Dynamic management method and system for optimizing full life cycle cost of special equipment

Abstract

The invention discloses a dynamic management method and system for optimizing the full life cycle cost of special equipment, and relates to the technical field of equipment management, and the method comprises the steps: collecting the operation data and environment data of the special equipment, preprocessing the data, and inputting the data into a fault prediction sub-model to predict the fault occurrence probability and fault parts of the special equipment; constructing a fuzzy comprehensive evaluation model to divide risk levels of the special equipment; combining the risk level of the special equipment and the prediction benefit of the benefit prediction sub-model, and making a dynamic maintenance strategy of the special equipment through the maintenance cost; and overhauling the special equipment according to the dynamic overhauling strategy, and optimizing the data model and the dynamic overhauling strategy according to the actual overhauling cost. By predicting the equipment fault occurrence probability, the optimal maintenance time of the equipment is dynamically determined, the maintenance cost and the shutdown loss cost are reduced, the maintenance efficiency is improved, the service life of the equipment is prolonged, and the whole life cycle cost of the equipment is controlled from the source.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment management, and more specifically to a dynamic management method and system for optimizing the full life cycle cost of special equipment. Background Art

[0002] The total life cycle cost refers to the sum of all costs incurred during the life cycle of the equipment, from initial investment, operation and maintenance to final retirement and scrapping. Special equipment usually requires large investment, complex technology, and harsh operating environment. The life cycle cost of the equipment involves multiple links. For special equipment enterprises, the equipment has a long service life. For the equipment that has been purchased, the initial purchase cost has been determined, but for equipment with an extended operating life, the maintenance cost of the equipment will increase significantly. By optimizing the equipment maintenance cost, the equipment life cycle cost can be optimized, and the lean management of the equipment life cycle cost can be achieved.

[0003] At present, most companies have two main ways to maintain equipment: one is to perform regular maintenance on the equipment, and the other is to rely on emergency maintenance after a failure occurs. These two maintenance systems are not sound, resulting in additional downtime losses, thereby increasing the life cycle cost of the equipment. An optimization management method for the full life cycle cost of equipment is urgently needed to reduce the sudden downtime caused by equipment failure during the production process and increase the full life cycle cost of the equipment.

[0004] The existing Chinese patent application with publication number CN117455710A discloses a method, device and computer equipment for determining the full life cycle cost of equipment, which includes obtaining the resource cost of the full life cycle of equipment under various maintenance strategies according to equipment management conditions; obtaining the carbon cost of the full life cycle of equipment under various maintenance strategies; and obtaining the full life cycle cost of equipment under various maintenance strategies according to the resource cost and carbon cost of the full life cycle of the equipment. The method can accurately calculate the full life cycle cost of equipment.

[0005] The existing Chinese patent with publication number CN111882162A discloses a method and device for equipment life cycle cost management, which includes the following steps: synchronizing the equipment hierarchy structure in the production management system to the asset management system; obtaining the maintenance work order from the production management system, and importing the maintenance work order into the asset management system; obtaining the equipment associated with the maintenance work order according to the equipment hierarchy result, and attributing the maintenance cost to the equipment associated with the maintenance work order; combining the maintenance cost of each equipment and the purchase cost of each equipment in the asset management system to achieve full life cycle cost management of each equipment. This invention can manage the cost of the equipment throughout its life cycle without repeated information entry, and has high management efficiency.

[0006] The above existing technologies manage the full life cycle cost of the equipment by calculating the equipment maintenance cost, but do not arrange maintenance plans in advance for possible failures of the equipment during operation, resulting in additional downtime loss costs and increasing the additional equipment life cycle cost caused by unreasonable equipment failure maintenance. Summary of the invention

[0007] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a dynamic management method and system for optimizing the full life cycle cost of special equipment. Through the equipment's operating data and environmental data, the probability of equipment failure is predicted, and the risk level of special equipment is divided. A dynamic maintenance strategy is formulated according to the failure probability and risk level, and the optimal maintenance time or maintenance method of special equipment is determined, thereby reducing the maintenance cost of special equipment and the production losses caused by equipment downtime, extending the service life of special equipment, and controlling the full life cycle cost of special equipment from the source.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] On the one hand, the present invention provides a dynamic management method for optimizing the life cycle cost of special equipment, which specifically includes the following steps:

[0010] Constructing a data model for the entire life cycle of special equipment, wherein the data model includes a fault prediction sub-model and a benefit prediction sub-model;

[0011] Collect the operation data and environmental data of special equipment, and input them into the fault prediction sub-model after pre-processing to predict the failure probability and faulty components of special equipment;

[0012] Based on the probability of failure and faulty components, combined with the characteristics of special equipment, a fuzzy comprehensive evaluation model is constructed to classify the risk level of special equipment;

[0013] Combining the risk level of special equipment and the predicted benefits of the benefit prediction sub-model, a dynamic maintenance strategy for special equipment is formulated through maintenance costs;

[0014] The special equipment is maintained according to the dynamic maintenance strategy, the actual maintenance cost is recorded, and the data model and the dynamic maintenance strategy are optimized according to the actual maintenance cost.

[0015] As a further improvement of the present invention, the construction of a fuzzy comprehensive evaluation model to classify the risk level of special equipment specifically includes:

[0016] Obtain the probability of failure and the faulty components, as well as the operating data, select evaluation factors, and determine the evaluation factor set;

[0017] The risk level of special equipment is divided into three levels, namely normal operation, low risk and high risk, to construct an evaluation level set;

[0018] Compare the relative importance of each evaluation factor in the evaluation factor set, construct a judgment matrix, calculate the eigenvector and maximum eigenvalue in the judgment matrix, and obtain the weight of each evaluation factor;

[0019] Determine the degree of membership of each evaluation factor to different risk levels and establish a fuzzy relationship matrix;

[0020] Through fuzzy synthesis operation, the comprehensive evaluation result vector is obtained, and the risk level of special equipment is determined according to the maximum membership principle.

[0021] As a further improvement of the present invention, the dynamic maintenance strategy includes a cost priority maintenance strategy and a time priority maintenance strategy, and the dynamic maintenance strategy for special equipment formulated by maintenance cost specifically includes:

[0022] Configure the fault probability threshold. When the fault probability of special equipment is less than or equal to the fault probability threshold or the risk level is normal operation, no dynamic maintenance strategy is formulated.

[0023] When the failure probability of special equipment is greater than the failure probability threshold and the risk level is normal operation or low risk, formulate a cost-priority maintenance strategy;

[0024] When the risk level of special equipment is high, formulate a time-priority maintenance strategy.

[0025] As a further improvement of the present invention, the specific steps of formulating the cost-priority maintenance strategy include:

[0026] S41: When the probability of a fault occurring is greater than a threshold value of the probability of a fault occurring, a warning signal is triggered to provide a reminder for equipment maintenance;

[0027] S42: taking the time of receiving the equipment maintenance reminder as the start time node of the equipment maintenance process, taking the latest maintenance time point as the end time node, and dividing by a preset first fixed time length as an interval to form a first maintenance time node sequence;

[0028] S43: Calculate the total maintenance cost that is expected to be generated when each maintenance time node in the first maintenance time node sequence starts to perform maintenance work, and associate the total maintenance cost with the first maintenance time node to form a first maintenance total cost set;

[0029] S44: configuring a total cost threshold, and adding the maintenance time nodes corresponding to the maintenance total cost in the first maintenance total cost set that are less than the total cost threshold to the time node set;

[0030] S45: judging and analyzing the time node set to form a second maintenance time node sequence, and calculating to obtain a second maintenance total cost set;

[0031] S46: sorting the total maintenance costs in the first total maintenance cost set and the second total maintenance cost set to form a maintenance cost sequence from low to high;

[0032] S47: Selecting the maintenance time node corresponding to the lowest total maintenance cost from the maintenance cost sequence as the optimal equipment maintenance time.

[0033] As a further improvement of the present invention, the step of judging and analyzing the time node set to form a second maintenance time node sequence and calculating the second maintenance total cost set includes:

[0034] Determine whether there are adjacent maintenance time nodes in the time node set. If there are adjacent maintenance time nodes, merge the adjacent maintenance time nodes.

[0035] The earliest maintenance time node among the adjacent maintenance time nodes is used as the secondary judgment start time node, the latest maintenance time node among the adjacent maintenance time nodes is used as the secondary judgment end time node, and the time length between the secondary judgment start time node and the secondary judgment end time node is used as the secondary judgment maintenance time period;

[0036] Divide the secondary judgment maintenance time period into intervals of a second fixed time length to form a second maintenance time node sequence, calculate the total maintenance cost that is expected to be generated when each maintenance time node in the second maintenance time node sequence starts to perform maintenance work, and associate the total maintenance cost with the maintenance time node to form a second maintenance total cost set;

[0037] If there are no adjacent maintenance time nodes, the second maintenance total cost set is defined as an empty set.

[0038] As a further improvement of the present invention, the calculation of the total maintenance cost expected to be generated when the maintenance work is started includes:

[0039] Obtain the downtime loss costs caused by downtime during maintenance through the accumulated production benefits of special equipment;

[0040] The equipment maintenance cost of special equipment in the expected maintenance process is calculated by labor costs, replacement costs of equipment parts and nonlinear equipment health status influencing factors; among which, labor costs are the product of unit labor hour cost and labor hours spent;

[0041] Obtain the probability of failure at each maintenance time node, and combine it with the average cost of historical maintenance costs to obtain potential future costs, where the potential future costs are used to represent the loss costs caused by failures that are expected to occur due to unmaintained maintenance from the initial time node to the maintenance time node;

[0042] The estimated total maintenance cost is obtained by taking the weighted sum of the downtime loss cost, equipment maintenance cost and potential future cost.

[0043] As a further improvement of the present invention, the acquisition of the downtime loss cost caused by downtime during the maintenance process specifically includes:

[0044] Obtain the cumulative production benefit value corresponding to the initial time node; the cumulative production benefit is the cumulative production volume multiplied by the economic benefits that can be created by the unit production volume;

[0045] The benefit prediction sub-model is used to predict the cumulative production benefit value corresponding to each maintenance time node;

[0046] The downtime loss cost is determined by the cumulative benefit difference between the cumulative production benefit value corresponding to the initial time node and the cumulative production benefit value corresponding to the maintenance time node, the time length between the initial time node and the maintenance time node, and the expected downtime duration.

[0047] As a further improvement of the present invention, the specific steps of formulating the time-priority maintenance strategy include:

[0048] The maintenance method is used as the decision node, the cost factor is used as the branch condition, and the final total maintenance cost is used as the leaf node; the maintenance method includes comprehensive maintenance and key component maintenance, and the cost factor data includes direct maintenance cost, downtime loss cost, and safety risk cost;

[0049] With the goal of minimizing the total maintenance cost, the decision tree model is trained by dividing the branches according to the value range of different cost factors;

[0050] When the risk level of special equipment is high risk, the cost factor data of special equipment are collected and preprocessed. The preprocessed cost factor data are input into the trained decision tree model. Starting from the root node, the decision tree is traversed downward according to the value of the cost factor until the leaf node is reached. The maintenance method corresponding to the leaf node is the output optimal maintenance method.

[0051] As a further improvement of the present invention, the fault prediction sub-model includes: an input layer, an LSTM layer, an attention layer, a fully connected layer and an output layer;

[0052] The input layer is used to input preprocessed data;

[0053] The LSTM layer contains a hidden layer of multiple LSTM units to capture long-term dependencies in dynamic data;

[0054] The attention layer is used to calculate the attention weights for the output sequence of the LSTM layer and obtain the attention output through weighted summation;

[0055] The fully connected layer integrates the features by connecting the attention output of the attention layer;

[0056] The output layer uses the Softmax function to predict the probability of failure and the failure of special equipment components.

[0057] In a second aspect, the present invention provides a dynamic management system for optimizing the life cycle cost of special equipment, including a model building module, a data acquisition module, a fault prediction module, a risk classification module, and a dynamic maintenance module, wherein:

[0058] The model building module is used to build a data model of the entire life cycle of special equipment, and the data model includes a fault prediction sub-model and a benefit prediction sub-model;

[0059] The data acquisition module is used to collect the operation data, environmental data, and accumulated production benefits of special equipment in real time and perform preprocessing;

[0060] The fault prediction module is used to predict the probability of failure of special equipment and the faulty components;

[0061] The risk classification module is used to classify the risk level of special equipment by building a fuzzy comprehensive evaluation model based on the probability of failure and faulty components, combined with the characteristics of special equipment;

[0062] The dynamic maintenance module is used to formulate a dynamic maintenance strategy for special equipment based on maintenance costs, combining the risk level of special equipment and the predicted benefits of the benefit prediction sub-model;

[0063] The feedback optimization module is used to record the actual maintenance cost and optimize the data model and dynamic maintenance strategy based on the actual maintenance cost.

[0064] Beneficial effects of the present invention:

[0065] By continuously collecting and analyzing the real-time operating status and environmental data of the equipment, and using machine learning technology to build a fault prediction sub-model for the equipment, the probability of equipment failure and the faulty components are predicted, so that equipment managers can foresee potential problems before failures occur and take preventive maintenance measures, significantly reducing downtime caused by sudden failures and ensuring the continuity and efficiency of equipment operation;

[0066] Different dynamic management strategies are formulated based on the probability of equipment failure and the risk level of special equipment. The optimal maintenance time of the equipment is dynamically determined through the total maintenance cost. This strategy can avoid unnecessary early maintenance and reduce maintenance costs. It can also prevent high repair costs caused by excessive wear of equipment and reduce production loss costs caused by equipment downtime. By accurately predicting and reasonably arranging maintenance time, the optimal allocation of maintenance resources is achieved, maintenance efficiency is improved, the service life of equipment is extended, and the full life cycle cost of equipment is controlled from the source. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A flow chart of a dynamic management method for optimizing the full life cycle cost of special equipment provided by the present invention;

[0068] Figure 2 A flow chart of the cost-first maintenance strategy provided by the present invention;

[0069] Figure 3 An interactive diagram of a dynamic management system module for optimizing the full life cycle cost of special equipment provided by the present invention. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] Example 1

[0072] refer to Figure 1 and Figure 2 As shown, a specific implementation method of the dynamic management method for optimizing the full life cycle cost of special equipment according to the present invention is shown.

[0073] Specifically, special equipment refers to equipment that involves life safety and has high risks during use. Such equipment usually needs to comply with strict regulations and standards, and requires regular inspection, maintenance and monitoring to ensure the safe operation of the special equipment. Common special equipment includes boilers, pressure vessels, elevators, cranes, etc.

[0074] A dynamic management method for optimizing the life cycle cost of special equipment specifically includes the following steps:

[0075] S1. Construct a data model for the entire life cycle of special equipment, wherein the data model includes a fault prediction sub-model and a benefit prediction sub-model;

[0076] By constructing a data model for special equipment, reliable analysis results can be provided for subsequent analysis of special equipment during operation. According to the analysis results, data support can be provided for maintenance and overhaul plans, improving the efficiency and accuracy of special equipment detection, thereby formulating reasonable maintenance and overhaul plans and optimizing the management costs of special equipment during operation.

[0077] S2. Collect the operation data and environmental data of the special equipment, and input them into the fault prediction sub-model after pre-processing to predict the failure probability and faulty components of the special equipment at a preset time point;

[0078] By comprehensively collecting the operating data of special equipment, such as operating time, load rate, fatigue strength and other data, a comprehensive data source is provided for the evaluation of the operating status of special equipment. By analyzing the impact of environmental data on special equipment during operation, such as excessively high ambient humidity will corrode special equipment, and excessively high ambient temperature will affect the material properties of special equipment and increase the risk or probability of failure, the potential failures or problems of special equipment can be comprehensively predicted by combining operating data with environmental data, which can increase the accuracy and reliability of special equipment failure prediction.

[0079] Preferably, the operation data of the special equipment includes static data and dynamic data, and the dynamic data is used to record the time series data monitored by the special equipment during operation, including operation time, load rate, vibration, corrosion rate, stress, load and fatigue strength, which are collected and acquired by corresponding sensors respectively;

[0080] The static data is used to record the historical data of special equipment, including the expected service life of special equipment, the actual service life of special equipment, historical maintenance data, special equipment maintenance type, replacement or repair parts records, the cost of each maintenance, and industry standard data; among which special equipment maintenance types include preventive maintenance, corrective maintenance, and emergency maintenance; industry standard data include national standards and regulations or industry association guidelines; the historical data of special equipment is obtained through the technical documents of special equipment or the internal management system of the enterprise.

[0081] Preferably, the fault prediction sub-model introduces the long short-term memory network in time series analysis in combination with the attention mechanism to perform in-depth analysis on the dynamic data and static data of special equipment to more accurately calculate the probability of future failures.

[0082] Furthermore, the specific steps of S2 include:

[0083] S21: Preprocessing of operation data and environmental data, including: using sliding window algorithm to identify and remove outliers, using multiple filling methods based on deep learning for missing values, using generative adversarial networks (GAN) combined with historical data to generate reasonable filling values ​​to ensure data integrity and accuracy; normalizing different types of data;

[0084] S22: Construct a fault prediction sub-model, including: an input layer, an LSTM layer, an attention layer, a fully connected layer and an output layer, wherein the input layer is used to input preprocessed data; the LSTM layer contains a hidden layer of multiple LSTM units, which is used to capture the long-term dependencies in time series data (dynamic data) and perform deep feature extraction on the input time series data; the attention layer is used to calculate the attention weights for the output sequence of the LSTM layer, and then obtain the attention output by weighted summation, so that the model can focus on the most important time step features for fault prediction; the fully connected layer integrates the features by connecting the attention output of the attention layer, wherein the fully connected layer includes at least two layers, and the number of neurons in each layer gradually decreases; the output layer uses a Softmax function for predicting the probability of fault occurrence and special equipment component failure. For the prediction of the probability of fault occurrence, the probability of the fault occurring and not occurring is output; for the prediction of special equipment component failure, the probability distribution of possible failure of each component is output.

[0085] S23: Divide the preprocessed data into a training set, a validation set, and a test set according to a preset ratio, use the validation set to train the fault prediction sub-model, perform multiple iterations on the training set to update the model parameters, use the validation set to validate the fault prediction sub-model, use the early stopping method to prevent overfitting, and stop training when the loss function on the validation set converges to obtain a trained fault prediction sub-model. The loss function uses the cross entropy loss function, and the Adam W optimizer is used to prevent model overfitting.

[0086] S24: Input the dynamic data collected and preprocessed in real time into the trained fault prediction sub-model, output the probability value of the fault occurrence, and at the same time, compare the predicted failure probability distribution of each equipment component in the output layer with the preset component probability threshold of each equipment component. When the predicted failure probability of the equipment component is greater than the corresponding component probability threshold, it is determined that a component failure may occur.

[0087] S3. Based on the probability of failure and faulty components, combined with the characteristics of special equipment, a fuzzy comprehensive evaluation model is constructed to classify the risk level of special equipment;

[0088] According to the failure probability output by the fault prediction sub-model, as well as the equipment components that may fail, combined with other data of special equipment, including operation data, historical failure rates and the complexity of the equipment operation environment, the risk level of special equipment is comprehensively assessed to provide a basis for the subsequent formulation of corresponding dynamic maintenance strategies, so as to formulate different maintenance methods or times according to different risk levels, optimize the costs incurred by special equipment due to failures during operation, and thus optimize the full life cycle cost of special equipment.

[0089] Furthermore, constructing a fuzzy comprehensive evaluation model to classify the risk level of special equipment specifically includes the following steps:

[0090] S31: Obtain the probability of failure, the faulty component, and the operation data, select evaluation factors, and determine the evaluation factor set;

[0091] Preferably, in addition to the failure probability and faulty components of special equipment, key data are selected as evaluation factors based on the operating data combined with the characteristics of special equipment. For example, for boilers, the pressure abnormality (the deviation ratio of pressure exceeding the normal range), temperature deviation (interpolation of actual temperature and equipment temperature), water quality (indicators such as impurities and pH in water), and historical failure probability are selected as key data to determine the evaluation factor set. For example, for cranes, the degree of deviation of lifting weight (the difference ratio between actual lifting weight and rated lifting weight), the degree of abnormal lifting speed (the deviation between actual lifting speed and standard lifting speed), the wear rate of running wheel (the degree of wear of running wheel, determined by measuring the ratio of the reduction in running wheel diameter to the initial diameter) are selected as key data to determine the evaluation factor set. The specific evaluation factor set is determined by the technicians in this field according to the characteristics of special equipment by selecting appropriate key factors. By selecting evaluation factors, the safety risk status of special equipment is reflected from multiple dimensions, and these factors are comprehensively considered to avoid the one-sidedness of a single factor.

[0092] S32: The risk level of special equipment is divided into three levels, namely normal operation, low risk and high risk, and an evaluation level set is constructed;

[0093] S33: performing pairwise comparisons on the relative importance of each evaluation factor in the evaluation factor set, constructing a judgment matrix, calculating the eigenvector and the maximum eigenvalue in the judgment matrix, and obtaining the weight of each evaluation factor;

[0094] S33 specifically includes the following steps:

[0095] The expert survey method is used to compare the relative importance of each evaluation factor in the evaluation factor set. For example, the impact of the failure probability and equipment operation data on the safety risk is compared, and the experts make a relative importance judgment based on their experience.

[0096] S331: Construct a judgment matrix based on the expert's judgment results , where n is the total number of evaluation factors, Indicates evaluation factors and evaluation factors The relative importance of is 1-9 and its reciprocal, 1 means both are equally important, 9 means Compare Extremely important;

[0097] S332: Calculate the product of the elements of each row of the judgment matrix, then take the nth root of the obtained product to obtain the initial weight, normalize the initial weight to obtain the weight of each evaluation factor, and form the eigenvector of the judgment matrix;

[0098] S333: Calculate the maximum eigenvalue based on the judgment matrix and the eigenvector; the calculation formula of the maximum eigenvalue is: ,in, is the maximum eigenvalue, W represents the eigenvector of the judgment matrix, represents the weight of the i-th evaluation factor;

[0099] S334: Calculate the consistency ratio by the maximum eigenvalue; the calculation formula of the consistency ratio is: ,in, represents the consistency ratio, represents the average random consistency index, obtained by looking up the table;

[0100] S335: Compare the consistency ratio with the preset threshold. If the consistency ratio is less than the preset threshold, the consistency of the judgment matrix is ​​acceptable, indicating that the calculated eigenvector as the weight of each evaluation factor is reasonable and reliable. If the consistency ratio is greater than or equal to the preset threshold, it means that the consistency of the judgment matrix is ​​poor, and it is necessary to readjust the expert's judgment or correct the judgment matrix. Repeat S332~S334, recalculate the eigenvector and the maximum eigenvalue, until the consistency test passes.

[0101] The weight reflects the relative importance of each evaluation factor in the safety risk assessment. The weight of the evaluation factor is determined by constructing a judgment matrix, and the rationality of the weight distribution is judged through a consistency test, so that the evaluation result is more in line with the actual situation and avoids estimation deviations caused by unreasonable weights of evaluation factors. For example, if the probability of a failure is higher, it means that it has a greater impact on the risk level and should be paid special attention to during the assessment.

[0102] S34: Determine the degree of membership of each evaluation factor to different risk levels and establish a fuzzy relationship matrix;

[0103] Through the statistical analysis of historical data, the membership of each evaluation factor to different risk levels is determined to form a membership vector, and the membership vectors of all evaluation factors are combined into a fuzzy relationship matrix; the fuzzy relationship matrix reflects the degree of association between each evaluation factor and different risk levels, and is the key data for fuzzy comprehensive evaluation. It converts the specific status of each factor into a membership relationship with different risk levels, providing a data basis for subsequent comprehensive evaluation.

[0104] S35: Through fuzzy synthesis operation, a comprehensive evaluation result vector is obtained, and the risk level of the special equipment is determined according to the maximum membership principle.

[0105] Through fuzzy synthesis operation, all evaluation factors and their weights are comprehensively considered to obtain a comprehensive evaluation result vector that fully reflects the safety risk status of special equipment. The maximum membership principle makes the determination of risk level clearer and more objective, and can quickly and accurately judge the risk level of equipment.

[0106] S4. Combine the risk level of special equipment and the predicted benefit of the benefit prediction sub-model to formulate a dynamic maintenance strategy for special equipment through maintenance cost;

[0107] Different maintenance strategies are formulated based on the risk level of special equipment and the probability of equipment failure. Different maintenance and repair methods are adopted for different situations to reduce unnecessary resource allocation and waste. For special equipment operating normally, no maintenance is required. For special equipment with certain risks, different maintenance times and methods are determined according to the maintenance cost, thereby achieving control over the maintenance cost of special equipment.

[0108] The dynamic maintenance strategy includes a cost-first maintenance strategy and a time-first maintenance strategy;

[0109] Configure the fault probability threshold. When the fault probability of special equipment is less than or equal to the fault probability threshold or the risk level is normal operation, no dynamic maintenance strategy is formulated. This indicates that the special equipment is in good operating condition and can maintain stable operation for a long time without maintenance, thus saving management costs.

[0110] When the failure probability of special equipment is greater than the failure probability threshold and the risk level is normal operation or low risk, a cost-priority maintenance strategy is formulated; this indicates that there may be some potential problems or failures in the operation status of the special equipment, and a maintenance plan needs to be arranged. Due to the production schedule of special equipment at different times, the accumulation of equipment operation status, etc., the cost of maintenance at different times may be different. Therefore, choosing the appropriate maintenance time node can optimize and reduce costs;

[0111] When the risk level of special equipment is high, a time-priority maintenance strategy is formulated; this indicates that the special equipment is very likely to have a fault, and if it continues to operate, it may have a greater impact and requires immediate shutdown and maintenance. However, different maintenance methods will incur different maintenance costs. By formulating a time-priority maintenance strategy and selecting the appropriate maintenance method for maintenance, costs can be further optimized.

[0112] Furthermore, the specific steps of the cost-first maintenance strategy include: S41: When the probability of a fault occurring is greater than a threshold value of the probability of a fault occurring, a warning signal is triggered to provide a reminder for equipment maintenance;

[0113] S42: The time of receiving the equipment maintenance reminder is used as the start time node of the equipment maintenance process, the latest maintenance time point is used as the end time node, and the first fixed time length is used as the interval to form a first maintenance time node sequence; wherein the first fixed time length and the latest maintenance time point are set by the management or production personnel according to actual needs;

[0114] S43: Calculate the total maintenance cost that is expected to be generated when each maintenance time node in the first maintenance time node sequence starts to perform maintenance work, and associate the total maintenance cost with the first maintenance time node to form a first maintenance total cost set;

[0115] S44: configuring a total cost threshold, adding the maintenance time nodes corresponding to the maintenance total cost in the first maintenance total cost set that are less than the total cost threshold to the time node set; wherein the total cost threshold is set by those skilled in the art according to actual needs;

[0116] S45: judging and analyzing the time node set to form a second maintenance time node sequence, and calculating to obtain a second maintenance total cost set;

[0117] The specific steps of S45 include:

[0118] Determine whether there are adjacent maintenance time nodes in the time node set. If there are adjacent maintenance time nodes, merge the adjacent maintenance time nodes.

[0119] The earliest maintenance time node among the adjacent maintenance time nodes is used as the secondary judgment start time node, the latest maintenance time node among the adjacent maintenance time nodes is used as the secondary judgment end time node, and the time length between the secondary judgment start time node and the secondary judgment end time node is used as the secondary judgment maintenance time period;

[0120] Divide the secondary judgment maintenance time period into intervals of a second fixed time length to form a second maintenance time node sequence, calculate the total maintenance cost that is expected to be generated when each maintenance time node in the second maintenance time node sequence starts to perform maintenance work, and associate the total maintenance cost with the maintenance time node to form a second maintenance total cost set;

[0121] If there are no adjacent maintenance time nodes, the second maintenance total cost set is defined as an empty set.

[0122] The second fixed time is at least less than half of the first fixed time, so that a new maintenance time node is added to the adjacent maintenance time nodes, and a more detailed maintenance time node is determined, so as to avoid missing the optimal maintenance time node due to the excessively large step size of the first fixed time, and also reduce the calculation burden increased due to the excessively small step size of the first fixed time. The second fixed time is specifically set by the management or production personnel according to actual needs;

[0123] S46: sorting the total maintenance costs in the first total maintenance cost set and the second total maintenance cost set to form a maintenance cost sequence from low to high;

[0124] S47: Selecting the maintenance time node corresponding to the lowest total maintenance cost from the maintenance cost sequence as the optimal equipment maintenance time.

[0125] For example, if a special equipment, such as a boiler, detects that the probability of a failure is greater than a threshold value of the probability of a failure (such as 0.7) at 9:00, an early warning signal is triggered, and an equipment maintenance reminder is provided. 9:00 is used as the start time node of the equipment maintenance process, and the management or production personnel sets the maintenance to be completed within 12 hours, that is, 21:00 is used as the end time node. The maintenance time nodes obtained are 9:00, 12:00, 15:00, 18:00, and 21:00, forming a first maintenance time node sequence. The total maintenance cost expected to be incurred when the maintenance work starts at each maintenance time node is calculated, such as They correspond to 16,000 yuan, 14,000 yuan, 17,000 yuan, 12,000 yuan, and 10,000 yuan respectively, forming the first maintenance total cost set. The total cost threshold is configured to be 15,000 yuan. Then, in the first maintenance total cost set, the maintenance time nodes corresponding to the maintenance total cost less than the corresponding total cost threshold are 12:00, 18:00, and 21:00. They are added to the time node set, and it is found that 18:00 and 21:00 are adjacent maintenance time nodes and are merged. 18:00 is used as the start time node of the second judgment, 21:00 is used as the end time node of the second judgment, and the three hours between 18:00 and 21:00 are used as the maintenance time period for the second judgment.

[0126] The secondary judgment maintenance time period is divided into intervals of the second fixed time length (such as 1 hour), and the obtained maintenance time nodes are 18:00, 19:00, 20:00, and 21:00, forming a second maintenance time node sequence. The total maintenance cost estimated to be incurred when the maintenance work starts at each maintenance time node in the second maintenance time node sequence is calculated, such as 12,000 yuan, 8,000 yuan, 9,000 yuan, and 10,000 yuan respectively, forming the second maintenance total cost set. All total maintenance costs are sorted to form a maintenance cost sequence from low to high, and the maintenance time node (19:00) corresponding to the lowest total maintenance cost (8,000 yuan) is selected as the optimal equipment maintenance time.

[0127] By dividing the maintenance time into different nodes and determining the optimal equipment maintenance time according to the total maintenance cost of each maintenance time node, the maintenance cost of special equipment can be optimized after the equipment maintenance reminder is issued, avoiding cost waste due to inappropriate maintenance time selection.

[0128] The calculation of the total maintenance cost expected to be incurred when starting maintenance work at each maintenance time node includes:

[0129] Through the accumulated production benefits of special equipment, the downtime loss cost caused by downtime during the maintenance process is obtained; specifically, the following steps are included:

[0130] Obtain the cumulative production benefit value corresponding to the initial time node; the cumulative production benefit is the cumulative production volume multiplied by the economic benefits that can be created by the unit production volume;

[0131] Predict the cumulative production benefit value corresponding to each maintenance time node through the benefit prediction sub-model; the benefit prediction sub-model is trained by obtaining historical production data and market data, and the benefit prediction sub-model is an ARIMA model or an LSTM model;

[0132] The downtime loss cost is determined by the cumulative benefit difference between the cumulative production benefit value corresponding to the initial time node and the cumulative production benefit value corresponding to the maintenance time node, the time length between the initial time node and the maintenance time node, and the expected downtime duration; the expected downtime duration is represented by the average downtime duration of the faults in the maintenance database that are the same or similar to the current fault in actual maintenance; the specific calculation method of the downtime loss cost is:

[0133]

[0134] in, represents the downtime loss cost, represents the cumulative production benefit value corresponding to the initial time node, represents the cumulative production benefit value corresponding to the kth time node, It represents the time length from the initial time node to the kth maintenance time node, and t represents the expected downtime maintenance duration.

[0135] The equipment maintenance cost of special equipment in the expected maintenance process is calculated by labor costs, replacement costs of equipment parts and nonlinear equipment health status influencing factors; among which, labor costs are the product of unit labor hour cost and labor hours spent;

[0136]

[0137] in, represents the equipment maintenance cost, H represents the number of man-hours required to repair the equipment, L represents the unit man-hour cost, and R represents the replacement cost of equipment parts. represents the influencing factor of the health status of nonlinear equipment, is the equipment health state decay rate constant, which is obtained by calculating the average rate of equipment health state degradation over time using the regression model through the historical maintenance data in the maintenance database;

[0138] It should be noted that the nonlinear equipment health status influencing factor is fitted by fitting the equipment maintenance cost with the maintenance time node and related health status indicators. The model parameters are estimated by the least squares method and other methods to obtain the quantitative relationship between the nonlinear equipment health status influencing factor and the maintenance time node. Since there may be potential problems or failures inside the special equipment, the later the maintenance time node, the greater the impact on the health status of the equipment, and the more difficult the maintenance and the longer the man-hours may be. By introducing the nonlinear equipment health status influencing factor, the nonlinear impact of the maintenance time node on the equipment maintenance cost is effectively considered, which increases the accuracy of the equipment maintenance cost.

[0139] The probability of failure at each maintenance time node is obtained, and the potential future cost is obtained by combining the average cost of historical maintenance costs. The potential future cost is used to represent the loss cost caused by the failure expected to occur due to unmaintained maintenance from the initial time node to the maintenance time node. The calculation formula for the potential future cost is:

[0140]

[0141] in, represents potential future costs, represents the probability of equipment failure at the kth maintenance time node predicted by the fault prediction sub-model, The average cost of historical maintenance costs is represented by the average cost spent in historical maintenance of faults in the maintenance database that are the same or similar to the current fault.

[0142] The estimated total maintenance cost is obtained by taking the weighted sum of the downtime loss cost, equipment maintenance cost and potential future cost.

[0143] Furthermore, the specific steps of the time-priority maintenance strategy include:

[0144] The maintenance method is used as the decision node, the cost factor as the branch condition, and the final total maintenance cost as the leaf node. The decision tree model is constructed through historical data and experience. During the model training process, the goal is to minimize the total cost, and the branches are divided according to the value range and mutual relationship of different cost factors. When encountering a new high-risk equipment maintenance decision, the cost factor data is input into the decision tree model, and the model outputs the optimal maintenance method according to the rules obtained through training.

[0145] Comprehensively collect maintenance case data of special equipment with high risk levels in the past, including data of different types and specifications. Maintenance case data include maintenance methods and cost factor data adopted for each maintenance, such as emergency comprehensive maintenance and key component maintenance. Cost factor data include direct maintenance cost, downtime loss cost, and safety risk cost. Direct maintenance cost refers to maintenance labor costs and parts replacement costs, downtime loss cost refers to production cost loss caused by downtime caused by equipment maintenance, and safety risk cost refers to the estimated risk loss that still exists after maintenance, such as accident loss caused by recurrence of faults in the short term of maintenance. It can be calculated through historical data statistics and risk assessment models, for example, based on the product of the probability of recurrence of faults after maintenance of similar equipment in history and the average loss caused by recurrence of faults;

[0146] Organize the collected data to ensure that the data format is unified and standardized, mark the maintenance method as the category label, and the cost factor as the characteristic data. At the same time, check the completeness and accuracy of the data, and handle missing values ​​and outliers;

[0147] Normalize the cost factors such as direct maintenance cost, downtime loss cost, and safety risk cost, and select the cost factors that have a greater impact on the maintenance method decision as features;

[0148] Initialize the decision tree, take the maintenance mode as the decision node, calculate the information gain or Gini index of each cost factor, select the cost factor with the largest information gain or the smallest Gini index as the optimal partition feature of the current node, and divide the sample set into different sub-nodes according to the value range of the selected partition feature. Each sub-node contains samples of the corresponding interval. For example, if the information gain of the direct maintenance cost is calculated to be the largest, then the direct maintenance cost is used as the first partition feature. If the direct maintenance cost is divided into three intervals of high, medium and low, then the sample set is divided into three sub-nodes.

[0149] Repeat the above process of calculating information gain or Gini index, selecting optimal partitioning features, and partitioning nodes for each child node, and recursively construct subtrees of the decision tree until the stopping condition is met. The stopping condition can be that the number of samples in the node is less than a preset threshold, or the depth of the decision tree reaches a preset value;

[0150] The constructed decision tree is trained using the organized maintenance case data. By continuously adjusting the structure and parameters of the decision tree, the decision tree can accurately predict the corresponding maintenance method based on the cost factor. During the training process, the goal is to minimize the total maintenance cost, that is, when the decision tree predicts the maintenance method, the total cost corresponding to the prediction result should be as close as possible to the actual minimum total cost.

[0151] To prevent the decision tree from overfitting, the trained decision tree is pruned. Pruning is divided into pre-pruning and post-pruning. Pre-pruning is to stop node division when certain conditions are met (such as information gain is less than a certain threshold) during the decision tree construction process; post-pruning is to evaluate the nodes from the leaf nodes step by step after the decision tree is built. If the performance of the model on the validation set (such as accuracy, F1 value, etc.) does not decrease after pruning a node, then the node will be pruned;

[0152] The trained decision tree is deployed in a decision application. When the risk level of special equipment is high, the data of various cost factors of special equipment are collected and preprocessed. The preprocessed data is input into the trained decision tree model. Starting from the root node, the decision tree is traversed downward according to the value of the cost factor until the leaf node is reached. The maintenance method corresponding to the leaf node is the output optimal maintenance method.

[0153] S5. Maintain the special equipment according to the dynamic maintenance strategy, record the actual maintenance cost, and optimize the data model and the dynamic maintenance strategy according to the actual maintenance cost.

[0154] A real-time data feedback mechanism is established to provide timely feedback on the operation data of the equipment after maintenance, the change of risk level, and the actual maintenance cost. Through the analysis of these real-time data, the dynamic maintenance strategy of balancing the risk and cost of special equipment is continuously optimized. For example, if after a certain maintenance, it is found that the actual operation status of the special equipment is better than expected, the risk level has dropped significantly, and the actual maintenance cost is lower than the expected cost, then when formulating the maintenance strategy for similar equipment in the future, this experience can be appropriately referred to to adjust the maintenance time and maintenance content.

[0155] Regularly conduct a comprehensive assessment of the dynamic maintenance strategy for special equipment risk and cost balance, and update and improve the strategy in combination with the latest safety regulations, technical standards and industry best practices. For example, conduct a comprehensive assessment of the maintenance strategy every quarter or every six months, analyze the problems and deficiencies in the implementation of the strategy, and adjust the parameters and rules in the strategy in a timely manner to ensure that the maintenance strategy always adapts to the safe operation and cost management needs of special equipment.

[0156] Example 2

[0157] refer to Figure 3 As shown, this embodiment is the second embodiment of the present invention, which is based on the same inventive concept as the first embodiment. This embodiment introduces a specific implementation method of a dynamic management system for optimizing the life cycle cost of special equipment, including a model building module, a data acquisition module, a fault prediction module, a risk classification module, a dynamic maintenance module, a feedback optimization module, and a maintenance record module, wherein:

[0158] The model building module is used to build a data model of the entire life cycle of special equipment, and the data model includes a fault prediction sub-model and a benefit prediction sub-model;

[0159] The data acquisition module is used to collect the operation data, environmental data, and accumulated production benefits of special equipment in real time and perform preprocessing;

[0160] The fault prediction module is used to input the preprocessed data into the fault prediction sub-model to predict the probability of failure of special equipment and the faulty components;

[0161] The risk classification module is used to classify the risk level of special equipment by building a fuzzy comprehensive evaluation model based on the probability of failure and faulty components, combined with the characteristics of special equipment;

[0162] The dynamic maintenance module is used to formulate a dynamic maintenance strategy for special equipment based on maintenance costs, combining the risk level of special equipment and the predicted benefits of the benefit prediction sub-model;

[0163] The feedback optimization module is used to record the actual maintenance cost and optimize the data model and dynamic maintenance strategy based on the actual maintenance cost.

[0164] The maintenance record module is used to record actual maintenance data, including maintenance time, replacement parts, maintenance results, and actual maintenance costs of this maintenance; so that staff can view the real-time operating status of the equipment, the probability of equipment failure, maintenance plans, and actual maintenance data.

[0165] Working principle and effect:

[0166] By continuously collecting and analyzing the real-time operating status and environmental data of the equipment, and using machine learning technology to build a fault prediction sub-model for the equipment, the probability of equipment failure and the faulty components are predicted, so that equipment managers can foresee potential problems before failures occur and take preventive maintenance measures, significantly reducing downtime caused by sudden failures and ensuring the continuity and efficiency of equipment operation;

[0167] Different dynamic management strategies are formulated based on the probability of equipment failure and the risk level of special equipment. The optimal maintenance time of the equipment is dynamically determined through the total maintenance cost. This strategy can avoid unnecessary early maintenance and reduce maintenance costs. It can also prevent high repair costs caused by excessive wear of equipment and reduce production loss costs caused by equipment downtime. By accurately predicting and reasonably arranging maintenance time, the optimal allocation of maintenance resources is achieved, maintenance efficiency is improved, the service life of equipment is extended, and the full life cycle cost of equipment is controlled from the source.

[0168] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0169] The above preset parameters or preset thresholds are all set by technicians in this field according to actual conditions or obtained through large-scale data simulation.

[0170] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the scope of protection of the purpose of the present invention and the claims, and all of these are within the protection of the present invention.

Claims

1. A dynamic management method for optimizing the life cycle cost of special equipment, characterized by: The following steps are involved: Constructing a data model for the entire life cycle of special equipment, wherein the data model includes a fault prediction sub-model and a benefit prediction sub-model; Collect the operation data and environmental data of special equipment, and input them into the fault prediction sub-model after pre-processing to predict the failure probability and faulty components of special equipment; Based on the probability of failure and faulty components, combined with the characteristics of special equipment, a fuzzy comprehensive evaluation model is constructed to classify the risk level of special equipment; Combining the risk level of special equipment and the predicted benefits of the benefit prediction sub-model, a dynamic maintenance strategy for special equipment is formulated based on maintenance costs; The special equipment is maintained according to the dynamic maintenance strategy, the actual maintenance cost is recorded, and the data model and the dynamic maintenance strategy are optimized according to the actual maintenance cost.

2. A dynamic management method for optimizing the life cycle cost of special equipment according to claim 1, characterized in that: The construction of the fuzzy comprehensive evaluation model to classify the risk level of special equipment specifically includes: Obtain the probability of failure and the faulty components, as well as the operating data, select evaluation factors, and determine the evaluation factor set; The risk level of special equipment is divided into three levels, namely normal operation, low risk and high risk, to construct an evaluation level set; Compare the relative importance of each evaluation factor in the evaluation factor set, construct a judgment matrix, calculate the eigenvector and maximum eigenvalue in the judgment matrix, and obtain the weight of each evaluation factor; Determine the degree of membership of each evaluation factor to different risk levels and establish a fuzzy relationship matrix; Through fuzzy synthesis operation, the comprehensive evaluation result vector is obtained, and the risk level of special equipment is determined according to the maximum membership principle.

3. A dynamic management method for optimizing the life cycle cost of special equipment according to claim 2, characterized in that: The dynamic maintenance strategy includes a cost-first maintenance strategy and a time-first maintenance strategy. The dynamic maintenance strategy for special equipment formulated based on maintenance cost specifically includes: Configure the fault probability threshold. When the fault probability of special equipment is less than or equal to the fault probability threshold or the risk level is normal operation, no dynamic maintenance strategy is formulated. When the failure probability of special equipment is greater than the failure probability threshold and the risk level is normal operation or low risk, formulate a cost-priority maintenance strategy; When the risk level of special equipment is high, formulate a time-priority maintenance strategy.

4. A dynamic management method for optimizing the life cycle cost of special equipment according to claim 3, characterized in that: The specific steps of formulating the cost-priority maintenance strategy include: S41: When the probability of a fault occurring is greater than a threshold value of the probability of a fault occurring, a warning signal is triggered to provide a reminder for equipment maintenance; S42: taking the time of receiving the equipment maintenance reminder as the start time node of the equipment maintenance process, taking the latest maintenance time point as the end time node, and dividing by a preset first fixed time length as an interval to form a first maintenance time node sequence; S43: Calculate the total maintenance cost that is expected to be generated when each maintenance time node in the first maintenance time node sequence starts to perform maintenance work, and associate the total maintenance cost with the maintenance time node to form a first maintenance total cost set; S44: configuring a total cost threshold, and adding the maintenance time nodes corresponding to the maintenance total cost in the first maintenance total cost set that are less than the total cost threshold to the time node set; S45: judging and analyzing the time node set to form a second maintenance time node sequence, and calculating to obtain a second maintenance total cost set; S46: sorting the total maintenance costs in the first total maintenance cost set and the second total maintenance cost set to form a maintenance cost sequence from low to high; S47: Selecting the maintenance time node corresponding to the lowest total maintenance cost from the maintenance cost sequence as the optimal equipment maintenance time.

5. A dynamic management method for optimizing the life cycle cost of special equipment according to claim 4, characterized in that: The judging and analyzing of the time node set to form a second maintenance time node sequence and calculating the second maintenance total cost set include: Determine whether there are adjacent maintenance time nodes in the time node set. If there are adjacent maintenance time nodes, merge the adjacent maintenance time nodes. The earliest maintenance time node among the adjacent maintenance time nodes is used as the secondary judgment start time node, the latest maintenance time node among the adjacent maintenance time nodes is used as the secondary judgment end time node, and the time length between the secondary judgment start time node and the secondary judgment end time node is used as the secondary judgment maintenance time period; Divide the secondary judgment maintenance time period into intervals of a second fixed time length to form a second maintenance time node sequence, calculate the total maintenance cost that is expected to be generated when each maintenance time node in the second maintenance time node sequence starts to perform maintenance work, and associate the total maintenance cost with the maintenance time node to form a second maintenance total cost set; If there are no adjacent maintenance time nodes, the second maintenance total cost set is defined as an empty set.

6. A dynamic management method for optimizing the life cycle cost of special equipment according to claim 4, characterized in that: The calculation of the total maintenance cost expected to be incurred when the maintenance work begins includes: Obtain the downtime loss costs caused by downtime during maintenance through the accumulated production benefits of special equipment; The equipment maintenance cost of special equipment in the expected maintenance process is calculated by labor costs, replacement costs of equipment parts and nonlinear equipment health status influencing factors; among which, labor costs are the product of unit labor hour cost and labor hours spent; Obtain the probability of failure at each maintenance time node, and combine it with the average cost of historical maintenance costs to obtain potential future costs, where the potential future costs are used to represent the loss costs caused by failures that are expected to occur due to unmaintained maintenance from the initial time node to the maintenance time node; The estimated total maintenance cost is obtained by taking the weighted sum of the downtime loss cost, equipment maintenance cost and potential future cost.

7. A dynamic management method for optimizing the life cycle cost of special equipment according to claim 6, characterized in that: The acquisition of downtime loss costs caused by downtime during the maintenance process specifically includes: Obtain the cumulative production benefit value corresponding to the initial time node; The benefit prediction sub-model is used to predict the cumulative production benefit value corresponding to each maintenance time node; The downtime loss cost is determined by the cumulative benefit difference between the cumulative production benefit value corresponding to the initial time node and the cumulative production benefit value corresponding to the maintenance time node, the time length between the initial time node and the maintenance time node, and the expected downtime duration.

8. A dynamic management method for optimizing the life cycle cost of special equipment according to claim 3, characterized in that: The specific steps of formulating the time-priority maintenance strategy include: The maintenance method is used as the decision node, the cost factor is used as the branch condition, and the final total maintenance cost is used as the leaf node; the maintenance method includes comprehensive maintenance and key component maintenance, and the cost factor data includes direct maintenance cost, downtime loss cost, and safety risk cost; With the goal of minimizing the total maintenance cost, the decision tree model is trained by dividing the branches according to the value range of different cost factors; When the risk level of special equipment is high risk, the cost factor data of special equipment are collected and preprocessed. The preprocessed cost factor data are input into the trained decision tree model. Starting from the root node, the decision tree is traversed downward according to the value of the cost factor until the leaf node is reached. The maintenance method corresponding to the leaf node is the output optimal maintenance method.

9. The dynamic management method for optimizing the life cycle cost of special equipment according to claim 1 is characterized in that: The fault prediction sub-model includes: an input layer, an LSTM layer, an attention layer, a fully connected layer and an output layer; The input layer is used to input preprocessed data; The LSTM layer contains a hidden layer of multiple LSTM units to capture long-term dependencies in dynamic data; The attention layer is used to calculate the attention weights for the output sequence of the LSTM layer and obtain the attention output through weighted summation; The fully connected layer integrates the features by connecting the attention output of the attention layer; The output layer uses the Softmax function to predict the probability of failure and the failure of special equipment components.

10. A dynamic management system for optimizing the full life cycle cost of special equipment, used to implement a dynamic management method for optimizing the full life cycle cost of special equipment as described in any one of claims 1 to 9, characterized in that: It includes model building module, data acquisition module, fault prediction module, risk classification module and dynamic maintenance module, among which: The model building module is used to build a data model of the entire life cycle of special equipment, and the data model includes a fault prediction sub-model and a benefit prediction sub-model; The data acquisition module is used to collect the operation data, environmental data, and accumulated production benefits of special equipment in real time and perform preprocessing; The fault prediction module is used to predict the probability of failure of special equipment and the faulty components; The risk classification module is used to classify the risk level of special equipment by building a fuzzy comprehensive evaluation model based on the probability of failure and faulty components, combined with the characteristics of special equipment; The dynamic maintenance module is used to formulate a dynamic maintenance strategy for special equipment based on maintenance costs, combining the risk level of special equipment and the predicted benefits of the benefit prediction sub-model; The feedback optimization module is used to record the actual maintenance cost and optimize the data model and dynamic maintenance strategy based on the actual maintenance cost.

Images