Maintenance management system and method for intelligent office equipment

Abstract

The invention discloses a maintenance management system and method for intelligent office equipment, and relates to the technical field of maintenance management of the intelligent office equipment, and the system comprises an equipment fault diagnosis grading module which is used for installing a multi-dimensional sensor on the intelligent office equipment, collecting the information of the multi-dimensional sensor in real time, carrying out the analysis and judgment of a fault type through combining historical fault data, and carrying out the maintenance management of the intelligent office equipment. A fault severity index is dynamically calculated, a maintenance priority is automatically divided, and a maintenance result is output and transmitted to a maintenance order generation module; the maintenance priority is automatically divided by dynamically calculating the fault severity index, the fault diagnosis accuracy is improved, the GIS positioning technology and the dynamic routing optimization algorithm are adopted, appropriate maintenance personnel are determined, the optimal maintenance work order is generated, intelligent order dispatching is achieved, and the maintenance efficiency is improved. Maintenance personnel can respond quickly and perform maintenance operation through dynamic authority management, so that the maintenance time is shortened, the maintenance efficiency is improved, the downtime of equipment is shortened, and the business loss caused by equipment faults is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent office equipment maintenance management, and particularly to a maintenance management system and method for intelligent office equipment. Background Art

[0002] With the continuous improvement of the intelligence level of modern office environments, intelligent office equipment such as computers, printers, conference systems, and intelligent whiteboards are increasingly widely used in large and small enterprises. However, with the increase in the number and complexity of intelligent office equipment, the maintenance and management of equipment have become an important challenge.

[0003] The existing maintenance management systems have the following defects:

[0004] Most of the existing fault diagnoses rely on the experience and professional knowledge of maintenance personnel. When facing a large number of equipment failures, it is difficult for maintenance personnel to quickly and accurately judge the severity of the failures, resulting in unreasonable priority allocation, waste of maintenance resources, delay in maintenance time, and affecting the normal operation of the equipment; when maintenance personnel operate without appropriate permissions, it may lead to equipment configuration errors, software conflicts, or security vulnerabilities, affecting the normal operation of the equipment and data security; the performance calibration after equipment maintenance usually depends on the skills and experience of engineers, and there may be significant differences in accuracy, resulting in unsatisfactory performance recovery after equipment maintenance and a high probability of re-failure; and the equipment usually uses passive alarm control, lacking continuous fault prediction and maintenance of Predictive Health Management (PHM), which easily leads to high maintenance costs, short equipment service life, and affects the operation efficiency of enterprises.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and thus it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The object of the present invention is to provide a maintenance management system and method for intelligent office equipment. The present invention uses multi-dimensional sensors, GIS positioning technology, dynamic routing optimization algorithm, dynamically grants temporary permissions to maintenance personnel, sets a time-limited prohibition mechanism, AI automatic execution of minimum deviation calibration technology, blockchain technology, and uses an LSTM-PHM hybrid model to solve the problems in the above background art.

[0007] To achieve the above object, the present invention provides the following technical solution: A maintenance management system for intelligent office equipment, including an equipment fault diagnosis and grading module: Install multi-dimensional sensors on the intelligent office equipment, collect multi-dimensional sensor information in real time, analyze and judge the fault type by combining historical fault data, dynamically calculate the fault severity index, automatically divide the maintenance priority level, and output the maintenance result to be transmitted to the maintenance order generation module;

[0008] Maintenance order generation module: Receive the maintenance result in the equipment fault diagnosis and grading module, combine the GIS positioning technology and the dynamic routing optimization algorithm to generate a maintenance work order and intelligently dispatch it, and synchronize the information of the maintenance work order to the blockchain maintenance file update module;

[0009] Maintenance mode adaptive permission module: Dynamically grant temporary permissions to maintenance personnel based on the role-based access control model, and set a time-limited permission automatic prohibition mechanism. Maintenance personnel need to input the work order number and the maintenance personnel ID to obtain the permission to access the equipment debugging and maintenance, and record the space-time label and operator identity of each permission change in the blockchain maintenance file update module. Based on the performance baseline data of the equipment PHM supervision module, restrict the maintenance personnel from making maintenance modifications to the equipment performance parameters that do not exceed the reasonable range;

[0010] Post-maintenance equipment performance accuracy deviation calibration module: Collect performance accuracy data of the maintained equipment, compare it with the original factory standard model to detect accuracy deviation, use AI to automatically execute the minimum deviation calibration technology to obtain the calibration result, and transmit the calibration result back to the blockchain maintenance file update module to update the equipment health file;

[0011] Blockchain maintenance file update module: Use blockchain technology to record the entire maintenance process data in each module, and provide the historical entire maintenance process data to the equipment PHM supervision module;

[0012] Equipment PHM supervision module: Integrate the real-time multi-dimensional sensor data in the equipment fault diagnosis and grading module and the historical entire maintenance process data in the blockchain maintenance file update module, and based on the LSTM-PHM hybrid model, predict the remaining useful life RUL of the maintained equipment and evaluate and analyze the probability of fault recurrence PFR.

[0013] Optionally, the calculation process of the fault severity index is as follows:

[0014] Deploy and install initialized multi-dimensional sensors on the intelligent office equipment. When a sudden fault response occurs during the operation of the intelligent office equipment, collect multi-dimensional sensor information in real time, including the equipment operation status data, environmental data, and operation logs, which are respectively calibrated as Osd, Evd, and Old;

[0015] Filter, detect outliers, and normalize the collected multi-dimensional sensor information to obtain standardized data;

[0016] The standardized data is transmitted with low latency using the MQTT protocol and stored in the blockchain historical data pool of the device;

[0017] Classify the current faults of the device based on the multi-dimensional sensor information and historical fault data, and use a causal inference engine to distinguish and determine the fault types of hardware faults, software faults, and environmental faults, which are respectively labeled as Hwf, Swf, and Evf;

[0018] According to the degree of influence of each fault type on the overall performance of the device, as well as the instantaneous Δt change and trend during sudden faults, introduce the concept of time window t, calculate the trend change amount within the time window t, and then dynamically adjust the weight values of the multi-dimensional sensor information;

[0019] The calculation formula for the fault severity index is FSI = SI × (α·Hwf + β·Swf + γ·Evf),

[0020]

[0021] and α + β + γ = 1, ω1 + ω2 + ω3 = 1,

[0022] In the formula, FSI represents the fault severity index,

[0023] SI represents the dynamic regulation index,

[0024] α, β, and γ respectively represent the initial weight values corresponding to the hardware fault Hwf, software fault Swf, and environmental fault Evf,

[0025] [[ID=z30]]ω1, ω2, and ω3 respectively represent the weights of the contribution of the operating status data Osd, environmental data Evd, and operation log Old in the multi-dimensional sensor information to the fault,

[0026] Osd(t), Evd(t), and Old(t) respectively represent the sensor readings of the corresponding operating status data Osd, environmental data Evd, and operation log Old at the current time t,

[0027] N Osd 、N Evd 、N Old respectively represent the historical maximum instantaneous change values corresponding to the operating status data Osd, environmental data Evd, and operation log Old,

[0028] M Osd 、M Evd 、M OldThey are respectively represented as the historical maximum trend change slopes corresponding to the operation status data Osd, environmental data Evd, and operation log Old.

[0029] δ is represented as the adjustment coefficient of the trend term, and δ ∈ (0, 1).

[0030] l(Osd, t), l(Evd, t), and l(Old, t) are respectively represented as the linear regression slopes of the operation status data Osd, environmental data Evd, and operation log Old within the time window t.

[0031] They are respectively represented as the instantaneous change terms corresponding to the operation status data Osd, environmental data Evd, and operation log Old.

[0032] They are respectively represented as the trend change terms corresponding to the operation status data Osd, environmental data Evd, and operation log Old.

[0033] According to the calculated fault severity index FSI, the maintenance priority is automatically divided. When FSI > 0.7, it is determined as high priority and urgent handling; when 0.4 < FSI ≤ 0.7, it is determined as medium priority and important handling; when FSI ≤ 0.4, it is determined as low priority and routine handling.

[0034] Optionally, the steps for generating the maintenance work order are as follows:

[0035] Match and analyze the candidate maintenance personnel who meet the skill certification requirements according to the maintenance results, where the maintenance results include the fault type, severity index, and maintenance priority.

[0036] Obtain the precise geographical location of the faulty device through GIS, and extract the real-time locations, skill qualifications, and current task loads of all available maintenance personnel from the maintenance database, and screen and retain the candidate maintenance personnel who meet the skill certification requirements. Among them, the geographical location of the faulty device located by GIS is L0 = (lat0, lon0), L0 represents the geographical location of the faulty device, and lat0 and lon0 respectively represent the latitude and longitude corresponding to the faulty device; the geographical location of the candidate maintenance personnel is L i =(lat i , lon i ), i = 1, 2, 3,..., n, representing the number of candidate maintenance personnel.

[0037] Use the dynamic routing optimization algorithm to make an intelligent decision and assign dispatch tasks to generate a maintenance work order. Among them, the dynamic routing optimization algorithm defines the total scheduling cost function as a multi-objective optimization problem, and the formula is

[0038] And

[0039] Wherein, Y represents the maintenance personnel selected to minimize the total scheduling cost in the dynamic routing optimization algorithm,

[0040] θ, μ, respectively represent the weight coefficients for dividing the priorities of the corresponding distance D(L0, L i ), time T(L0, L i ), and failure severity index (1 - FSI),

[0041] D(L0, L i ) represents the geographical distance between the geographical location L0 of the faulty device and the geographical location L of the maintenance personnel, i

[0042] T(L_{0}, L i ) represents the estimated arrival time of the maintenance personnel from their own geographical location L i to the geographical location L0 of the faulty device,

[0043] R represents the radius of the earth, and R = 6371 km,

[0044] represents the average vehicle speed when the maintenance personnel travel to the geographical location of the faulty device,

[0045] τ represents the traffic congestion coefficient;

[0046] The generation of the maintenance work order is simultaneously and synchronously updated to the blockchain maintenance file update module. Among them, the information content of the maintenance work order includes the work order number, fault type and severity index, equipment location, matching maintenance personnel ID, skill qualification, and estimated arrival time.

[0047] Optionally, the setting steps of the time - limited permission automatic prohibition mechanism are as follows:

[0048] Based on the role - based access control model, permission application and verification are carried out. The maintenance personnel input the work order number and maintenance personnel ID information through the terminal to request permission to access the debugging and maintenance of the faulty device;

[0049] The system reads the maintenance work order in the blockchain maintenance file update module for information comparison, verifies the work order number and maintenance personnel ID information of the maintenance personnel. After verification, the system dynamically grants the maintenance personnel temporary access permission, otherwise it does not grant;

[0050] According to the fault type of the equipment and the PHM performance baseline data, the operation permissions required to access the equipment and the permission to modify the equipment performance parameters are restricted, and a timer is used for permission time - limitation;

[0051] ​When the time-limited permission times out, the system immediately terminates the current operation permission and the device performance parameter modification permission, locks the device debugging interface, realizes the forced recovery of permissions, and if the device fault has not been resolved after the permission recovery, the maintenance priority is automatically increased, triggering a new work order dispatch process;

[0052] Use the blockchain maintenance file update module to record the maintenance event status, including repaired after timeout, not repaired after timeout, and actively completed maintenance status within the time limit, the maintenance personnel ID and the timestamp.

[0053] Optionally, the steps for obtaining the deviation of the performance accuracy data are as follows:

[0054] Execute the standard test procedure of the device on the repaired device, and use multi-dimensional sensors to collect the performance data of the repaired device under the operating state, calibrated as X j , where j represents the number of performance types of the intelligent office device;

[0055] For the performance data X j Perform filtering, denoising and normalization processing to obtain the normalized performance accuracy data with the same dimension as the factory standard model. Among them, the formula for the normalized performance accuracy data is In the formula, represents the performance data X j The normalized performance accuracy data after standardization processing, represents the benchmark mean value of the factory device standard performance parameters stored in the blockchain, represents the standard deviation of the factory device standard performance parameters stored in the blockchain;

[0056] Send the normalized performance accuracy data to the benchmark value of the device factory standard performance parameters stored in the blockchain, calibrated as X j0 , and use the AI algorithm to analyze the performance data to identify the performance deviation ΔX j , where the AI algorithm formula for the performance deviation is In the formula, ΔX j represents multi-dimensional anomaly detection, represents the multi-dimensional vector of the normalized performance accuracy data, represents the benchmark mean vector, T represents the transpose, represents the inverse matrix of the covariance matrix;

[0057] When is greater than the chi-square distribution threshold, trigger the calibration process.

[0058] Optionally, the process for the AI to automatically perform minimum deviation calibration is as follows:

[0059] Send calibration instructions through the programmable logic controller (PLC) in the system, establish a reinforcement learning (RL) dynamic optimization model, and perform minimization deviation calibration calculations based on the performance deviation ΔX j Perform minimization deviation calibration calculations;

[0060] Use the normalized performance accuracy data As the state variable, reflecting the deviation distribution between the current repaired equipment and the factory equipment benchmark, and define the adjustable equipment performance control parameter as a continuous or discrete action, calibrated as z j ;

[0061] Design a reward function to simultaneously minimize the performance deviation and the adjustment amplitude of the control parameter. Then the expression of the reward function In the formula, Represents the reward function for minimizing the deviation of equipment performance, ΔX j Represents the performance deviation value, Represents the regularization coefficient, ||z j || 2 Represents the performance control parameter penalty term;

[0062] Input the performance deviation ΔX j And the reward function to train the RL dynamic optimization model, and deploy the trained model to the edge computing unit of the equipment. Input the real-time collected performance data X j Into the model for minimizing deviation optimization, and output the calibration result through RL decision-making;

[0063] Among them, optimize the parameters of the RL dynamic optimization model through gradient ascent.

[0064] Optionally, the steps for constructing the blockchain maintenance file are as follows:

[0065] Collect the full-process maintenance data of the faulty equipment. The full-process maintenance data includes the maintenance results recorded in the equipment fault diagnosis and grading module, the maintenance work orders in the maintenance order generation module, the time-limited permission automatic prohibition mechanism in the maintenance mode adaptive permission module, and the calibration results in the post-maintenance equipment performance accuracy deviation calibration module;

[0066] Perform data standardization processing on the full-process maintenance data to generate JSON structured data;

[0067] Use data hashing technology and digital signature technology for the JSON structured data, and package the data hash, signature, and timestamp as transaction data;

[0068] Use the consensus mechanism of the blockchain to automatically verify whether the transaction data is compliant. And the transactions are packaged into blocks in chronological order to form a chain structure. Only when it is judged to be compliant is it allowed to be uploaded to the chain, otherwise not;

[0069] The new block is broadcast to all participating nodes through the gRPC protocol, and the maintenance file is stored in the blockchain state database in the form of key-value pairs;

[0070] The blockchain maintenance file update module shares the entire maintenance process data to the PHM device PHM supervision module.

[0071] Optionally, the prediction and evaluation logic of the LSTM-PHM hybrid model is as follows:

[0072] Extract the characteristic data of the multi-dimensional sensor data of the device in real time, and the historical entire maintenance process data is encoded using the BERT model;

[0073] Adopt the dynamic time warping DTW algorithm to unify the data with different sampling frequencies in the multi-dimensional sensor data to the same time axis, perform time alignment and then normalization processing. Through the multi-modal data fusion technology and the use of the gated attention mechanism, integrate the characteristic data of the real-time multi-dimensional sensor data with the historical entire maintenance process data;

[0074] Extract the characteristic time series data related to the remaining useful life and the probability of fault recurrence of the device from the characteristic data of the multi-dimensional sensor data. After inputting it into the LSTM-PHM hybrid model, use the long short-term memory network LSTM model to train and process the characteristic time series data, and capture the prediction results of the long-term degradation trend of the device;

[0075] Input the prediction results of the LSTM model into the Prognostics and Health Management PHM model for fusion, and then predict the probability of the device from the current moment to the recurrence of the next serious fault or the remaining time until the end of the service life,

[0076] Among them, the calculation formula of the probability of fault recurrence PFR is and In the formula, PFR represents the probability of fault recurrence evaluated by the LSTM-PHM hybrid model, A represents the mapping of the hidden state output by the LSTM model from the multi-dimensional sensor data and the historical entire maintenance process data to the risk value, h1 represents the characteristic of the current historical entire maintenance process data, and h0 represents the characteristic of the historical entire maintenance process data;

[0077] In the learning of the LSTM-PHM hybrid model, define the logarithmic Cauchy loss function of the remaining useful life RUL and the loss function for solving class imbalance of the probability of fault recurrence PFR respectively,

[0078] Among them, the expression of the logarithmic Cauchy loss function is In the formula, L RUL represents the logarithmic Cauchy loss function of the remaining useful life RUL, n and N respectively represent the nth sample and the total number of samples N, b represents the scale parameter of the Cauchy distribution, RUL nDenoted as the true RUL value of the nth sample of the device, Denoted as the RUL value predicted by the LSTM-PHM hybrid model for the device;

[0079] The expression for solving the class imbalance loss function is In the formula, L PFR Denoted as the class imbalance loss function for solving the probability of fault recurrence PFR, c is denoted as the class weight factor, p n Denoted as the probability of fault recurrence output by the LSTM-PHM hybrid model, and d is denoted as the focusing factor.

[0080] A maintenance management method for intelligent office equipment, including the following steps:

[0081] S1. Real-time fault monitoring and intelligent grading: Real-time collect multi-dimensional sensor data through the built-in sensors of intelligent office equipment, and combine historical fault data analysis to judge the fault type, dynamically calculate the fault severity index, and automatically divide the maintenance priority;

[0082] S2. Dynamic work order generation and intelligent dispatching: Receive the maintenance results, combine the GIS positioning technology and the dynamic routing optimization algorithm to determine the corresponding maintenance personnel, generate a maintenance work order and perform intelligent dispatching, and synchronize the maintenance work order information to the blockchain;

[0083] S3. Dynamic permission control and maintenance execution: Dynamically grant temporary permissions to maintenance personnel and set a time-limited permission automatic prohibition mechanism, restrict operation safety based on PHM performance baseline data, and record the spatio-temporal tags of maintenance operations on the blockchain;

[0084] S4. AI calibration and performance verification after equipment maintenance: Collect accuracy data to compare the detection accuracy deviation from the original factory standard model, and AI automatically executes the minimum deviation calibration technology to obtain the calibration result;

[0085] S5. Chain storage and evidence preservation of the entire maintenance process data: Use blockchain technology to write the entire maintenance process data to ensure that the data is tamper-proof and traceable;

[0086] S6. Remaining useful life RUL prediction and fault recurrence PFR assessment: Input the real-time multi-dimensional sensor data and the records of historical maintenance process data, use the LSTM-PHM hybrid model to predict the remaining useful life RUL of the equipment after maintenance, and evaluate and analyze the probability of fault recurrence PFR.

[0087] A computer device, including: a memory and a processor; the memory stores a computer program, and when the processor executes the computer program, it implements the steps of a maintenance management system for an intelligent office equipment according to any one of claims 1 to 7.

[0088] A computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of a maintenance management system for an intelligent office device according to any one of claims 1 to 7 are implemented.

[0089] In the above technical solution, the technical effects and advantages provided by the present invention:

[0090] The present invention automatically divides the maintenance priority by dynamically calculating the fault severity index, improves the accuracy of fault diagnosis, adopts GIS positioning technology and dynamic routing optimization algorithm to determine the appropriate maintenance personnel and generate the optimal maintenance work order while realizing intelligent dispatching. Dynamic permission management enables maintenance personnel to respond quickly and perform maintenance operations, shortening the maintenance time, improving the maintenance efficiency, reducing the equipment downtime, and reducing the business losses caused by equipment failures; and the blockchain technology ensures the immutability and traceability of the entire maintenance process data, making the data more secure and reliable, running through the entire maintenance process, and providing a credible basis for management decisions; and the LSTM-PHM hybrid model can make full use of real-time and historical data to accurately predict the remaining service life and fault recurrence probability of the equipment, providing a scientific basis for the preventive maintenance of the equipment, enabling the enterprise to arrange resources more reasonably, formulate maintenance plans, and improve the management level and operation efficiency of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0092] Figure 1 It is a schematic diagram of the modules of the maintenance management system for the intelligent office device of the present invention.

[0093] Figure 2 It is a flowchart of the maintenance management method for the intelligent office device of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0094] Now, the exemplary embodiments will be described more fully with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these exemplary embodiments are provided so that this disclosure will be more complete and comprehensive, and will fully convey the concept of the exemplary embodiments to those skilled in the art.

[0095] Embodiment 1

[0096] The present invention provides as Figure 1A maintenance management system for an intelligent office device as shown, including a device fault diagnosis and grading module: Install multi-dimensional sensors on the intelligent office device, collect multi-dimensional sensor information in real time, analyze and judge the fault type in combination with historical fault data, dynamically calculate the fault severity index, automatically divide the maintenance priority, and output the maintenance result to be transmitted to the maintenance order generation module;

[0097] Specifically, the calculation process of the fault severity index is as follows:

[0098] Deploy and install initialized multi-dimensional sensors on the intelligent office device. When a sudden fault response occurs during the operation of the intelligent office device, collect multi-dimensional sensor information in real time, including the device's operating status data, environmental data, and operation logs, which are respectively calibrated as Osd, Evd, and Old;

[0099] Perform filtering, outlier detection, and normalization preprocessing on the collected multi-dimensional sensor information to obtain standardized data;

[0100] The standardized data uses the MQTT protocol to achieve low-latency transmission and is stored in the device's blockchain historical data pool for subsequent fault type analysis and diagnosis;

[0101] According to the multi-dimensional sensor information and historical fault data, classify the current fault of the device, use a causal inference engine to distinguish and determine the fault types of hardware faults, software faults, and environmental faults, which are respectively calibrated as Hwf, Swf, and Evf, and define the initial weight values for the fault types according to historical fault data;

[0102] According to the degree of influence of each fault type on the overall performance of the device, as well as the instantaneous Δt change and trend during a sudden fault, introduce the concept of time window t, calculate the trend change amount within the time window t, and then dynamically adjust the weight values of the multi-dimensional sensor information;

[0103] The calculation formula for the fault severity index is FSI = SI × (α·Hwf + β·Swf + γ·Evf),

[0104]

[0105] And α + β + γ = 1, ω1 + ω2 + ω3 = 1,

[0106] In the formula, FSI represents the fault severity index,

[0107] SI represents the dynamic regulation index,

[0108] α, β, and γ respectively represent the initial weight values corresponding to the hardware fault Hwf, software fault Swf, and environmental fault Evf,

[0109] ω1, ω2, and ω3 respectively represent the weights of the contribution degrees of the operating status data Osd, environmental data Evd, and operation log Old in the multi-dimensional sensor information to the fault.

[0110] Osd(t), Evd(t), and Old(t) respectively represent the sensor readings of the corresponding operating status data Osd, environmental data Evd, and operation log Old at the current time t.

[0111] N Osd 、N Evd 、N Old respectively represent the historical maximum instantaneous change values of the corresponding operating status data Osd, environmental data Evd, and operation log Old.

[0112] M Osd 、M Evd 、M Old respectively represent the historical maximum trend change slopes of the corresponding operating status data Osd, environmental data Evd, and operation log Old.

[0113] δ represents the adjustment coefficient of the trend term, and δ ∈ (0, 1).

[0114] l(Osd, t), l(Evd, t), and l(Old, t) respectively represent the linear regression slopes of the operating status data Osd, environmental data Evd, and operation log Old within the time window t.

[0115] respectively represent the instantaneous change terms of the corresponding operating status data Osd, environmental data Evd, and operation log Old.

[0116] respectively represent the trend change terms of the corresponding operating status data Osd, environmental data Evd, and operation log Old.

[0117] According to the calculated fault severity index FSI, the maintenance priority is automatically divided. When FSI > 0.7, it is determined as high priority and dealt with urgently; when 0.4 < FSI ≤ 0.7, it is determined as medium priority and dealt with importantly; when FSI ≤ 0.4, it is determined as low priority and dealt with routinely.

[0118] It is further explained that by installing multi-dimensional sensors to collect multi-dimensional data of operating status, environment, and operation logs in real time, combined with filtering, outlier detection, and normalization preprocessing, the problems of incomplete coverage of traditional single data sources, noise interference, and data scale differences are solved, ensuring data comprehensiveness and calculation stability. The blockchain historical data pool storage makes the data tamper-proof and traceable, solving the trust problem that traditional data storage is easily disturbed or lost. The causal reasoning engine is used to distinguish the fault type for accurate classification, solving the misjudgment problem caused by reliance on manual experience. By fusing instantaneous change items and trend change items, combined with dynamic weights, the impact of faults on equipment performance is accurately quantified, solving the defect that traditional threshold methods cannot integrate multi-dimensional dynamic characteristics, and solving the problem of uneven distribution of maintenance resources or delayed response based on automatic division of maintenance priorities, providing a scientific basis for maintenance decisions.

[0119] Maintenance order generation module: Receives maintenance results from the equipment fault diagnosis and grading module, combines GIS positioning technology and dynamic routing optimization algorithms, generates maintenance work orders and intelligently dispatches them, and synchronizes the maintenance work order information to the blockchain maintenance file update module;

[0120] Specifically, the steps for generating a maintenance work order are as follows:

[0121] Matching and analyzing candidate maintenance personnel who meet the skill certification requirements for maintenance based on maintenance results, where maintenance results include fault type, severity index, and maintenance priority;

[0122] The precise geographic location of the faulty equipment is obtained through GIS, and the real-time location, skill qualifications, and current task load of all available maintenance personnel are extracted from the maintenance database. Candidate maintenance personnel who meet the skill certification are screened and retained. The geographic location of the faulty equipment located by GIS is L0 = (lat0, lon0), where L0 represents the geographic location of the faulty equipment, lat0, lon0 represent the latitude and longitude corresponding to the faulty equipment respectively; the geographic location of the candidate maintenance personnel is L i =(lat i ,lon i ), i = 1, 2, 3, ..., n, representing the number of candidate maintenance personnel;

[0123] The dynamic routing optimization algorithm is used to intelligently decide and allocate dispatch tasks and generate maintenance work orders. The dynamic routing optimization algorithm defines the total dispatch cost function as a multi-objective optimization problem, which is used to reduce cross-regional dispatch costs while avoiding dispatch errors caused by information islands. The formula is:

[0124] and

[0125] In the formula, Y represents the maintenance personnel selected to minimize the total scheduling cost in the dynamic routing optimization algorithm,

[0126] θ, μ, respectively represent the weight coefficients for prioritizing (1 - FSI) corresponding to the distance D(L0, L i ), time T(L0, L i ), and failure severity index,

[0127] D(L0, L i ) represents the geographical distance between the geographical location L0 of the faulty device and the geographical location L of the maintenance personnel, i

[0128] T(L0, L i ) represents the estimated arrival time of the maintenance personnel from their own geographical location L i to the geographical location L0 of the faulty device,

[0129] R represents the radius of the earth, and R = 6371 km,

[0130] represents the average driving speed when the maintenance personnel travel to the geographical location of the faulty device,

[0131] τ represents the traffic congestion coefficient;

[0132] The generation of the maintenance work order is simultaneously and synchronously updated to the blockchain maintenance file update module to ensure the transparency and immutability of the maintenance records. Among them, the information content of the maintenance work order includes the work order number, fault type and severity index, equipment location, matching maintenance personnel ID, skill qualifications, and estimated arrival time. The work order number is generated through a hash value, including a timestamp and an equipment ID.

[0133] ​Furthermore, by using GIS positioning technology to accurately obtain the longitude and latitude coordinates of the device and the maintenance personnel, and using the Haversine formula for distance calculation, it optimizes the scheduling deviation caused by the traditional dispatch relying on fuzzy location descriptions and the resource waste caused by unclear geographical locations in cross-regional dispatching. It is an important basis for eliminating manual description errors, improving the matching accuracy of the device location and the maintenance personnel location, and significantly reducing the ineffective dispatching cost. By using the dynamic routing optimization algorithm, taking the total dispatching cost function as a multi-objective optimization to screen the maintenance personnel, combined with dynamic factors such as traffic congestion, weather, and task priority, it generates the optimal path in real time, meeting the real-time adaptability and multi-objective optimization. Furthermore, it solves the problems that static path planning cannot handle sudden traffic jams or changes in task priority, and the low efficiency and easy conflicts of manual dispatching when multiple tasks are concurrent. That is, GIS positioning accurately matches the "nearest personnel", and dynamic routing plans the "optimal path". The combination of the two realizes the optimization of global resources, avoids cross-regional dispatching redundancy, improves efficiency, and makes the system have good robustness and scalability at the same time.

[0134] Maintenance mode adaptive permission module: Dynamically grants temporary permissions to maintenance personnel based on the role-based access control model, and sets a time-limited permission automatic prohibition mechanism. Maintenance personnel need to input the work order number and maintenance personnel ID to obtain the permission to access the device debugging and maintenance, and record the spatio-temporal tags and operator identities of each permission change in the blockchain maintenance file update module. Based on the performance baseline data of the device PHM supervision module, it restricts the maintenance personnel from making maintenance modifications to the device performance parameters that do not exceed the reasonable range.

[0135] Specifically, the setting steps of the time-limited permission automatic prohibition mechanism are as follows:

[0136] Based on the role-based access control model, permission application and verification are carried out. The maintenance personnel input the work order number and maintenance personnel ID information through the terminal to request the permission to access the debugging and maintenance of the faulty device.

[0137] The system reads the maintenance work orders in the blockchain maintenance file update module for information comparison, verifies the work order number and maintenance personnel ID information of the maintenance personnel. After verification, the system dynamically grants the maintenance personnel temporary access permissions, otherwise it does not grant, to ensure that the current maintenance personnel meet the requirements of the intelligent dispatch in the maintenance work order to screen out the maintenance personnel with the ability to handle the current fault type.

[0138] According to the fault type of the device and the PHM performance baseline data, it restricts the operation permissions required to access the device and the modification permissions of the device performance parameters, and uses a timer for permission time limit. For example, the time limit is 2 hours. If the maintenance task is not completed within the time-limited permission, the system automatically withdraws the permission and forces the maintenance work to exit, to prevent the long-term downtime of the device fault from affecting the hardware performance and precision deviation.

[0139] When the time-limited permission expires, the system immediately terminates the current operation permission and the permission to modify device performance parameters, locks the device debugging interface, realizes the forced recovery of permissions, and if the device failure is still not resolved after the permission recovery, the maintenance priority is automatically increased, triggering a new work order dispatching process;

[0140] Use the blockchain maintenance record update module to record the status of maintenance events, including repaired after timeout, not repaired after timeout, and actively completed maintenance status within the time limit, the maintenance personnel ID and timestamp, as well as the changed space-time label and operator identity record in the case of not repaired after timeout, to ensure transparency and immutability.

[0141] Furthermore, by setting up a time-limited permission automatic prohibition mechanism, adaptive permission management is realized, solving problems such as abuse of permissions, maintenance delays, and device performance deviations by maintenance personnel during the maintenance of traditional faulty devices, as well as causing device data leakage, interface being attacked by viruses, and affecting device operation. Dynamically granting temporary permissions to maintenance personnel based on the role-based access control model significantly enhances the security, transparency, and efficiency of the maintenance process, while improving user trust and satisfaction. This mechanism not only optimizes resource utilization but also realizes the full traceability of the maintenance process and forms an intelligent operation and maintenance paradigm of "closed-loop permission governance" by cooperating with blockchain technology to record permission changes.

[0142] Device performance accuracy deviation calibration module after maintenance: Collect performance accuracy data of the device after maintenance, compare it with the original factory standard model to detect accuracy deviations, and use AI to automatically execute the minimum deviation calibration technology to obtain calibration results, so as to improve the performance accuracy of the device after maintenance, and transmit the calibration results back to the blockchain maintenance record update module to update the device health record;

[0143] Specifically, the steps to obtain the deviation of performance accuracy data are as follows:

[0144] Execute the standard test procedure of the device on the device after maintenance, and use multi-dimensional sensors to collect the performance data of the device after maintenance under the operating state, calibrated as X j , where j represents the number of performance types of intelligent office devices, such as load, rotation speed, response delay, electrical fluctuation, etc.;

[0145] Perform filtering and denoising and standardization processing on the performance data X j to obtain normalized performance accuracy data with the same dimension as the factory standard model. Among them, the formula for the normalized performance accuracy data is In the formula, represents the normalized performance accuracy data of the performance data X j after being standardized, Represents the benchmark mean of the standard performance parameters of the factory equipment stored in the blockchain, Represents the standard deviation of the standard performance parameters of the factory equipment stored in the blockchain;

[0146] The normalized performance accuracy data is sent through the MQTT protocol And the benchmark value of the standard performance parameters of the factory equipment stored in the blockchain, calibrated as X j0 , and use the AI algorithm to analyze the performance data to identify the performance deviation ΔX of the equipment j , where the AI algorithm formula for the performance deviation is In the formula, ΔX j Represents multi-dimensional anomaly detection, Represents the multi-dimensional vector of the normalized performance accuracy data, Represents the benchmark mean vector, T represents transpose, Represents the inverse matrix of the covariance matrix;

[0147] When Of the chi-square distribution threshold, trigger the calibration process.

[0148] Specifically, the process of the AI automatically performing the minimum deviation calibration is as follows:

[0149] Send a calibration instruction through the programmable logic controller PLC in the system, establish a reinforcement learning RL dynamic optimization model, and perform minimum deviation calibration calculation according to the performance deviation ΔX j ;

[0150] Take the normalized performance accuracy data As the state quantity, reflecting the deviation distribution between the currently repaired equipment and the factory equipment benchmark, and define the adjustable equipment performance control parameter as a continuous or discrete action, calibrated as z j ;

[0151] Design a reward function to simultaneously minimize the performance deviation and the adjustment amplitude of the control parameter, then the expression of the reward function In the formula, Represents the reward function for minimizing the deviation of the equipment performance, ΔX j Represents the performance deviation value, Represents the regularization coefficient to prevent the equipment from becoming unstable due to over-adjustment, ||z j || 2 Represents the performance control parameter penalty term;

[0152] Input the performance deviation ΔX j And the reward function to train the reinforcement learning RL dynamic optimization model, and deploy the trained model to the edge computing unit of the equipment, and collect the real-time performance data X jIn the input model, minimize the deviation optimization and output the calibration result through RL decision-making;

[0153] Among them, optimize the parameters of the reinforcement learning RL dynamic optimization model through gradient ascent optimization.

[0154] Furthermore, through the collection of performance accuracy data and comparison with the original factory standard model, establish a reinforcement learning RL dynamic optimization model, use AI to automatically execute the calibration technology, accurately identify millimeter-level accuracy deviations, effectively identify the performance accuracy of the repaired equipment and minimize the performance deviation of the equipment, ensure that the performance parameters of the equipment are close to being restored to the original factory standard, not only improve the accuracy and reliability of the equipment, but also record the calibration process through blockchain technology, enhancing the transparency and traceability of the repair process. And the calibration process optimizes the performance recovery of the repaired equipment, enhancing user trust and satisfaction.

[0155] Blockchain repair file update module: Use blockchain technology to record the entire repair process data in each module, ensure that the data is tamper-proof and traceable, and provide the historical entire repair process data to the device PHM supervision module for predicting the remaining life of the device;

[0156] Specifically, the construction steps of the blockchain repair file are as follows:

[0157] Collect the entire repair process data of the faulty equipment. The entire repair process data includes recording the repair results in the equipment fault diagnosis and grading module, the repair work order in the repair order generation module, the time-limited permission automatic prohibition mechanism in the repair mode adaptive permission module, and the calibration results in the performance accuracy deviation calibration module of the repaired equipment;

[0158] Perform data standardization processing on the entire repair process data to generate JSON structured data;

[0159] Use data hashing technology and digital signature technology for the JSON structured data, and package the data hash, signature, and timestamp into transaction data to ensure the authenticity and consistency of the data and ensure that the data is tamper-proof;

[0160] Use the consensus mechanism of the blockchain to automatically verify whether the transaction data is compliant, and the transactions are packaged into blocks in chronological order to form a chain structure. Only when it is judged to be compliant is it allowed to be uploaded to the chain, otherwise it is not;

[0161] The new block is broadcast to all participating nodes through the gRPC protocol, and the repair file is stored in the blockchain state database in the form of key-value pairs;

[0162] The blockchain repair file update module shares the entire repair process data to the PHM device PHM supervision module for predicting the remaining life of the device and evaluating the probability of fault recurrence.

[0163] Furthermore, by using blockchain technology, all the data of the entire maintenance process, including fault diagnosis results, work orders, permission changes, calibration results, etc., are encrypted and uploaded to the blockchain in the form of hash values. Any modification will disrupt the hash association on the chain, ensuring data authenticity, and solving the problem that traditional manual maintenance records using paper / electronic spreadsheets are easily tampered with or lost, resulting in difficulties in quality traceability. By recording the data generation sequence through timestamps and chain structures, the maintenance operations at any node can be quickly traced back, avoiding the scattered storage of historical maintenance data and the time-consuming query of historical records.

[0164] Device PHM supervision module: Integrate the real-time multi-dimensional sensor data in the device fault diagnosis and grading module with the historical maintenance process data in the blockchain maintenance file update module, and based on the LSTM-PHM hybrid model, predict the remaining useful life RUL of the device after maintenance and evaluate and analyze the probability of fault recurrence PFR.

[0165] Specifically, the prediction and evaluation logic of the LSTM-PHM hybrid model is as follows:

[0166] Extract the characteristic data of the multi-dimensional sensor data of the device in real time, and encode the historical maintenance process data using the BERT model;

[0167] Adopt the dynamic time warping DTW algorithm to unify the data with different sampling frequencies in the multi-dimensional sensor data to the same time axis, perform time alignment and then normalization processing, and integrate the characteristic data of the real-time multi-dimensional sensor data with the historical maintenance process data through multi-modal data fusion technology and the use of gated attention mechanism;

[0168] Extract the characteristic time series data related to the remaining useful life and the probability of fault recurrence from the characteristic data of the multi-dimensional sensor data, input it into the LSTM-PHM hybrid model, and use the long short-term memory network LSTM model to train and process the characteristic time series data, and capture the prediction results of the long-term degradation trend of the device;

[0169] Input the prediction results of the LSTM model into the fault prediction and health management PHM model for fusion, and then predict the probability of the device from the current moment to the recurrence of the next serious fault or the remaining time until the end of the service life,

[0170] Among them, the calculation formula for the probability of fault recurrence PFR is And In the formula, PFR represents the probability of fault recurrence evaluated by the LSTM-PHM hybrid model, A represents the mapping of the hidden state output by the LSTM model from the multi-dimensional sensor data and the historical maintenance process data to the risk value, h1 represents the characteristic of the current historical maintenance process data, and h0 represents the characteristic of the historical maintenance process data;

[0171] In the learning of the LSTM-PHM hybrid model, the logarithmic Cauchy loss function of the remaining useful life (RUL) and the loss function for solving class imbalance of the probability of fault recurrence (PFR) are respectively defined.

[0172] Among them, the expression of the logarithmic Cauchy loss function is In the formula, L RUL represents the logarithmic Cauchy loss function of the remaining useful life (RUL). n and N respectively represent the nth sample and the total number of samples N. b represents the scale parameter of the Cauchy distribution, which is used to control the sensitivity of the loss to outliers. RUL n represents the true RUL value of the nth sample of the device. represents the RUL value predicted by the LSTM-PHM hybrid model for the device;

[0173] The expression of the loss function for solving class imbalance is In the formula, L PFR represents the loss function for solving class imbalance of the probability of fault recurrence (PFR). c represents the class weight factor, and p n represents the probability of fault recurrence output by the LSTM-PHM hybrid model. d represents the focusing factor, which is used to reduce the weight of easily classified samples.

[0174] Furthermore, by using the LSTM-PHM hybrid model, the long short-term memory (LSTM) network model can process sequential data, capture the time-dependent relationship between multi-dimensional sensor data and the entire process data of historical maintenance, and learn the change pattern of the device state over time. The PHM model is used to describe the degradation process of the remaining life and fault recurrence of the device. By combining the two, their respective advantages are fully utilized for data analysis and prediction. Thus, it can relatively accurately predict the remaining useful life (RUL) of the device after maintenance, provide a scientific basis for the maintenance and replacement plan of the device, and accurately evaluate and analyze the probability of fault recurrence (PFR) so as to take preventive measures in advance and reduce the possibility of fault recurrence.

[0175] Embodiment 2

[0176] The present invention provides a maintenance management method for an intelligent office device as Figure 2 shown, including the following steps:

[0177] S1. Real-time fault monitoring and intelligent grading: Multidimensional sensor data is collected in real time through the built-in sensors of the intelligent office device, and the fault type is judged by combining historical fault data. After dynamically calculating the fault severity index, the maintenance priority is automatically divided.

[0178] S2. Dynamic work order generation and intelligent dispatching: Receive the repair results, combine GIS positioning technology and dynamic routing optimization algorithm to determine the corresponding repair personnel, generate repair work orders and conduct intelligent dispatching, and synchronize the repair work order information to the blockchain;

[0179] S3. Dynamic permission control and repair execution: Dynamically grant temporary permissions to repair personnel and set a time-limited automatic prohibition mechanism for permissions, restrict operation safety based on PHM performance baseline data, and record the spatio-temporal tags of repair operations on the blockchain;

[0180] S4. AI calibration and performance verification after equipment repair: Collect accuracy data, compare the detection accuracy deviation with the original factory standard model, and let AI automatically execute the minimum deviation calibration technology to obtain the calibration result;

[0181] S5. Chain storage and certification of the entire repair process data: Use blockchain technology to write the entire repair process data to ensure that the data is tamper-proof and traceable;

[0182] S6. Remaining useful life (RUL) prediction and probability of failure recurrence (PFR) assessment: Input the real-time multi-dimensional sensor data and the records of the historical entire repair process data, and use the LSTM-PHM hybrid model to predict the remaining useful life (RUL) of the equipment after repair and evaluate and analyze the probability of failure recurrence (PFR).

[0183] Furthermore, by using the multi-dimensional sensors built into intelligent office equipment, GIS positioning technology, dynamic routing optimization algorithm, dynamically granting temporary permissions to repair personnel, setting a time-limited prohibition mechanism, AI automatically executing the minimum deviation calibration technology, blockchain technology, and using the LSTM-PHM hybrid model to predict the remaining useful life of the equipment and evaluate the probability of failure recurrence, it solves the problems that traditional equipment fault diagnosis may rely on manual experience, lack real-time performance and accuracy, the allocation of repair personnel may lack a scientific basis, resulting in long repair response time and low efficiency; there is no effective control over the operation permissions and operation safety of repair personnel, which may lead to improper repairs or data tampering, and it is impossible to accurately predict the remaining useful life of the equipment and the probability of failure recurrence, which is not conducive to formulating a reasonable equipment maintenance plan. This improves the repair efficiency of intelligent office equipment while enhancing the enterprise's management level and operational benefits.

[0184] The repair management method of an intelligent office equipment provided by the embodiment of the present invention is implemented through the above-mentioned repair management system of an intelligent office equipment. The specific methods and processes of the repair management method of an intelligent office equipment are detailed in the embodiments of the above-mentioned repair management system of an intelligent office equipment, and will not be elaborated here.

[0185] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0186] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0187] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0188] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0189] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all of them should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A maintenance management system for intelligent office equipment, characterized in that, It includes a device fault diagnosis and grading module: Install multi-dimensional sensors on intelligent office devices, collect multi-dimensional sensor information in real time, combine historical fault data to analyze and judge the fault type, dynamically calculate the fault severity index, automatically divide the maintenance priority, and transmit the maintenance result to the maintenance order generation module; Maintenance order generation module: Receive the maintenance result in the device fault diagnosis and grading module, combine GIS positioning technology and dynamic routing optimization algorithm to generate a maintenance work order and intelligently assign orders, and synchronize the information of the maintenance work order to the blockchain maintenance file update module; Maintenance mode adaptive permission module: Dynamically grant temporary permissions to maintenance personnel based on the role-based access control model, and set a time-limited permission automatic prohibition mechanism. Maintenance personnel need to input the work order number and maintenance personnel ID to obtain the permission to access device debugging and maintenance, and record the space-time tag and operator identity of each permission change in the blockchain maintenance file update module. Based on the performance baseline data of the device PHM supervision module, limit the maintenance personnel from modifying the device performance parameters beyond a reasonable range; Module for calibrating performance accuracy deviation of the device after maintenance: Collect performance accuracy data of the device after maintenance, compare it with the original factory standard model to detect accuracy deviation, use AI to automatically execute the minimum deviation calibration technology to obtain the calibration result, and transmit the calibration result back to the blockchain maintenance file update module to update the device health file; Blockchain maintenance file update module: Use blockchain technology to record the entire maintenance process data in each module, and provide the historical entire maintenance process data to the device PHM supervision module; Device PHM supervision module: Integrate the real-time multi-dimensional sensor data in the device fault diagnosis and grading module and the historical entire maintenance process data in the blockchain maintenance file update module, and based on the LSTM-PHM hybrid model, predict the remaining useful life RUL of the device after maintenance and evaluate and analyze the probability of fault recurrence PFR.

2. The maintenance management system of an intelligent office device according to claim 1, characterized in that, The calculation process of the fault severity index is as follows: Deploy and install initialized multi-dimensional sensors on intelligent office devices. When a sudden fault response occurs during the operation of the intelligent office device, collect multi-dimensional sensor information in real time, including the device's operating status data, environmental data, and operation logs, which are respectively calibrated as Osd, Evd, and Old; Perform filtering, outlier detection, and normalization preprocessing on the collected multi-dimensional sensor information to obtain standardized data; The standardized data uses the MQTT protocol to achieve low-latency transmission and is stored in the device's blockchain historical data pool; According to the multi-dimensional sensor information and historical fault data, classify the current fault of the device, and use a causal inference engine to distinguish and determine the fault types of hardware faults, software faults, and environmental faults, which are respectively calibrated as Hwf, Swf, and Evf; According to the impact degree of each fault type on the overall performance of the device, as well as the instantaneous Δt change and trend during the sudden fault, introduce the concept of time window t, calculate the trend change amount within the time window t, and then dynamically adjust the weight value of the multi-dimensional sensor information; The calculation formula of the fault severity index is FSI = SI × (α·Hwf + β·Swf + γ·Evf), and α + β + γ = 1, ω1 + ω2 + ω3 = 1, where FSI represents the fault severity index, SI represents the dynamic regulation index, α, β, and γ respectively represent the initial weight values corresponding to the hardware fault Hwf, software fault Swf, and environmental fault Evf, ω1, ω2, and ω3 respectively represent the weights of the contribution of the operation status data Osd, environmental data Evd, and operation log Old in the multi-dimensional sensor information to the fault, Osd(t), Evd(t), and Old(t) respectively represent the sensor readings of the corresponding operation status data Osd, environmental data Evd, and operation log Old at the current time t, N Osd 、N Evd 、N Old respectively represent the historical maximum instantaneous change values corresponding to the operation status data Osd, the environmental data Evd, and the operation log Old. M Osd 、M Evd 、M Old are respectively represented as the historical maximum trend change slopes corresponding to the operation status data Osd, the environmental data Evd, and the operation log Old. δ represents the adjustment coefficient of the trend term, and δ ∈ (0, 1), l(Osd, t), l(Evd, t), and l(Old, t) respectively represent the linear regression slopes of the operation status data Osd, environmental data Evd, and operation log Old within the time window t, Respectively represented as instantaneous change items corresponding to the operation status data Osd, environmental data Evd, and operation log Old, They are respectively represented as trend change items corresponding to the operation status data Osd, the environmental data Evd, and the operation log Old; According to the calculated fault severity index FSI, the maintenance priority is automatically divided. When FSI > 0.7, it is determined as a high priority and urgent processing is required; when 0.4 < FSI ≤ 0.7, it is determined as a medium priority and important processing is required; when FSI ≤ 0.4, it is determined as a low priority and routine processing is required.

3. The maintenance management system of an intelligent office device according to claim 2, characterized in that, The steps for generating the maintenance work order are as follows: Match and analyze the candidate maintenance personnel who meet the skill certification required for maintenance according to the maintenance results, where the maintenance results include the fault type, severity index, and maintenance priority; Obtain the precise geographical location of the faulty device through GIS, extract the real-time locations, skill qualifications, and current task loads of all available maintenance personnel from the maintenance database, and screen and retain the candidate maintenance personnel who meet the skill certifications. Among them, the geographical location of the faulty device located by GIS is L0 = (lat0, lon0), where L0 represents the geographical location of the faulty device, and lat0 and lon0 respectively represent the latitude and longitude corresponding to the faulty device; the geographical location of the candidate maintenance personnel is L i = (lat i , lon i ), i = 1, 2, 3,..., n, representing the number of candidate maintenance personnel; Use the dynamic routing optimization algorithm to make intelligent decisions and allocate dispatch tasks, generating maintenance work orders. Among them, the dynamic routing optimization algorithm defines the total scheduling cost function as a multi-objective optimization problem, and the formula is and where Y represents the maintenance personnel selected by the dynamic routing optimization algorithm to minimize the total scheduling cost, θ, μ, are respectively represented as the weight coefficients corresponding to the distance D(L0, L i ), the time T(L0, L i ), and the priority division (1 - FSI) of the fault severity index D(L0,L i ) represents the geographical distance between the geographical location L0 of the faulty device and the geographical location L i of the maintenance personnel, T(L0,L i ) represents the estimated time of arrival for maintenance personnel from their own geographical location L i to the geographical location L0 of the faulty device, R represents the radius of the earth, and R = 6371 km, Indicates the average vehicle speed when the maintenance personnel travel to the geographical location of the faulty equipment, τ represents the traffic congestion coefficient; The generation of the maintenance work order is simultaneously updated to the blockchain maintenance file update module. The information content of the maintenance work order includes the work order number, fault type and severity index, equipment location, matching maintenance personnel ID, skill qualification, and estimated arrival time.

4. The maintenance management system of an intelligent office device according to claim 3, wherein The setting steps of the time-limited permission automatic prohibition mechanism are as follows: Based on the role-based access control model, permission application and verification are carried out. The maintenance personnel input the work order number and maintenance personnel ID information through the terminal to request permission to access the debugging and maintenance of the faulty equipment; The system reads the maintenance work order in the blockchain maintenance file update module for information comparison, verifies the work order number and maintenance personnel ID information of the maintenance personnel. After verification, the system dynamically grants the maintenance personnel temporary access permission, otherwise it does not grant; According to the fault type of the equipment and the PHM performance baseline data, restrict the operation permission and the permission to modify the equipment performance parameters required to access the equipment, and use a timer for permission time limit; When the time-limited permission times out, the system immediately terminates the current operation permission and the permission to modify the equipment performance parameters, and locks the equipment debugging interface to achieve forced recovery of the permission. If the equipment fault is still not resolved after the permission is recovered, the maintenance priority is automatically increased, triggering a new work order dispatching process; Use the blockchain maintenance record update module to record the status of maintenance events, including the status of being repaired after timeout, not being repaired after timeout, actively completing the maintenance within the time limit, the ID of the maintenance personnel, and the timestamp.

5. The maintenance management system of an intelligent office device according to claim 4, characterized in that, The steps for obtaining the deviation of the performance accuracy data are as follows: Execute the standard test procedure of the equipment on the repaired equipment, and collect the performance data of the repaired equipment under the operating state by using multi-dimensional sensors, calibrated as X j , where j represents the number of performance types of intelligent office equipment; For the performance data X j Perform filtering, denoising, and normalization processing to obtain normalized performance accuracy data with the same dimension as the factory standard model. Among them, the formula for the normalized performance accuracy data is In the formula,[[]] Represents the performance data X j The normalized performance accuracy data after normalization processing,[[]] Represents the reference mean value of the factory equipment standard performance parameters stored in the blockchain,[[]] Represents the standard deviation of the factory equipment standard performance parameters stored in the blockchain; Normalized performance accuracy data is sent through the MQTT protocol and the benchmark values of the device's factory standard performance parameters stored in the blockchain, calibrated as X j0 , and AI algorithms are used to analyze the performance data to identify the performance deviation ΔX of the device j , where the AI algorithm formula for performance deviation is In the formula, ΔX j represents multi-dimensional anomaly detection represents the multi-dimensional vector of normalized performance accuracy data represents the benchmark mean vector, T represents the transpose represents the inverse matrix of the covariance matrix; When the chi-square distribution threshold is reached, the calibration process is triggered.

6. The maintenance management system of an intelligent office device according to claim 5, characterized in that, The process for the AI to automatically perform minimum deviation calibration is as follows: Send calibration instructions through the programmable logic controller (PLC) in the system, establish a reinforcement learning (RL) dynamic optimization model, and perform minimization deviation calibration calculations based on the performance deviation ΔX j Perform minimization deviation calibration calculations; Take the normalized performance accuracy data as a state quantity to reflect the deviation distribution of the equipment after current maintenance from the benchmark of the factory equipment, and define the adjustable equipment performance control parameter as a continuous or discrete action, calibrated as z j ; Design a reward function to simultaneously minimize the performance deviation and the adjustment amplitude of the control parameter. Then the expression of the reward function In the formula,[[]] represents the reward function for minimizing the deviation of the device performance, and ΔX j represents the performance deviation value,[[]] represents the regularization coefficient, ||z j || 2 represents the penalty term of the performance control parameter; Input performance deviation ΔX j Train a reinforcement learning RL dynamic optimization model with the sum reward function, and deploy the trained model to the edge computing unit of the device. Input the real-time collected performance data X j into the model for minimizing deviation optimization, and output the calibration result through RL decision-making; Among them, the parameters of the reinforcement learning RL dynamic optimization model are optimized through gradient ascent.

7. The maintenance management system of an intelligent office device according to claim 6, characterized in that, The prediction evaluation logic of the LSTM-PHM hybrid model is as follows: Extract the feature data of the multi-dimensional sensor data of the device in real time, and encode the historical full-process maintenance data using the BERT model; Adopt the dynamic time warping DTW algorithm to unify the data with different sampling frequencies in the multi-dimensional sensor data to the same time axis, perform time alignment and then normalization processing, and integrate the feature data of the real-time multi-dimensional sensor data with the historical full-process maintenance data through multi-modal data fusion technology and using the gated attention mechanism; Extract the characteristic time series data related to the remaining useful life and the probability of fault recurrence of the device from the feature data of the multi-dimensional sensor data, input it into the LSTM-PHM hybrid model, and use the long short-term memory network LSTM model to train and process the characteristic time series data, and capture the prediction results of the long-term degradation trend of the device; Input the prediction results of the LSTM model into the fault prediction and health management PHM model for fusion, and then predict the probability of the device experiencing the recurrence of the next serious fault from the current moment or the remaining time until the end of the service life; Among them, the calculation formula of the probability of failure recurrence PFR is and In the formula, PFR represents the probability of failure recurrence evaluated by the LSTM-PHM hybrid model, A represents the mapping of the hidden state output by the LSTM model from multi-dimensional sensor data and historical maintenance full-process data to the risk value, h1 represents the feature of the current historical maintenance full-process data, and h0 represents the feature of the historical maintenance full-process data; In the learning of the LSTM-PHM hybrid model, define the logarithmic Cauchy loss function of the remaining useful life RUL and the loss function for solving class imbalance of the probability of fault recurrence PFR respectively; Among them, the expression of the logarithmic Cauchy loss function is In the formula, L RUL represents the logarithmic Cauchy loss function of the remaining useful life RUL. n and N respectively represent the nth sample and the total number of samples N. b represents the scale parameter of the Cauchy distribution. RUL n represents the true RUL value of the nth sample of the device, and represents the RUL value predicted by the LSTM-PHM hybrid model for the device; The expression of the loss function for solving class imbalance is In the formula, L PFR represents the loss function for solving class imbalance as the probability of failure recurrence (PFR), c represents the class weight factor, p n represents the probability of failure recurrence output by the LSTM-PHM hybrid model, and d represents the focusing factor.

8. A maintenance management method for an intelligent office device, which is implemented by a maintenance management system for an intelligent office device according to any one of claims 1-7, characterized in that, The steps include the following: S1. Real-time fault monitoring and intelligent classification: Real-time collect multi-dimensional sensor data through the built-in sensors of intelligent office equipment, combine historical fault data analysis to judge the fault type, dynamically calculate the fault severity index, and automatically divide the maintenance priority; S2. Dynamic work order generation and intelligent dispatching: Receive the maintenance results, combine the GIS positioning technology and the dynamic routing optimization algorithm to determine the corresponding maintenance personnel, generate a maintenance work order and perform intelligent dispatching, and synchronize the maintenance work order information to the blockchain; S3. Dynamic permission control and maintenance execution: Dynamically grant temporary permissions to maintenance personnel and set a time-limited permission automatic prohibition mechanism, restrict operation safety based on the PHM performance baseline data, and record the space-time tags of maintenance operations on the blockchain; S4. AI calibration and performance verification after equipment maintenance: Collect accuracy data to compare the detection accuracy deviation with the original factory standard model, and the AI automatically performs minimum deviation calibration technology to obtain the calibration results; S5. Chain storage and evidence preservation of the full-process maintenance data: Use blockchain technology to write the full-process maintenance data to ensure that the data is immutable and traceable; S6. Prediction of the remaining useful life RUL and evaluation of the probability of fault recurrence PFR: Input the records of real-time multi-dimensional sensor data and historical full-process maintenance data, and use the LSTM-PHM hybrid model to predict the remaining useful life RUL of the equipment after maintenance and evaluate and analyze the probability of fault recurrence PFR.

9. A computer device, comprising: A memory and a processor; the memory stores a computer program, characterized in that: when the processor executes the computer program, the steps of a maintenance management system for an intelligent office device according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of a maintenance management system for an intelligent office device according to any one of claims 1 to 7 are implemented.

Images