the invention, optionally, the risk warning information of the current environment is a comprehensive safety risk description data that integrates equipment fault alarms and on-site environmental exceedances; wherein, the risk warning information of the current environment may include equipment fault risk, environmental exceedance risk alone, and dual high-risk warning information of equipment and environment. For example, the risk warning information of the current environment may include environmental exceedance items, exceedance values, environmental risk level, and a description of the safety hazards after superimposed with equipment faults.

[0080] In this embodiment of the invention, optionally, the isolated passive switch interlock signal is a switch on / off signal output by a relay passive dry contact after optocoupler electrical isolation. It has no built-in power supply and only provides the contact on / off status. It is used to connect to the equipment emergency stop circuit and the plant's main alarm system to control the alarm component to continuously alarm at the highest level, or to directly control the target electromechanical equipment to shut down.

[0081] In this embodiment of the invention, optionally, obtaining the real-time environmental information corresponding to the target device based on the alarm result may include: Based on the alarm results, the risk level of the target equipment is determined, and based on the risk level of the target equipment, the data acquisition command is automatically sent to the data acquisition protocol to synchronously collect data from the surrounding environmental sensors of the electromechanical equipment in order to generate real-time environmental information corresponding to the target equipment.

[0082] In this embodiment of the invention, optionally, the risk warning information for determining the current environment of the target device based on real-time environmental information may include: Based on real-time environmental information, safety thresholds are compared item by item. When environmental parameters exceed the standard and the equipment itself has a high-risk fault, it is determined that there is a double safety hazard. The equipment fault and environmental hazard content are integrated to generate a structured current environmental risk warning information.

[0083] In this embodiment of the invention, optionally, the generation of the isolated passive switch interlock signal based on the risk warning information may include: Based on the risk warning information, a relay hardware control instruction is generated, which outputs a continuous high-level signal to the GPIO pin P3.

[0084] In this embodiment of the invention, optionally, for example, a freewheeling diode D2 is connected in parallel across the coil of a preset relay RL1 to protect the circuit. The normally open contact of RL1 outputs a passive dry contact switch signal, i.e., an isolated passive switch quantity interlocking signal, through the terminal TB1. The optocoupler OC1 achieves complete electrical isolation between the internal control circuit of the terminal and the external factory power circuit, avoiding interference and short circuit damage to the main control chip.

[0085] It is evident that implementation Figure 2The described electromechanical equipment evaluation method based on large-scale model intelligent agent collaboration can determine the abnormal parameters of the target equipment according to the equipment evaluation report, generate equipment alarm information based on the abnormal parameters, generate alarm control parameters corresponding to the alarm components based on the equipment alarm information and pre-determined alarm components, and control the alarm components to perform alarm operations matching the alarm control parameters to obtain alarm results. Based on the alarm results, the method obtains the real-time environmental information corresponding to the target equipment, determines the risk warning information of the current environment in which the target equipment is located, and generates isolated passive switch interlocking signals based on the risk warning information. It can use the standardized equipment evaluation report generated by the cloud-based large-scale model intelligent agent as the alarm data source. The report integrates multi-dimensional information, rather than relying solely on a single simple threshold judgment, which can accurately filter out abnormal parameters and avoid the ambiguity and large amount of meaningless alarms in traditional single-threshold alarms. Addressing false alarms and missed alarms improves on-site maintenance efficiency. Hardware-layered driving of audible and visual alarm devices makes on-site risks readily perceptible, with a highly deterministic response chain. On-site personnel can quickly identify the severity of risks, enhancing user intuitiveness and convenience. The entire alarm process is traceable, facilitating subsequent maintenance review and safety audits. Automatic collection of real-time environmental information surrounding equipment enables simultaneous identification of external environmental hazards, achieving integrated assessment of equipment malfunctions and on-site environmental risks. This covers all dimensions of equipment operation safety risks, improving the safety and reliability of equipment evaluation. Furthermore, it can execute shutdown operations based on passive interlock signals, maximizing production and operational efficiency while ensuring safety standards. This enhances the overall intelligent level of safety management for large-scale electromechanical equipment clusters, improving the accuracy and reliability of electromechanical equipment evaluation, and increasing the efficiency of safety assessments.

[0086] In an optional embodiment, the evaluation feature data is sent to a large model cloud platform, and agent analysis data is generated based on the large model cloud platform, including: Based on the evaluation feature data, an evaluation data threshold matching the evaluation feature data is determined in a pre-defined judgment rule database, and a data comparison operation is performed between the evaluation feature data and the evaluation data threshold to obtain the data comparison results; Based on the data comparison results, the risk decision parameters corresponding to the data comparison results are determined. The risk decision parameters include at least the risk level of the target equipment, the risk recommendation measures, and the confidence score corresponding to the data comparison results. By sending risk decision parameters, data comparison results, and target device data to the big model cloud platform, the full-link data corresponding to the target device can be extracted through the big model cloud platform, and intelligent agent analysis data can be generated based on the full-link data. The end-to-end data includes at least the original time-series data corresponding to the target device, evaluation feature data, risk decision conclusions, device identifier, and evaluation timestamp.

[0087] In this optional embodiment, the process of determining the evaluation data threshold matching the evaluation feature data in a pre-determined judgment rule database based on the evaluation feature data may include: Read the evaluation feature data and, based on a pre-determined evaluation rule database, match the evaluation data thresholds of the corresponding target device in the evaluation rule database, wherein the number of evaluation data thresholds is at least one.

[0088] In this optional embodiment, the above-mentioned data comparison operation between the evaluation feature data and the evaluation data threshold to obtain the data comparison result may include: Each evaluation feature field is numerically compared with the matched evaluation data threshold dimension by dimension to obtain sub-comparison results. Based on all sub-comparison results, a structured data comparison result is generated, and thus the data comparison result is obtained.

[0089] In this optional embodiment, the evaluation feature data may optionally include the equipment health comprehensive score, fault types of each component, fault confidence, remaining life of the component, fault trend rate, vibration, temperature, and current core exceedance indicators; the pre-determined judgment rule database may include normal thresholds, warning thresholds, high-risk thresholds, and fault classification judgment logic for each monitoring indicator of different electromechanical equipment; the evaluation data thresholds may include the upper limit of normal values, the lower limit of normal values, the warning trigger threshold, and the emergency shutdown threshold; the data comparison results may include a complete structured dataset output by cloud comparison calculation, recording the deviation of each feature indicator from the threshold; further, it may specifically include the indicator name, actual feature value, matching threshold, difference, exceedance mark, and deviation percentage.

[0090] In this optional embodiment, the above-mentioned determination of risk decision parameters corresponding to the data comparison results based on the data comparison results may include: Based on the data comparison results, all out-of-limit indicators, deviation ranges, and fault feature matching tags are extracted. Based on the risk classification mapping matrix built into the early warning intelligent agent, standardized risk decision parameters are automatically matched and output according to the comparison results. Among them, the risk level of the target equipment can include the number of out-of-limit indicators and the deviation range; risk recommendation measures can include continuous monitoring, planned shutdown for maintenance, immediate power outage, and component replacement; confidence score can include comprehensive fault feature matching degree, data integrity, and sensor accuracy.

[0091] In this optional embodiment, the original time-series data may include the original continuous sampling data of the device sensors; the device feature data may include the time-domain and spectral fault features extracted by the cloud signal processing agent; the risk decision conclusion may include the complete set of risk decision parameters output in the previous step; the device identifier may include the unique hardware number of the device, used to distinguish different electromechanical devices in the cloud; and the evaluation timestamp may include the standard time markers for the start and completion of this evaluation task.

[0092] As can be seen, implementing this optional embodiment can determine the assessment data threshold matching the assessment feature data in a pre-determined judgment rule database based on the assessment feature data, and perform a data comparison operation between the assessment feature data and the assessment data threshold to obtain the data comparison result. Based on the data comparison result, the risk decision parameters corresponding to the data comparison result are determined. The risk decision parameters, data comparison result, and target device data are sent to the large model cloud platform to extract the full-link data corresponding to the target device through the large model cloud platform. Based on the full-link data, intelligent agent analysis data is generated. It can match the corresponding device-specific assessment data threshold based on a unified and professional judgment rule database, which helps to improve the reliability and accuracy of the assessment data threshold and prevent the judgment deviation of the assessment threshold. It can also simultaneously generate three types of core risk decision parameters, namely risk level, risk recommendation measures, and confidence score, based on the data comparison result, forming a complete and verifiable set of risk decision parameters. The terminal hardware analyzes the structured risk dataset, and the terminal simultaneously uploads three core data categories: risk decision parameters, data comparison results, and original target equipment data. The cloud platform can conduct comprehensive analysis by combining the real-time assessment data with the equipment's historical archived full-link data. The full-link data covers original time-series sampling data, assessment characteristics, risk decision conclusions, unique equipment identifiers, and assessment timestamps, completely preserving all process information of the equipment from data collection to risk assessment. This enables full-process tracing and in-depth root cause analysis of equipment failures. It also enables secure interaction, with risk decisions, comparison results, and original equipment data transmitted in encrypted form and verified by hardware signatures throughout the process. While achieving accurate and intelligent assessment, it ensures that sensitive equipment operating data is not leaked or tampered with, balancing intelligent analysis capabilities with industrial data security requirements. This improves the security of data processing and transmission, which is conducive to improving the accuracy and reliability of electromechanical equipment assessment and the efficiency of safety assessment of electromechanical equipment.

[0093] In another optional embodiment, based on the device alarm information and the pre-determined alarm components, alarm control parameters corresponding to the alarm components are generated, including: Based on the device alarm information and the pre-determined alarm components, the structured instructions corresponding to the device alarm information are parsed into predefined hardware control commands through the application processor's fixed scheduler. The hardware control commands determine the level state of the corresponding pins through the GPIO controller, and generate alarm control parameters corresponding to the alarm components based on the level state.

[0094] In this optional embodiment, the above-mentioned parsing of the structured instructions corresponding to the device alarm information into predefined hardware control commands through the application processor's fixed scheduler, based on the device alarm information and the pre-determined alarm components, may include: Read device alarm information, determine the binding relationship of alarm components, extract the core fields in the alarm information, and filter the alarm components that need to be activated this time based on the binding relationship of alarm components; The equipment alarm information is decomposed into fields to obtain the equipment alarm decomposition results. The equipment alarm decomposition results may include equipment number, alarm trigger time, abnormal parameters, risk level, and handling suggestions. Based on the device alarm analysis results, and according to the risk level, the predefined hardware action rule library in the flash memory is matched to translate the upper-layer business structured alarm instructions into hardware control commands in a fixed format that can be recognized by the lower-layer chip, so as to generate hardware control specifications.

[0095] In this optional embodiment, optionally, the hardware control command determines the level state of the corresponding pin through the GPIO controller, and generates alarm control parameters corresponding to the alarm component based on the level state, which may include: The U3 chip integrates a hardware GPIO controller, which directly reads the hardware control commands stored in the U3's on-chip cache. Based on the hardware control commands, the GPIO controller locates the dedicated pins of the mapped alarm components, and the hardware latch pins output levels. The GPIO controller has a built-in hardware readback circuit to collect the current pin output status in real time: level high / low, PWM frequency, waveform duty cycle, and level latch duration; the U3 fixed scheduler reads all pin hardware status data, packages them according to alarm components, and generates independent alarm control parameters for each alarm component.

[0096] In this optional embodiment, the GPIO controller (General Purpose Input / Output Controller) is a hardware peripheral module integrated inside the terminal application processor (the U3 chip in Document 1, such as i.MX8MPlus). Its function is to directly control the level state of the chip pins and it is the core hardware bridge connecting "cloud intelligent decision-making" and "local physical execution".

[0097] In this optional embodiment, the U3 chip is optionally the core main control chip of the intelligent agent collaborative scheduling module, equipped with a hardware security area and a built-in GPIO hardware controller, used to carry the fixed scheduling program, complete instruction parsing, and output pin level control signals.

[0098] In this optional embodiment, the solidified scheduler is optionally an immutable microprogram stored in flash memory and written to the hardware security zone. It is a dedicated hardware scheduling kernel used to read alarm information, match alarm component mapping tables, parse structured alarm instructions, and output hardware control commands.

[0099] In this optional embodiment, the structured instructions are standardized layered instruction texts encapsulated from device alarm information, which are machine-readable and can distinguish risk levels and alarm types; the predefined hardware control commands are low-level binary instructions of the chip, which are only used to drive the GPIO controller to adjust the pin level, have no business semantics, and are pure hardware drive instructions. In this optional embodiment, the level state of the corresponding pin is the electrical signal state output in real time by the GPIO pin of each alarm component, which is divided into two categories: DC level and PWM pulse waveform; the alarm control parameters are a complete set of driving parameters for a single alarm component, which are quantized and encapsulated by the pin level state and are the underlying standard parameters for driving the hardware sound and light and relay actions.

[0100] As can be seen, implementing this optional embodiment can parse the structured instructions corresponding to the device alarm information into predefined hardware control commands based on the device alarm information and the pre-determined alarm components through the application processor's fixed scheduler. The hardware control commands are then used by the GPIO controller to determine the level state of the corresponding pins, and alarm control parameters corresponding to the alarm components are generated based on the level state. This approach uses a hardware-fixed scheduler to parse alarm instructions, eliminating software dependency. The alarm control process is stable, reliable, and tamper-proof, which improves the security and reliability of data processing and alarm handling. Furthermore, it enables standardized parsing of structured alarm information without the need for manual configuration of conversion scripts, thus achieving alarm functionality. The fully automated instruction parsing reduces operational errors and configuration omissions caused by manual configuration, which helps improve the accuracy and reliability of alarms, as well as their security, intelligence, and efficiency. It can also convert structured business instructions into standardized hardware commands, resulting in stronger compatibility and scalability, significantly reducing the cost and development cycle of terminal hardware modification. Furthermore, it can generate standardized alarm control parameters based on hardware levels, making alarm actions fully quantifiable, traceable, and easy for operation and maintenance audits. This provides complete hardware-level evidence for equipment fault tracing and safety production audits, improving safety performance. In turn, it helps improve the accuracy and reliability of electromechanical equipment assessments, as well as the efficiency of safety assessments of electromechanical equipment.

[0101] In yet another optional embodiment, transmitting the target device data to a preset data buffer for storing the target device data includes: The target device data is transmitted to a preset data buffer through a unique and fixed transmission channel. The original data inside the data buffer is retrieved. Based on the original data inside the data buffer, hardware-level encryption and digital signature operations are performed on the target device data to obtain encrypted device data. Data processing operations are performed on the encrypted data of the device to obtain the device-stored data, and the device-stored data is stored in the data buffer; the data processing operations include data cleaning operations, data removal operations, and data feature extraction operations.

[0102] In this optional embodiment, optionally, the process of transmitting the target device data to a preset data buffer via a unique fixed transmission channel and retrieving the original data within the data buffer may include: The target device data is transmitted unidirectionally to a preset data buffer via a single fixed transmission channel. Based on the target device data uploaded to the data buffer, the cached internal raw data within the raw data partition is read by the memory hardware controller. This internal raw data may include data that serves as a basis for digital signature hash calculations, used to verify the continuity and consistency between the currently collected data and historical data, and to identify data gaps and malicious tampering.

[0103] In this optional embodiment, optionally, the above-mentioned hardware-level encryption and digital signature operation on the target device data based on the internal original data to obtain encrypted device data may include: Based on the internal raw data, the algorithm performs hardware encryption on the internal raw data and generates a hash digest by combining the internal raw data in the buffer. An immutable digital signature is then attached to the encrypted ciphertext to obtain the device encrypted data.

[0104] In this optional embodiment, the above-described data processing operation on the encrypted device data to obtain the device-stored data, and storing the device-stored data in a data buffer, may include: The encrypted data from the device undergoes three layers of processing in a fixed serial order to obtain the device's stored data. The processing object is the plaintext data after decryption and verification within the encrypted data packet. The data cleaning operation includes filtering out empty values, garbled characters, invalid sampling points due to sensor jumps, and abnormal values ​​exceeding the range; unifying the data format; and filling in missing timestamps. The data removal operation includes removing instantaneous impacts during device start-up and shutdown, fluctuations at power-on, and instantaneous data with no business value generated by human debugging, while retaining steady-state valid operating data. The data feature extraction operation includes extracting basic shallow features from the complete time-series data, including simple features such as vibration mean / peak value, temperature extreme value, current fluctuation variance, speed range, and load average value. The processed device storage data is stored in a data buffer, which is partitioned by hardware isolation and physically isolated from the encryption and analysis partitions to prevent data overlap.

[0105] In this optional embodiment, the device storage data may optionally include encrypted valid original time-series data, extracted basic device shallow features, device digital signature, and complete timestamp.

[0106] In this optional embodiment, data encryption can be performed via a secure encryption module, specifically using a security chip SE1 (e.g., model SJT2001) certified to Level 2 or higher of the Chinese national cryptographic standard as the core. The SE1's SPI slave interface (MISO, MOSI, SCLK, CS) is directly connected to the application processor U3's SPI host interface via PCB traces. Additionally, one of the SE1's USB Device interfaces is directly connected to the USBHost interface of the secure communication module MD1 via an onboard connector. The SE1's general-purpose I / O pins (e.g., GPIO0) are connected to the indicator light (e.g., power / status indicator) driver circuit in the local alarm and display driver module via pull-up resistors. The core innovation of this module lies in establishing a mandatory hardware encryption link. All business data (e.g., raw data, agent requests) sent from U3 to the cloud-based large-scale intelligent agent platform must first be transmitted to SE1 via the SPI physical link, where SE1 performs encryption and digital signature. Conversely, all data packets received from the cloud from MD1 must also be sent to SE1 via the USB link for decryption and signature verification. The SPI and USB physical connections constitute an indispensable hardware security barrier for the data inflow and outflow scheduling module, ensuring "data remains confidential, and encrypted data is transmitted only." This provides hardware-based, high-security encryption, decryption, and authentication services for all data interactions between the terminal and the cloud. Specifically, this includes encrypting and signing business data sent by the U3 to ensure confidentiality, integrity, and non-repudiation; and decrypting and verifying the signature of data received by the MD1 to ensure its authenticity and tamper-proof nature. This hardware-based design eliminates the risk of side-channel attacks at the software level and forms the cornerstone of the terminal's overall security architecture.

[0107] In this optional embodiment, storing data in a data buffer can be achieved via a secure communication module, which may be an industrial-grade 5G communication module MD1. Its PCIe interface directly interfaces with the PCIe interface of U3 via an onboard connector, and its USB interface connects to the USB interface of SE1 (for transmitting encrypted service data streams); it provides a physical wireless channel. Its dual hardware interfaces with both U3 and SE1 ensure reliable transmission of encrypted data streams.

[0108] As can be seen, implementing this optional embodiment enables the target device data to be transmitted to a preset data buffer through a unique and fixed transmission channel. The original data within the data buffer is retrieved, and based on this original data, hardware-level encryption and digital signature operations are performed on the target device data to obtain encrypted device data. Data processing operations are then performed on the encrypted device data to obtain the device's stored data, which is then stored in the data buffer. This method allows for the transmission of target device data through a unique and fixed transmission channel, preventing malicious data injection attacks at the hardware level and ensuring that all data entering the terminal is genuine, original signal collected by the device. Compared to multi-channel universal gateways, this significantly improves the authenticity and traceability of original industrial operating data. The encryption mechanism enables fully cached, encrypted data storage. By generating verification hashes based on historical raw data within the buffer, real-time collected data is bound and verified with historical archived data to ensure data integrity. Furthermore, by integrating three layers of preprocessing—cleaning, removal, and feature extraction—the overall efficiency of end-to-cloud collaboration is improved. This reduces the amount of cache space occupied by invalid data, increases buffer storage utilization, saves industrial network communication costs, and shortens cloud data parsing time. It also improves the accuracy of equipment fault diagnosis and risk assessment from the data source level. Layered processing generates standardized equipment storage data with a unified cache data structure, facilitating subsequent full-process retrieval and parsing. This ensures the stable parallel operation of multiple rounds of data collection tasks, thereby improving the accuracy and reliability of electromechanical equipment assessment and increasing the efficiency of safety assessments.

[0109] In another optional embodiment, based on the data encryption result and the target device data, a device evaluation result for the target device is generated, including: Hardware decryption and signature validity verification are uniformly performed on the encrypted data results to obtain the data verification results, and it is determined whether the data verification results meet the preset verification conditions. When it is determined that the data verification result meets the preset verification conditions, the plaintext feature data corresponding to the data encryption result and the target device data is determined, and the plaintext feature data is stored in an independent cache partition; Read all independent storage data stored in the independent cache partition, generate multi-dimensional evaluation results based on the independent storage data, and generate device evaluation results for the target device based on the multi-dimensional evaluation results; The multi-dimensional evaluation results include at least the target equipment's health score, fault type, fault confidence, remaining lifespan, and fault trend.

[0110] In this optional embodiment, the above-mentioned unified hardware decryption and signature validity verification of the data encryption result to obtain the data verification result may include: The hardware circuit executes the national cryptographic symmetric algorithm on the encrypted data to decrypt the ciphertext, restore the data analysis and warning plaintext stored internally, and extract the digital signature bound to the ciphertext. It then calculates the standard hash digest by combining it with the original data of the target device in the cache and compares whether the signature hash matches the local benchmark hash to obtain the hash matching result. Finally, it integrates the decryption status, signature matching status, data integrity flag, and hash matching identifier into a structured receipt data packet to generate the data verification result.

[0111] In this optional embodiment, the above-mentioned determination of whether the data verification result meets the preset verification conditions may include determining whether the ciphertext format matches, determining whether the digital signature and the local original data hash completely match, determining whether the data packet is free of missing characters, garbled characters, and over-range tampering marks, determining whether the device ID, evaluation timestamp and the local cached target device data correspond one-to-one. If all of the above are verified, it is determined that the data verification result meets the preset verification conditions.

[0112] In this optional embodiment, if it is determined that the data verification result does not meet the preset verification conditions, the process can be terminated, or if the verification fails, an exception log can be directly generated and stored in the flash memory, terminating the subsequent evaluation process and not generating a device evaluation result.

[0113] In this optional embodiment, the hardware decryption and signature validity verification operation may include hardware symmetric decryption, hash digest hardware calculation, digital signature comparison, and data packet integrity verification; the data verification result may include decryption status, signature matching flag, data integrity flag, device ID matching flag, and verification timestamp; the preset verification conditions may include decryption being normal, signature being valid, data being complete, and device metadata matching being true simultaneously.

[0114] In this optional embodiment, the process of determining the plaintext feature data corresponding to the data encryption result and the target device data, and storing the plaintext feature data in an independent cache partition, may include: The decrypted plaintext data analysis data, device warning plaintext data, and original plaintext time-series data of the target device read from the cache are extracted and merged. A unified set of quantitative indicators for the device across all dimensions is extracted to form plaintext feature data. A pre-defined independent cache partition is then called, and the complete plaintext feature data is written in batches to this independent cache partition. This independent cache partition is physically isolated from the original data partition and the encrypted partition. The memory controller has separate read / write permissions, and upper-layer software has no right to modify the partition configuration. It is specifically used to store the verified plaintext feature data.

[0115] In this optional embodiment, the process of reading all independent storage data stored in the independent cache partition, generating a multi-dimensional evaluation result based on the independent storage data, and generating a device evaluation result for the target device based on the multi-dimensional evaluation result may include: The memory controller initiates a hardware batch read instruction to completely read all plaintext feature data in the independent cache partition. It uses the hardware floating-point arithmetic unit to perform fusion calculations on all independently stored data and outputs five core quantitative indicators in parallel to form a complete multi-dimensional evaluation result. The solidified scheduler packages and integrates the multi-dimensional evaluation result with the underlying basic data, standardizes and structures the encapsulation, and generates the device evaluation result. Furthermore, the multi-dimensional assessment results can include equipment health scores, ranging from 0 to 100 points, calculated by weighting temperature, vibration, current, and load degradation; fault types, including fault categories identified by matching feature libraries (bearing wear, rotor imbalance, winding aging, insufficient oil pressure, etc.); fault confidence scores, ranging from 0 to 100 points, characterizing the reliability of the current fault identification results; remaining lifespan, including the estimated usable time of components based on degradation rate and historical operating data; and fault trends, including short-term and medium-to-long-term degradation increase and decrease rate curves.

[0116] In this optional embodiment, the equipment evaluation result of the target device may be a complete comprehensive evaluation data package that integrates basic equipment information, all multi-dimensional quantitative indicators, and risk warning levels, including basic equipment identifier, original operating data summary, a complete set of multi-dimensional evaluation indicators, risk classification, and local evaluation timestamp.

[0117] As can be seen, implementing this optional embodiment can uniformly perform hardware decryption and signature validity verification on the data encryption result, obtain the data verification result, and determine whether the data verification result meets the preset verification conditions. When it is determined that the data verification result meets the preset verification conditions, the plaintext feature data corresponding to the data encryption result and the target device data is determined, and the plaintext feature data is stored in an independent cache partition. All independently stored data stored in the independent cache partition is read, and a multi-dimensional evaluation result is generated based on the independently stored data. Based on the multi-dimensional evaluation result, a device evaluation result for the target device is generated. It can perform dual verification through hardware decryption and signature validity verification to ensure the authenticity and reliability of the evaluation result, and is conducive to improving the security and reliability of data processing. The verified plaintext is stored in the independent cache partition. Hardware isolated storage realizes data layer protection, avoids data cross-contamination, and ensures the accuracy of multi-dimensional evaluation calculations. Based on the stability of the source, an independent partition is dedicated to storing trusted plaintext features, distinguishing various types of business data within the cache. When reading the full set of feature data, there is no need to filter irrelevant and redundant data, improving the reading speed of multi-dimensional evaluation calculations and shortening the latency of generating equipment evaluation results. This is conducive to improving the intelligence and efficiency of data processing and the generation of evaluation results and reports. By integrating multi-dimensional indicators to generate integrated equipment evaluation results, it provides a unified and standardized data source for backend alarms, cloud analysis, and report generation, eliminating the need for repeated data format conversions, simplifying the data interaction logic throughout the entire process, reducing software and hardware development costs, and enabling dual support for local real-time control and cloud-based long-term operation and maintenance analysis. It also enables refined and digital evaluation of equipment status, improving the level of intelligent safety management of electromechanical equipment, thereby improving the accuracy and reliability of electromechanical equipment evaluation and the efficiency of safety evaluation of electromechanical equipment.

[0118] In another optional embodiment, risk decision parameters, data comparison results, and target device data are sent to a large-scale model cloud platform. The large-scale model cloud platform can then extract the full-link data corresponding to the target device, and generate agent analysis data based on this full-link data, including: Based on risk decision parameters, data comparison results and target device data, data to be analyzed is generated, and the data to be analyzed is sent to the big model cloud platform. The big model cloud platform then generates the corresponding data message information of the data to be analyzed. Determine whether the data message information is used to indicate that the data to be analyzed meets the preset data analysis conditions; When it is determined that the data message information indicates that the data to be analyzed meets the preset data analysis conditions, the full-link data corresponding to the target device is extracted through the large model cloud platform, and the full-link data is packaged to obtain the link data packaging result. Based on the link data packaging result, intelligent agent analysis data is generated. When it is determined that the data message information is used to indicate that the data to be analyzed does not meet the preset data analysis conditions, a judgment log information corresponding to the data message information is generated based on the data to be analyzed, the data message information, and the preset data analysis conditions. An exception prompt information is generated based on the judgment log information, and an adjustment operation is performed on the data message information according to the exception prompt information to obtain the adjusted data message information. Then, the operation of determining whether the data message information is used to indicate that the data to be analyzed meets the preset data analysis conditions is triggered again.

[0119] In this optional embodiment, the process of generating data to be analyzed based on risk decision parameters, data comparison results, and target device data, sending the data to be analyzed to the large model cloud platform, and generating corresponding data message information through the large model cloud platform may include: Based on risk decision parameters, data comparison results, and target device data, the three types of business data and metadata are merged into a standardized binary data package to generate data to be analyzed. The data to be analyzed is uploaded to the large model cloud platform. The cloud message verification engine receives the encrypted data packet, completes decryption and format parsing, and automatically generates a structured verification message corresponding to the data to be analyzed, so as to generate data message information.

[0120] In this optional embodiment, the risk decision parameters are optionally a standardized risk judgment set output by the cloud-based early warning agent based on threshold comparison, which may include equipment risk level, risk recommendation measures, and fault confidence score; the large model cloud platform may include a cloud computing power service cluster that deploys four collaborative agents: signal processing, model calculation, early warning, and report generation; the data message information may include a compliance verification message automatically generated by the cloud after receiving the data to be analyzed, used to determine whether the uploaded data conforms to the cloud parsing standard, and may further include message number, device ID, data packet size, field integrity flag, data format check code, missing field list, and message generation timestamp.

[0121] In this optional embodiment, the determination of whether the data message information is used to indicate that the data to be analyzed meets the preset data analysis conditions may include: Iterate through all check flags in the data message information, compare them item by item with the preset data analysis conditions, output binary judgment flags, and determine whether the binary judgment flags are used to indicate that all messages in the data message information meet the standards. When it is determined that the binary decision flag indicates that all messages in the data message information meet the criteria, the data message information is used to indicate that the data to be analyzed meets the preset data analysis conditions; when it is determined that the binary decision flag indicates that all messages in the data message information do not meet the criteria, the data message information is used to indicate that the data to be analyzed does not meet the preset data analysis conditions.

[0122] In this optional embodiment, the process of extracting end-to-end data corresponding to the target device through a large-scale model cloud platform, performing data packaging operations on the end-to-end data to obtain the packaged data, and generating intelligent agent analysis data based on the packaged data may include: The large model cloud platform parses the device ID and evaluation timestamp in the data to be analyzed, retrieves the cloud device archive database, and retrieves all historical data from the entire process of this evaluation task from data collection to risk assessment, and integrates them into a complete end-to-end dataset. The cloud performs standardized compression and layered encapsulation of the entire data chain to form a packaged result of the data chain. Perform multi-dimensional fusion reasoning on the packaged end-to-end data, and after the reasoning is completed, encapsulate and output the intelligent agent analysis data in a structured manner, cache it in the cloud, and generate intelligent agent analysis data.

[0123] In this optional embodiment, the end-to-end data may include all raw, intermediate, and conclusion-type archived data generated during a single complete evaluation process of a single device, such as raw time-series data, device characteristic data, risk decision conclusions, device identifiers, and evaluation timestamps. The data packaging operation is used to hierarchically classify various types of end-to-end data, compress to reduce transmission volume, and generate global hash check codes. The end-to-end data packaging result includes standardized data packets after the end-to-end data has been hierarchically encapsulated, compressed, and appended with check hashes. The agent analysis data is a comprehensive analysis dataset generated by cloud-based reports after the agent integrates the end-to-end packaged data to complete deep inference of a large model. It may include fault root cause deduction, detailed prediction of component remaining life, long-term fault degradation trends, graded operation and maintenance solutions, and horizontal comparison conclusions from multiple rounds of evaluation.

[0124] In this optional embodiment, optionally, a judgment log information corresponding to the data packet information is generated based on the data to be analyzed, the data packet information, and preset data analysis conditions. An exception prompt information is generated based on the judgment log information, and an adjustment operation is performed on the data packet information according to the exception prompt information to obtain the adjusted data packet information. This may include: Based on the data to be analyzed, the data message information, and the preset data analysis conditions, the fields are compared, all non-compliant items are recorded, and all are uniformly encapsulated into structured judgment log information; The system analyzes and judges all abnormal items in the log, generates abnormal prompts, corrects the original data to be analyzed locally according to the correction rules in the abnormal prompts, re-encapsulates the data packet after correction, encrypts it again and uploads it to the cloud, and the cloud re-parses and generates a brand new adjusted data packet information. Local correction of the original data to be analyzed may include completing missing metadata, removing invalid samples that exceed the range, standardizing data field formats, and completing missing sub-data packets.

[0125] In this optional embodiment, the judgment log information is optionally used to record a complete audit log of all non-compliant items, violation basis, original data, and original messages of the uploaded data packet. It may include device ID, message number, all verification items that do not meet the conditions, corresponding preset condition clauses, original data summary to be analyzed, and log generation timestamp. The abnormal prompt information is a standardized fault correction prompt extracted from the judgment log, which clarifies the problems and repair rules of the data packet. It may include missing assessment timestamp, empty confidence field of risk decision parameter, and out-of-range garbled characters in the original device data.

[0126] As can be seen, implementing this optional embodiment can generate data to be analyzed based on risk decision parameters, data comparison results, and target device data. This data is then sent to a large-scale model cloud platform, which generates corresponding data message information. The platform determines whether the data message information indicates that the data to be analyzed meets preset data analysis conditions. If it does, the platform extracts the full-link data corresponding to the target device and performs data packaging to obtain the link data packaging result. Based on this result, intelligent agent analysis data is generated. If not, the platform generates judgment log information corresponding to the data message information based on the data to be analyzed, the data message information, and the preset data analysis conditions. An error message is generated based on the judgment log information, and adjustments are made to the data message information according to the error message to obtain the adjusted data message information. The operation to determine whether the data message information indicates that the data to be analyzed meets the preset data analysis conditions is then retried. This allows for the analysis of multiple... Core data is fused and encapsulated into data to be analyzed. End-to-cloud interaction data is standardized and unified, reducing transmission redundancy and parsing costs. No additional matching indexes are needed, simplifying cloud data parsing logic, reducing cloud program computation overhead, and enhancing the cloud's ability to concurrently process device data. Furthermore, it can perform pre-compliance verification based on cloud-generated data message information, reducing platform operating costs, ensuring overall platform stability, and unifying equipment assessment standards. It also retrieves complete end-to-end data for comprehensive reasoning of compliant data, providing complete analytical dimensions and significantly improving the depth and accuracy of intelligent agent data analysis. This provides comprehensive and reliable data support for professional equipment assessment reports, improving the reliability and accuracy of generated assessment results and reports. The data to be analyzed is transmitted via a terminal hardware encrypted channel throughout the process, standardizing data interaction formats while ensuring the security of sensitive industrial data transmission such as equipment operating conditions and fault risks. This improves the security and reliability of data processing, thereby enhancing the accuracy and reliability of electromechanical equipment assessments and increasing the efficiency of safety assessments for electromechanical equipment.

[0127] Example 3 Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of another electromechanical equipment evaluation device based on large-scale intelligent agent collaboration disclosed in an embodiment of the present invention. Figure 3 As shown, the electromechanical equipment evaluation device based on large-scale model intelligent agent collaboration may include: Acquisition module 301 is used to acquire data from the target device; The transmission module 302 is used to transmit the target device data to a preset data buffer for storing the target device data; Reading module 303 is used to read target device data from a preset data buffer; Analysis module 304 is used to perform data analysis operations on target device data using a pre-determined model to calculate the intelligent agent and obtain data analysis results; The generation module 305 is used to generate the device early warning result of the target device corresponding to the target device data based on the data analysis results; Encryption module 306 is used to perform data encryption operations on data analysis results and equipment warning results to obtain encrypted data results; The generation module 305 is also used to generate the device evaluation result of the target device based on the data encryption result and the target device data; Extraction module 307 is used to extract evaluation feature data of equipment evaluation results; The generation module 305 is also used to send the evaluation feature data to the large model cloud platform; generate agent analysis data based on the large model cloud platform; and generate an equipment evaluation report for the target device based on the agent analysis data.

[0128] It is evident that implementation Figure 3The described device can collect target equipment data, transmit the target equipment data to a data buffer for storage, read the target equipment data and call the model to calculate the intelligent agent to perform data analysis operations to obtain data analysis results, generate equipment early warning results based on the data analysis results, perform data encryption operations on the data analysis results and equipment early warning results to obtain encrypted data results, generate equipment evaluation results based on the encrypted data results and target equipment data, extract evaluation feature data from the equipment evaluation results and send it to the large model cloud platform to generate intelligent agent analysis data, and generate equipment evaluation reports for the target equipment based on the intelligent agent analysis data. It can solve the risk of industrial equipment data tampering and leakage through end-to-end encrypted storage and transmission; the data analysis results and equipment early warning results are generated with unified hardware encryption. The data encryption process ensures that evaluation results are generated only after successful decryption and verification. This enhances the security and reliability of the data used in the evaluation process. Furthermore, the hierarchical collaboration of multi-level large-scale model agents significantly improves the accuracy and reliability of electromechanical equipment evaluation. The model computation agent performs multi-dimensional fault analysis based on complete equipment time-series data, outputting standardized data analysis results. This provides high-quality evaluation feature data for the cloud agent, further improving the accuracy of cloud-based root cause analysis and report generation. It can also automatically schedule multi-level cloud agents to complete analysis, early warning, evaluation, and report generation, thereby improving the intelligence and efficiency of evaluation results and report generation. This contributes to the accuracy and reliability of electromechanical equipment evaluation and the efficiency of safety evaluation of electromechanical equipment.

[0129] In an optional embodiment, such as Figure 4 As shown, the device also includes: The determination module 308 is used to determine the abnormal parameters of the target equipment based on the equipment evaluation report. The generation module 305 is also used to generate equipment alarm information based on equipment abnormal parameters; and to generate alarm control parameters corresponding to the alarm components based on the equipment alarm information and the pre-determined alarm components. The control module 309 is used to control the alarm component to perform alarm operations that match the alarm control parameters and obtain alarm results; The acquisition module 301 is also used to obtain real-time environmental information corresponding to the target device based on the alarm results; The determination module 308 is also used to determine risk warning information of the current environment in which the target device is located based on real-time environmental information; The generation module 305 is also used to generate an isolated passive switch interlock signal based on the risk warning information. The isolated passive switch interlock signal is used to control the alarm component or target equipment to perform a shutdown operation.

[0130] It is evident that implementation Figure 4The described device can determine the abnormal parameters of the target device based on the equipment assessment report, generate equipment alarm information based on the abnormal parameters, generate alarm control parameters corresponding to the alarm components based on the equipment alarm information and pre-determined alarm components, and control the alarm components to perform alarm operations matching the alarm control parameters to obtain alarm results. Based on the alarm results, it obtains the real-time environmental information corresponding to the target device, determines the risk warning information of the current environment in which the target device is located based on the real-time environmental information, and generates isolated passive switch interlocking signals based on the risk warning information. It can use the standardized equipment assessment report generated by the cloud-based large-scale model intelligent agent as the alarm data source. The report integrates multi-dimensional information, rather than relying solely on a single simple threshold judgment, which can accurately filter out abnormal parameters and avoid the problems of ambiguity, a large number of meaningless false alarms, and missed alarms in traditional single-threshold alarms. It improves on-site operation and maintenance efficiency and can drive audible and visual alarm devices through hardware layering, making on-site risks intuitively perceptible and the response chain highly deterministic. On-site personnel can quickly identify the severity of risks, improving user intuitiveness and convenience. Moreover, the entire alarm process is traceable, facilitating subsequent operation and maintenance review and safety audit. It automatically collects real-time environmental information around the equipment and can simultaneously identify external environmental hazards, realizing integrated comprehensive judgment of equipment failure and on-site environmental hazards, covering all dimensions of equipment operation safety risks, which is conducive to improving the safety and reliability of equipment assessment. It can also execute shutdown operations based on passive interlock signals to maximize production and operation efficiency while ensuring the bottom line of safe production, improve the overall intelligent level of safety management of large electromechanical equipment clusters, and improve the accuracy and reliability of electromechanical equipment assessment, as well as the efficiency of electromechanical equipment safety assessment.

[0131] In another alternative embodiment, such as Figure 4 As shown, the generation module 305 sends the evaluation feature data to the large model cloud platform, and generates agent analysis data based on the large model cloud platform in the following specific ways: Based on the evaluation feature data, an evaluation data threshold matching the evaluation feature data is determined in a pre-defined judgment rule database, and a data comparison operation is performed between the evaluation feature data and the evaluation data threshold to obtain the data comparison results; Based on the data comparison results, the risk decision parameters corresponding to the data comparison results are determined. The risk decision parameters include at least the risk level of the target equipment, the risk recommendation measures, and the confidence score corresponding to the data comparison results. By sending risk decision parameters, data comparison results, and target device data to the big model cloud platform, the full-link data corresponding to the target device can be extracted through the big model cloud platform, and intelligent agent analysis data can be generated based on the full-link data. The end-to-end data includes at least the original time-series data corresponding to the target device, evaluation feature data, risk decision conclusions, device identifier, and evaluation timestamp.

[0132] It is evident that implementation Figure 4 The described device can determine the assessment data threshold matching the assessment feature data in a pre-defined judgment rule database based on the assessment feature data, and perform a data comparison operation between the assessment feature data and the assessment data threshold to obtain the data comparison result. Based on the data comparison result, it determines the risk decision parameters corresponding to the data comparison result, and sends the risk decision parameters, data comparison result, and target device data to a large model cloud platform. The large model cloud platform can extract the full-link data corresponding to the target device, and generate intelligent agent analysis data based on the full-link data. It can match the corresponding device-specific assessment data threshold based on a unified and professional judgment rule database, which helps to improve the reliability and accuracy of the assessment data threshold and prevent the judgment bias of the assessment threshold. It can also simultaneously generate three types of core risk decision parameters based on the data comparison result: risk level, risk recommendation measures, and confidence score, forming a complete set of parameters that can be used by the terminal hardware. The structured risk dataset is parsed, and the terminal synchronously uploads three core data types: risk decision parameters, data comparison results, and original target equipment data. The cloud platform can combine the real-time assessment data with the equipment's historical archived full-link data to conduct comprehensive analysis. The full-link data covers the original time-series sampling data, assessment characteristics, risk decision conclusions, unique equipment identifiers, and assessment timestamps, completely preserving all process information of the equipment from data collection to risk assessment. This enables full-process tracing and in-depth root cause analysis of equipment failures. It also enables secure interaction, with risk decisions, comparison results, and original equipment data transmitted in encrypted form and verified by hardware signatures throughout the process. While achieving accurate and intelligent assessment, it ensures that sensitive equipment operating data is not leaked or tampered with, balancing intelligent analysis capabilities with industrial data security requirements. This improves the security of data processing and transmission, which is conducive to improving the accuracy and reliability of electromechanical equipment assessment and the efficiency of safety assessment of electromechanical equipment.

[0133] In yet another alternative embodiment, such as Figure 4 As shown, the generation module 305 generates alarm control parameters corresponding to alarm components based on device alarm information and pre-determined alarm components in the following specific ways: Based on the device alarm information and the pre-determined alarm components, the structured instructions corresponding to the device alarm information are parsed into predefined hardware control commands through the application processor's fixed scheduler. The hardware control commands determine the level state of the corresponding pins through the GPIO controller, and generate alarm control parameters corresponding to the alarm components based on the level state.

[0134] It is evident that implementation Figure 4 The described device, based on device alarm information and pre-determined alarm components, uses a fixed scheduler in the application processor to parse the structured instructions corresponding to the device alarm information into predefined hardware control commands. The hardware control commands are then used by a GPIO controller to determine the voltage levels of corresponding pins, and alarm control parameters corresponding to the alarm components are generated based on these voltage levels. This hardware-based fixed scheduler parses alarm commands, eliminating software dependency and ensuring a stable, reliable, and tamper-proof alarm control process. This improves the security and reliability of data processing and alarm handling. Furthermore, the device achieves standardized parsing of structured alarm information without requiring manual configuration of conversion scripts, thus enabling alarm command decryption. Full automation reduces operational errors and omissions caused by manual configuration, improving the accuracy and reliability of alarms, as well as their security, intelligence, and efficiency. It also converts structured business instructions into standardized hardware commands, enhancing compatibility and scalability, significantly reducing terminal hardware modification costs and development cycles. Furthermore, it generates standardized alarm control parameters based on hardware levels, ensuring fully quantified, traceable, and easily auditable alarm actions. This provides complete hardware-level evidence for equipment fault tracing and safety production audits, improving safety performance and ultimately enhancing the accuracy and reliability of electromechanical equipment assessments, as well as the efficiency of safety assessments.

[0135] In yet another alternative embodiment, such as Figure 4 As shown, the specific method by which the transmission module 302 transmits target device data to a preset data buffer for storing target device data includes: The target device data is transmitted to a preset data buffer through a unique and fixed transmission channel. The original data inside the data buffer is retrieved. Based on the original data inside the data buffer, hardware-level encryption and digital signature operations are performed on the target device data to obtain encrypted device data. Data processing operations are performed on the encrypted data of the device to obtain the device-stored data, and the device-stored data is stored in the data buffer; the data processing operations include data cleaning operations, data removal operations, and data feature extraction operations.

[0136] It is evident that implementation Figure 4The described device can transmit target device data to a preset data buffer through a unique, fixed transmission channel. It retrieves the original data from the data buffer and, based on this original data, performs hardware-level encryption and digital signature operations on the target device data to obtain encrypted device data. Data processing operations are then performed on the encrypted device data to obtain the device's stored data, which is stored in the data buffer. This device can transmit target device data through a unique, fixed transmission channel, preventing malicious data injection attacks at the hardware level and ensuring that all data entering the terminal is genuine, original signal collected by the device. Compared to multi-channel universal gateways, this significantly improves the authenticity and traceability of original industrial operating data. The encryption mechanism enables fully cached encrypted data storage, relying on… The buffer generates verification hashes from historical raw data, binding and verifying real-time collected data with historical archived data to ensure data integrity. It also improves the overall efficiency of end-to-cloud collaboration through integrated three-layer preprocessing: cleaning, removal, and feature extraction. This reduces the cache space occupied by invalid data, increases the storage utilization of the buffer, saves industrial network communication costs, and shortens the cloud data parsing time. It improves the accuracy of equipment fault diagnosis and risk assessment from the data source level. Layered processing generates standardized equipment storage data with a unified cache data structure, facilitating subsequent full-process calling and parsing. This ensures the stable parallel operation of multiple rounds of data collection tasks, thereby improving the accuracy and reliability of electromechanical equipment evaluation and the efficiency of safety assessment of electromechanical equipment.

[0137] In yet another alternative embodiment, such as Figure 4 As shown, the specific methods by which the generation module 305 generates the device evaluation result of the target device based on the data encryption result and the target device data include: Hardware decryption and signature validity verification are uniformly performed on the encrypted data results to obtain the data verification results, and it is determined whether the data verification results meet the preset verification conditions. When it is determined that the data verification result meets the preset verification conditions, the plaintext feature data corresponding to the data encryption result and the target device data is determined, and the plaintext feature data is stored in an independent cache partition; Read all independent storage data stored in the independent cache partition, generate multi-dimensional evaluation results based on the independent storage data, and generate device evaluation results for the target device based on the multi-dimensional evaluation results; The multi-dimensional evaluation results include at least the target equipment's health score, fault type, fault confidence, remaining lifespan, and fault trend.

[0138] It is evident that implementation Figure 4The described device can uniformly perform hardware decryption and signature validity verification on the encrypted data results, obtain the data verification results, and determine whether the data verification results meet the preset verification conditions. When the data verification results are determined to meet the preset verification conditions, the device determines the plaintext feature data corresponding to the encrypted data and the target device data, and stores the plaintext feature data in an independent cache partition. It reads all independently stored data in the independent cache partition, generates multi-dimensional evaluation results based on the independently stored data, and generates a device evaluation result for the target device based on the multi-dimensional evaluation results. This dual verification through hardware decryption and signature validity verification ensures the authenticity and reliability of the evaluation results, and improves the security and reliability of data processing. Verified plaintext is stored in the independent cache partition. Hardware-isolated storage achieves layered data protection, avoids data cross-contamination, and ensures the data source for multi-dimensional evaluation calculations. Stability is ensured by an independent partition dedicated to storing trusted plaintext features, distinguishing various business data within the cache, eliminating the need to filter irrelevant and redundant data when reading full feature data, improving the reading speed of multi-dimensional evaluation calculations, shortening the latency of equipment evaluation result generation, and enhancing the intelligence and efficiency of data processing and the generation of evaluation results and reports. By integrating multi-dimensional indicators to generate unified equipment evaluation results, it provides a unified and standardized data source for backend alarms, cloud analysis, and report generation, eliminating the need for repeated data format conversions, simplifying the entire process of data interaction, reducing software and hardware development costs, and enabling dual support for local real-time control and long-term cloud-based operation and maintenance analysis. It also enables refined and digital evaluation of equipment status, improving the level of intelligent safety management of electromechanical equipment, thereby improving the accuracy and reliability of electromechanical equipment evaluation and increasing the efficiency of safety evaluation of electromechanical equipment.

[0139] In yet another alternative embodiment, such as Figure 4 As shown, the generation module 305 sends risk decision parameters, data comparison results, and target device data to the large model cloud platform to extract the full-link data corresponding to the target device. The specific methods for generating intelligent agent analysis data based on the full-link data include: Based on risk decision parameters, data comparison results and target device data, data to be analyzed is generated, and the data to be analyzed is sent to the big model cloud platform. The big model cloud platform then generates the corresponding data message information of the data to be analyzed. Determine whether the data message information is used to indicate that the data to be analyzed meets the preset data analysis conditions; When it is determined that the data message information indicates that the data to be analyzed meets the preset data analysis conditions, the full-link data corresponding to the target device is extracted through the large model cloud platform, and the full-link data is packaged to obtain the link data packaging result. Based on the link data packaging result, intelligent agent analysis data is generated. When it is determined that the data message information is used to indicate that the data to be analyzed does not meet the preset data analysis conditions, a judgment log information corresponding to the data message information is generated based on the data to be analyzed, the data message information, and the preset data analysis conditions. An exception prompt information is generated based on the judgment log information, and an adjustment operation is performed on the data message information according to the exception prompt information to obtain the adjusted data message information. Then, the operation of determining whether the data message information is used to indicate that the data to be analyzed meets the preset data analysis conditions is triggered again.

[0140] It is evident that implementation Figure 4 The described device can generate data to be analyzed based on risk decision parameters, data comparison results, and target device data. It then sends this data to a large-scale model cloud platform and generates corresponding data message information. The device determines whether the data message information indicates that the data meets preset data analysis conditions. If so, it extracts the full-link data corresponding to the target device through the large-scale model cloud platform, performs data packaging on the full-link data to obtain the packaged link data, and generates agent analysis data based on the packaged link data. If not, it generates judgment log information corresponding to the data message information based on the data to be analyzed, the data message information, and the preset data analysis conditions. It also generates anomaly prompts based on the judgment log information and adjusts the data message information according to the anomaly prompts to obtain the adjusted data message information. Finally, it re-triggers the operation to determine whether the data message information indicates that the data meets the preset data analysis conditions, enabling the analysis of multiple core... Data fusion and encapsulation result in data to be analyzed. End-to-cloud interaction data is standardized and unified, reducing transmission redundancy and parsing costs. No additional matching indexes are required, simplifying cloud data parsing logic, reducing cloud program computation overhead, and enhancing the cloud's ability to concurrently process device data. Furthermore, it enables pre-compliance verification of cloud-generated data messages, reducing platform operating costs and ensuring overall platform stability. Equipment assessment standards are unified, and comprehensive reasoning based on complete end-to-end data retrieval for compliant data provides complete analytical dimensions, significantly improving the depth and accuracy of intelligent agent data analysis. This provides comprehensive and reliable data support for professional equipment assessment reports, enhancing the reliability and accuracy of generated assessment results and reports. The data to be analyzed is transmitted via a terminal hardware encrypted channel throughout the process, standardizing data interaction formats while ensuring the security of sensitive industrial data transmission such as equipment operating conditions and fault risks. This improves the security and reliability of data processing, thereby enhancing the accuracy and reliability of electromechanical equipment assessments and increasing the efficiency of safety assessments for electromechanical equipment.

[0141] Example 4 Please see Figure 5 , Figure 5This is a schematic diagram of the structure of another electromechanical equipment evaluation device based on large-scale intelligent agent collaboration disclosed in an embodiment of the present invention. Figure 5 As shown, the electromechanical equipment evaluation device based on large-scale model intelligent agent collaboration may include: Memory 401 storing executable program code; Processor 402 coupled to memory 401; The processor 402 calls the executable program code stored in the memory 401 to execute some or all of the steps in any of the electromechanical equipment evaluation methods based on large model intelligent agent collaboration in Embodiment 1 of the present invention.

[0142] Example 5 This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to execute some or all of the steps in any of the electromechanical equipment evaluation methods based on large model intelligent agent collaboration disclosed in Embodiment 1 of this invention.

[0143] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0144] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0145] Finally, it should be noted that the above embodiments are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating electromechanical equipment based on large-scale intelligent agent collaboration, characterized in that, The method includes: Collect target device data and transfer the target device data to a preset data buffer for storage; The target device data is read from the preset data buffer, and a pre-determined model computing agent is invoked to perform data analysis operations on the target device data to obtain data analysis results; Based on the data analysis results, generate the device early warning result for the target device corresponding to the target device data; The data analysis results and the device warning results are encrypted to obtain encrypted data results. Based on the encrypted data results and the target device data, the device evaluation results of the target device are generated. The evaluation feature data of the device evaluation results is extracted, the evaluation feature data is sent to the large model cloud platform, and intelligent agent analysis data is generated based on the large model cloud platform. Finally, the device evaluation report of the target device is generated based on the intelligent agent analysis data.

2. The electromechanical equipment evaluation method based on large-scale intelligent agent collaboration according to claim 1, characterized in that, The method further includes: Based on the equipment evaluation report, determine the abnormal parameters of the target equipment, and generate equipment alarm information based on the abnormal parameters. Based on the device alarm information and the pre-determined alarm components, alarm control parameters corresponding to the alarm components are generated, and the alarm components are controlled to perform alarm operations that match the alarm control parameters to obtain alarm results. Based on the alarm result, real-time environmental information corresponding to the target device is obtained. Based on the real-time environmental information, risk warning information of the current environment where the target device is located is determined. Based on the risk warning information, an isolated passive switch interlock signal is generated. The isolated passive switch interlock signal is used to control the alarm component or the target device to perform a shutdown operation.

3. The electromechanical equipment evaluation method based on large-scale intelligent agent collaboration according to claim 1 or 2, characterized in that, The step of sending the evaluation feature data to the large model cloud platform and generating agent analysis data based on the large model cloud platform includes: Based on the evaluation feature data, an evaluation data threshold matching the evaluation feature data is determined in a pre-determined judgment rule database, and a data comparison operation is performed between the evaluation feature data and the evaluation data threshold to obtain the data comparison result; Based on the data comparison results, risk decision parameters corresponding to the data comparison results are determined, wherein the risk decision parameters include at least the risk level of the target device, risk recommendation measures, and the confidence score corresponding to the data comparison results; The risk decision parameters, the data comparison results, and the target device data are sent to the large model cloud platform to extract the full-link data corresponding to the target device through the large model cloud platform, and intelligent agent analysis data is generated based on the full-link data. The end-to-end data includes at least the original time-series data, evaluation feature data, risk decision conclusions, device identifier, and evaluation timestamp corresponding to the target device.

4. The electromechanical equipment evaluation method based on large-scale intelligent agent collaboration according to claim 2, characterized in that, The step of generating alarm control parameters corresponding to the alarm components based on the device alarm information and the pre-determined alarm components includes: Based on the device alarm information and the pre-determined alarm components, the structured instructions corresponding to the device alarm information are parsed into predefined hardware control commands through the application processor's fixed scheduler. The hardware control commands are used to determine the level state of the corresponding pins through the GPIO controller, and alarm control parameters corresponding to the alarm component are generated based on the level state.

5. The electromechanical equipment evaluation method based on large-scale intelligent agent collaboration according to claim 1, characterized in that, The step of transmitting the target device data to a preset data buffer for storing the target device data includes: The target device data is transmitted to a preset data buffer through a unique fixed transmission channel. The original data inside the data buffer is retrieved. Based on the original data inside the data buffer, hardware-level encryption and digital signature operations are performed on the target device data to obtain encrypted device data. Data processing operations are performed on the encrypted data of the device to obtain the device stored data, and the device stored data is stored in the data buffer; wherein, the data processing operations include data cleaning operations, data removal operations, and data feature extraction operations.

6. The electromechanical equipment evaluation method based on large-scale intelligent agent collaboration according to claim 1, characterized in that, The step of generating a device evaluation result for the target device based on the data encryption result and the target device data includes: Hardware decryption and signature validity verification are uniformly performed on the encrypted data to obtain the data verification result, and it is determined whether the data verification result meets the preset verification conditions. When it is determined that the data verification result meets the preset verification conditions, the plaintext feature data corresponding to the data encryption result and the target device data is determined, and the plaintext feature data is stored in an independent cache partition; Read all independent storage data stored in the independent cache partition, generate multi-dimensional evaluation results based on the independent storage data, and generate device evaluation results for the target device based on the multi-dimensional evaluation results; The multi-dimensional evaluation results include at least the target device's health score, fault type, fault confidence, remaining lifespan, and fault trend.

7. The electromechanical equipment evaluation method based on large-scale intelligent agent collaboration according to claim 3, characterized in that, The step of sending the risk decision parameters, the data comparison results, and the target device data to the large model cloud platform, so as to extract the full-link data corresponding to the target device through the large model cloud platform, and generate intelligent agent analysis data based on the full-link data, includes: Based on the risk decision parameters, the data comparison results and the target device data, data to be analyzed is generated, the data to be analyzed is sent to the large model cloud platform, and the data message information corresponding to the data to be analyzed is generated through the large model cloud platform. Determine whether the data message information is used to indicate that the data to be analyzed meets preset data analysis conditions; When it is determined that the data message information indicates that the data to be analyzed meets the preset data analysis conditions, the full-link data corresponding to the target device is extracted through the large model cloud platform, and a data packaging operation is performed on the full-link data to obtain the link data packaging result. Based on the link data packaging result, intelligent agent analysis data is generated. When it is determined that the data packet information indicates that the data to be analyzed does not meet the preset data analysis conditions, a judgment log information corresponding to the data packet information is generated based on the data to be analyzed, the data packet information, and the preset data analysis conditions. An exception prompt information is generated based on the judgment log information, and an adjustment operation is performed on the data packet information according to the exception prompt information to obtain the adjusted data packet information. The operation of determining whether the data packet information is used to indicate that the data to be analyzed meets the preset data analysis conditions is then triggered again.

8. An electromechanical equipment evaluation device based on large-scale intelligent agent collaboration, characterized in that, The device includes: The acquisition module is used to acquire data from the target device; The transmission module is used to transmit the target device data to a preset data buffer for storing the target device data; A reading module is used to read data from the target device in a preset data buffer; The analysis module is used to calculate the intelligent agent to perform data analysis operations on the target device data using a pre-determined model, and obtain data analysis results; The generation module is used to generate a device warning result for the target device corresponding to the target device data based on the data analysis results. The encryption module is used to perform data encryption operations on the data analysis results and the device warning results to obtain data encryption results; The generation module is also used to generate a device evaluation result for the target device based on the data encryption result and the target device data; An extraction module is used to extract the evaluation feature data of the equipment evaluation results; The generation module is also used to send the evaluation feature data to the large model cloud platform; generate agent analysis data based on the large model cloud platform; and generate an equipment evaluation report for the target device based on the agent analysis data.

9. An electromechanical equipment evaluation device based on large-scale intelligent agent collaboration, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the electromechanical equipment evaluation method based on large model intelligent agent cooperation as described in any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the electromechanical equipment evaluation method based on large-scale intelligent agent collaboration as described in any one of claims 1-7.

Images