Electromechanical equipment maintenance prediction system based on big data

Abstract

The invention discloses an electromechanical equipment maintenance prediction system based on big data, and relates to the technical field of electromechanical equipment maintenance, and the system comprises the steps: collecting a historical data set, building a fault time prediction model, obtaining the historical data corresponding to a fault time point and a normal state, outputting a first parameter signal, and building an optimization objective function and a constraint condition. And calculating a future time step length and a control input parameter, outputting a second parameter signal, sending the first parameter signal and the second parameter signal to a control end, and carrying out early warning and regulation and control on the electromechanical equipment. A fault time prediction model is established through a forward-backward algorithm, accurate prediction is realized, for equipment with insufficient data, the potential fault trend is speculated based on trend prediction of a dynamic model, collaborative maintenance prediction of equipment with various data volumes is realized, and the fault prediction precision and adaptability of the equipment under different data scales are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electromechanical equipment maintenance, and in particular to an electromechanical equipment maintenance prediction system based on big data. Background Art

[0002] With the continuous improvement of the level of industrial automation, the demand for intelligent operation and maintenance of electromechanical equipment is increasing. The traditional maintenance model is usually based on regular maintenance strategies or experience-based judgments, but this method is prone to excessive or insufficient maintenance and cannot achieve early prediction of faults. In recent years, the rapid development of big data, the Internet of Things and artificial intelligence technologies has provided technical support for the preventive maintenance of equipment. Equipment failure prediction technology based on big data can collect and analyze equipment operation data, use prediction models to identify potential failure risks in advance, and provide a scientific basis for equipment operation and maintenance.

[0003] At present, a Chinese invention patent with application number CN202311355664.5 discloses a mechanical and electrical equipment detection and maintenance system, including a main control terminal, to which are connected: a sensor network unit for collecting operating data of the mechanical and electrical equipment; a data processing unit connected to the sensor network unit, for receiving and processing data collected by the sensor network; an intelligent prediction unit connected to the data processing unit, and by analyzing historical data and environmental factors, predicting possible future failures of the mechanical and electrical equipment, and generating maintenance suggestions in advance; an adaptive learning unit, connected in parallel with the data processing unit and the intelligent prediction unit; the present invention can help users take preventive maintenance measures, reduce equipment failures and downtime, and improve the reliability and availability of equipment.

[0004] The above technology is difficult to perform adaptive fault prediction based on different amounts of historical data, which affects the maintenance prediction effect when electromechanical equipment with different historical data collection times operate together. Summary of the invention

[0005] The technical problem solved by the present invention is that the above technology is difficult to perform adaptive fault prediction based on different amounts of historical data, which affects the maintenance prediction effect when electromechanical equipment with different historical data collection times operate together.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] The electromechanical equipment maintenance prediction system based on big data includes data partitioning module, model building module, trend prediction module and decision-making and control module:

[0008] The data partitioning module is used to collect historical data sets, analyze the historical data sets according to a preset historical data number threshold, and output a selection signal according to the analysis result, wherein the selection signal includes a rich data signal and a simple data signal;

[0009] The model building module is used to receive a selection signal. If the selection signal is a rich data signal, the historical data set of the data partitioning module is obtained, the historical data set is preprocessed, the model parameters are set, and a failure time prediction model is established according to a forward-backward algorithm. The input data of the failure time prediction model is the current parameter data, and the output data is the failure time point. The current parameter data is obtained and used as the input data of the failure time prediction model, and the historical data corresponding to the failure time point and the normal state are obtained, and the output is a first parameter signal;

[0010] The trend prediction module is used to receive a selection signal. If the selection signal is a simple data signal, a dynamic model is established, a time interval is set, a state change is predicted according to the dynamic model, an optimization objective function and constraint conditions are established, current parameter data is input into the objective function, future time steps and control input parameters are calculated, and the output is a second parameter signal;

[0011] The decision-making and control module is used to send the first parameter signal and the second parameter signal to the control end to perform early warning and control on the electromechanical equipment.

[0012] Preferably, the data partitioning module includes a data acquisition unit and a data analysis unit;

[0013] The data acquisition unit is used to collect historical data sets, which include temperature data, vibration data, current data, voltage data, speed data, torque data, pressure data and corresponding real-time data under normal operating conditions, temperature data, vibration data, current data, voltage data, speed data, torque data, pressure data and corresponding real-time data under abnormal operating conditions, and initial time data of the device at an initial moment;

[0014] The data analysis unit is used to analyze the historical data according to the historical data number threshold;

[0015] If the number of data in the historical data set is greater than the historical data number threshold, a rich data signal is output;

[0016] If the number of data in the historical data set is less than the historical data number threshold, a simple data signal is output;

[0017] The rich data signal and the simple data signal are combined and output as a selection signal.

[0018] Preferably, the model building module includes a first signal receiving unit, a data processing unit, a parameter setting unit, a first model building unit and a first output processing unit;

[0019] The first signal receiving unit is used to receive a selection signal, and if the selection signal is a rich data signal, trigger the data processing unit, the parameter setting unit, the model building unit and the output processing unit;

[0020] The data processing unit is used to preprocess the historical data set, and the preprocessing includes removing outliers and reducing noise. When removing outliers, if any data in the historical data set is removed, all data in the same instant time data and the same initial time data as the data will be removed at the same time.

[0021] Preferably, the parameter setting unit is used to set model parameters, and the model parameters include:

[0022] State parameters, the state parameters including normal operating state and fault operating state;

[0023] The number of observation values, the number of observation values ​​is all parameter data sets of a normal operating state and a fault operating state, the parameter data set of the normal operating state includes a normal temperature data set, a normal vibration data set, a normal current data set, a normal voltage data set, a normal speed data set, a normal torque data set and a normal pressure data set, and the parameter data set of the fault operating state includes a fault temperature data set, a fault vibration data set, a fault current data set, a fault voltage data set, a fault speed data set, a fault torque data set and a fault pressure data set;

[0024] Initial state distribution, where the initial state distribution is the probability of the electromechanical equipment being in a normal operating state and a faulty operating state at the initial time data in the historical data set;

[0025] A transition probability matrix, wherein the transition probability matrix is ​​the probability of the electromechanical equipment transitioning between a normal operating state and a faulty operating state;

[0026] The observation probability matrix is ​​the probability that the current parameter data collected is in a certain data set in the parameter data set when the state parameter is in a normal operation state or a fault operation state.

[0027] Preferably, the first model building unit is used to build a failure time prediction model using a forward-backward algorithm:

[0028] After initializing the transfer probability matrix and the observation probability matrix, the forward-backward algorithm is used to continuously update the transfer probability matrix, the observation probability matrix and the initial state distribution based on the historical data set, and the log likelihood is calculated. When the log likelihood is less than the preset log likelihood threshold, convergence is stopped and the output is the failure time prediction model.

[0029] The first output processing unit is used to input the current parameter data into the fault time prediction model, obtain the state parameters and transition probability matrix corresponding to the current parameter data, and if the probability of transitioning from the normal operation state to the fault operation state in the transition probability matrix is ​​greater than a preset probability threshold, then obtain the time length for the current state parameters to transfer to the fault operation state and the historical data corresponding to the normal operation state closest to the fault operation state, calculate the fault time point based on the time length, and output the fault time point and the historical data corresponding to the normal state as the first parameter signal.

[0030] Preferably, the trend prediction module includes a second signal receiving unit, a second model building unit and a second output processing unit;

[0031] The second signal receiving unit is used to receive a selection signal, and if the selection signal is a simple data signal, the second model building unit and the second output processing unit are triggered;

[0032] The second model building unit is used to build a kinetic model, and the kinetic model is:

[0033] x k+1 =Ax k +Bu k

[0034] Among them, x is parameter data, u is control input data, x k+1 is the parameter data at the k+1th moment, the parameter data including temperature data, vibration data, current data, voltage data, speed data, torque data and pressure data, x k is the parameter data at the kth moment, u k is the control input data at the kth moment, A is the state matrix, and B is the input matrix;

[0035] Set the time interval n, and predict the state change within a time interval according to the kinetic model. The mathematical expression is:

[0036] x n =Ax n-1 +Bu n-1

[0037] Among them, x n is the parameter data at the nth moment, x n-1 is the parameter data at the n-1th moment, u n-1 is the control input data at the n-1th moment, n=1,2,3,,,n;

[0038] The output is a state sequence, the state sequence is x1,x2,x3,,,x n ;

[0039] Setting a constraint condition, wherein the constraint condition is a constraint range of each data in the parameter data;

[0040] Establish the optimization objective function J and find the input sequence u0,u1,,,u when the objective function is the smallest n , the function expression is:

[0041]

[0042] Among them, x ref is the target state, the target state is the parameter data at the preset standard parameter threshold, R is the input weight matrix, Q is the error weight matrix, and T is the predicted future time step.

[0043] Preferably, the second output processing unit is used to input the current parameter data into the function expression when the objective function is minimized, obtain the future time step and the input sequence, output the input sequence as a control input parameter, calculate the fault time point according to the future time step, and output the fault time point and the control input parameter as a second parameter signal.

[0044] Preferably, the decision-making and control module includes an early warning processing unit and an equipment control unit;

[0045] The warning processing unit is used to receive the first parameter signal and the second parameter signal, extract the fault time point in the first parameter signal and the fault time point in the second parameter signal, and output the extracted fault time point as a warning signal to send to the warning terminal;

[0046] The equipment control unit is used to extract historical data from the first parameter signal and control input parameters from the second parameter signal, and output the extracted historical data and control input parameters as a target control signal to be sent to the electromechanical equipment control terminal.

[0047] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the electromechanical equipment maintenance prediction system based on big data is implemented.

[0048] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the electromechanical equipment maintenance prediction system based on big data.

[0049] Beneficial effects of the present invention: The present invention adaptively selects fault prediction methods according to the different amounts of historical data. For equipment with rich data, a fault time prediction model is established through a forward-backward algorithm to achieve accurate prediction. For equipment with insufficient data, trend prediction based on a dynamic model is used to infer potential fault trends, thereby achieving collaborative maintenance prediction for equipment with various amounts of data, effectively improving the fault prediction accuracy and adaptability of equipment under different data scales. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A schematic diagram of the basic flow of a big data-based electromechanical equipment maintenance prediction system provided for one embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0052] Example, see Figure 1 , provides a big data-based electromechanical equipment maintenance prediction system, including data partitioning module, model building module, trend prediction module and decision-making and control module:

[0053] The data partitioning module is used to collect historical data sets, analyze the historical data sets according to a preset historical data number threshold, and output a selection signal according to the analysis result, wherein the selection signal includes a rich data signal and a simple data signal.

[0054] The model building module is used to receive a selection signal. If the selection signal is a rich data signal, the historical data set of the data partitioning module is obtained, the historical data set is preprocessed, the model parameters are set, and a fault time prediction model is established according to the forward-backward algorithm. The input data of the fault time prediction model is the current parameter data, and the output data is the fault time point. The current parameter data is obtained and used as the input data of the fault time prediction model, and the historical data corresponding to the fault time point and the normal state are obtained, and the output is the first parameter signal.

[0055] The trend prediction module is used to receive the selection signal. If the selection signal is a simple data signal, a dynamic model is established, the time interval is set, the state change is predicted according to the dynamic model, the optimization objective function and constraint conditions are established, the current parameter data is input into the objective function, the future time step and control input parameters are calculated, and the output is the second parameter signal.

[0056] The decision-making and control module is used to send the first parameter signal and the second parameter signal to the control end to provide early warning and control for the electromechanical equipment.

[0057] The data partitioning module includes a data acquisition unit and a data analysis unit.

[0058] The data acquisition unit is used to collect historical data sets, which include temperature data, vibration data, current data, voltage data, speed data, torque data, pressure data and corresponding real-time data under normal operating conditions, temperature data, vibration data, current data, voltage data, speed data, torque data, pressure data and corresponding real-time data under abnormal operating conditions, and initial time data of the equipment at an initial moment.

[0059] The data acquisition unit is responsible for acquiring historical data of the equipment under different conditions, providing basic information for subsequent analysis and prediction models, and ensuring the acquisition of comprehensive operating status information.

[0060] The data analysis unit is used to analyze the historical data according to the historical data quantity threshold.

[0061] If the number of data in the historical data set is greater than the historical data number threshold, a rich data signal is output.

[0062] If the number of data in the historical data set is less than the historical data number threshold, a simple data signal is output.

[0063] The rich data signal and the simple data signal are combined and output as a selection signal.

[0064] The data analysis unit is used to evaluate whether the amount of historical data reaches the set threshold. When the amount of data exceeds the threshold, it outputs a "rich data signal", and when it is insufficient, it outputs a "simple data signal". Through this analysis, the optimal prediction model is selected to provide effective prediction basis for devices with different data scales.

[0065] The main effect of the data partitioning module is to classify the quantity and quality of the equipment historical data set, determine the subsequent prediction strategy by analyzing the richness of the data volume, and provide adaptive fault prediction solutions for data-rich and data-limited equipment, thereby improving the prediction accuracy and adaptability under different data scales.

[0066] The model building module includes a first signal receiving unit, a data processing unit, a parameter setting unit, a first model building unit and a first output processing unit.

[0067] The first signal receiving unit is used to receive a selection signal. If the selection signal is a rich data signal, the data processing unit, the parameter setting unit, the model building unit and the output processing unit are triggered.

[0068] The first signal receiving unit is used to receive the selection signal and activate the subsequent units upon receiving the "enriched data signal" to prepare for the subsequent model construction.

[0069] The data processing unit is used to preprocess the historical data set, and the preprocessing includes removing outliers and reducing noise. When removing outliers, if any data in the historical data set is removed, all data in the same instant time data and the same initial time data as the data will be removed at the same time.

[0070] The data processing unit preprocesses the historical data set, including removing outliers and reducing noise. When removing abnormal data, ensure that all instant time and initial time data associated with the data are also removed, so as to ensure data consistency and improve the accuracy of the prediction model.

[0071] The parameter setting unit is used to set the model parameters, which include:

[0072] Status parameters, which include normal operation status and fault operation status.

[0073] The number of observation values, the number of observation values ​​is all parameter data sets of normal operating state and fault operating state, the parameter data set of normal operating state includes normal temperature data set, normal vibration data set, normal current data set, normal voltage data set, normal speed data set, normal torque data set and normal pressure data set, the parameter data set of fault operating state includes fault temperature data set, fault vibration data set, fault current data set, fault voltage data set, fault speed data set, fault torque data set and fault pressure data set.

[0074] Initial state distribution: The initial state distribution is the probability of the electromechanical equipment being in the normal operation state and the faulty operation state at the initial time data in the historical data set.

[0075] The transition probability matrix is ​​the probability of the electromechanical equipment transferring between the normal operating state and the fault operating state.

[0076] The observation probability matrix is ​​the probability that the current parameter data collected is in a certain data set in the parameter data set when the state parameter is in a normal operation state or a fault operation state.

[0077] The parameter setting unit sets the key parameters of the prediction model, including state parameters, number of observations, initial state distribution, transition probability matrix and observation probability matrix, providing a statistical basis for the establishment and operation of the model and ensuring that the model can accurately simulate equipment state transitions and observation patterns.

[0078] The first model building unit is used to build a failure time prediction model using a forward-backward algorithm:

[0079] After initializing the transfer probability matrix and the observation probability matrix, the forward-backward algorithm is used to continuously update the transfer probability matrix, the observation probability matrix and the initial state distribution based on the historical data set, and the log likelihood is calculated. When the log likelihood is less than the preset log likelihood threshold, the convergence is stopped and the output is the failure time prediction model.

[0080] The first model building unit builds a failure time prediction model based on the forward-backward algorithm, initializes and continuously updates the transition probability matrix, observation probability matrix and initial state distribution, iteratively calculates the log-likelihood value and stops converging when the preset threshold is met, thereby generating an accurate failure time prediction model.

[0081] The first output processing unit is used to input the current parameter data into the fault time prediction model, obtain the state parameters and transition probability matrix corresponding to the current parameter data, and if the probability of transitioning from the normal operation state to the fault operation state in the transition probability matrix is ​​greater than a preset probability threshold, then obtain the time length for the current state parameters to transfer to the fault operation state and the historical data corresponding to the normal operation state closest to the fault operation state, calculate the fault time point according to the time length, and output the fault time point and the historical data corresponding to the normal state as the first parameter signal.

[0082] The first output processing unit inputs the current parameter data into the fault time prediction model to obtain the corresponding state and transition probability matrix. If the probability of transitioning to the fault state is higher than the threshold, the expected fault time point is calculated and the first parameter signal is output to provide a fault prediction basis for the early warning processing unit.

[0083] The effect of the model building module is to establish an accurate failure time prediction model based on rich data to realize the prediction of future failure time points of the equipment. When receiving the "rich data signal", it will automatically start the data processing unit, parameter setting unit, first model building unit and first output processing unit, and finally output the equipment failure time prediction signal to provide a basis for fault warning and maintenance regulation.

[0084] The trend prediction module includes a second signal receiving unit, a second model building unit and a second output processing unit.

[0085] The second signal receiving unit is used for receiving a selection signal, and if the selection signal is a simple data signal, the second model building unit and the second output processing unit are triggered.

[0086] The second signal receiving unit receives the selection signal, and triggers the second model building unit and the second output processing unit upon receiving the "simple data signal", thereby entering the trend prediction mode.

[0087] The second model building unit is used to build a kinetic model, and the kinetic model is:

[0088] x k+1 =Ax k +Bu k

[0089] Among them, x is parameter data, u is control input data, x k+1 is the parameter data at the k+1th moment, the parameter data including temperature data, vibration data, current data, voltage data, speed data, torque data and pressure data, x k is the parameter data at the kth moment, u k is the control input data at the kth moment, A is the state matrix, and B is the input matrix.

[0090] Set the time interval n, and predict the state change within a time interval according to the kinetic model. The mathematical expression is:

[0091] x n =Ax n-1 +Bu n-1

[0092] Among them, x n is the parameter data at the nth moment, x n-1 is the parameter data at the n-1th moment, u n-1 is the control input data at the n-1th moment, n=1,2,3,,,n.

[0093] The output is a state sequence, the state sequence is x1,x2,x3,,,x n .

[0094] Set a constraint condition, where the constraint condition is a constraint range for each data in the parameter data.

[0095] Establish the optimization objective function J and find the input sequence u0,u1,,,u when the objective function is the smallest n , the function expression is:

[0096]

[0097] Among them, x ref is the target state, the target state is the parameter data at the preset standard parameter threshold, R is the input weight matrix, Q is the error weight matrix, and T is the predicted future time step.

[0098] The second model building unit constructs a dynamic model, predicts the state changes within a time interval by setting the time interval, outputs the state sequence, sets parameter constraints, limits the data range, and constructs an optimization objective function to seek the input sequence when the objective function is the smallest, so as to provide the most reasonable trend prediction model.

[0099] The second output processing unit is used to input the current parameter data into the function expression when the objective function is minimized, obtain the future time step and the input sequence, output the input sequence as a control input parameter, calculate the fault time point according to the future time step, and output the fault time point and the control input parameter as a second parameter signal.

[0100] The second output processing unit inputs the current parameter data into the optimal expression of the objective function, obtains the future time step and input sequence, and outputs the control input parameters. It calculates the fault time point according to the time step, and outputs the fault time point and control input parameters to ensure the maintenance prediction effect of the equipment under simple data conditions.

[0101] The trend prediction module uses a dynamic model to perform status prediction based on a small amount of data. It is suitable for equipment that lacks rich historical data. When receiving a "simple data signal", it builds a simple time prediction model to calculate the status changes of the equipment in future time intervals and determine the predicted failure time point and control input parameters to provide real-time reference for maintenance and operation.

[0102] The decision-making and control module includes an early warning processing unit and an equipment control unit;

[0103] The warning processing unit is used to receive the first parameter signal and the second parameter signal, extract the fault time point in the first parameter signal and the fault time point in the second parameter signal, and output the extracted fault time point as a warning signal to be sent to the warning end.

[0104] The early warning processing unit receives the first parameter signal and the second parameter signal, extracts the fault time point information therein, generates an early warning signal and sends it to the early warning terminal for early intervention.

[0105] The equipment control unit is used to extract the historical data in the first parameter signal and the control input parameters in the second parameter signal, and output the extracted historical data and control input parameters as a target control signal to be sent to the electromechanical equipment control terminal.

[0106] The equipment control unit extracts historical data from the first parameter signal, extracts control input parameters from the second parameter signal, converts the extracted information into a target control signal and transmits it to the equipment control terminal, thereby realizing automatic control of the equipment status and improving the safety and operation efficiency of the equipment.

[0107] The decision-making and control module is used to receive and process fault prediction information, combine early warning signals with target control signals, coordinate maintenance and equipment control, and realize early warning and control of equipment status by analyzing the fault time point and equipment control parameters from the first and second parameter signals, ensuring that electromechanical equipment receives timely response in abnormal conditions.

[0108] The present invention can flexibly adjust the prediction method according to the richness of historical data, determine whether the data meets the set richness threshold through the "data partitioning module", and output a "rich data signal" or a "simple data signal", and automatically select the appropriate prediction strategy according to different signals, thereby realizing adaptive prediction of the operating status of equipment with different amounts of historical data, and effectively improving the scope of application of the prediction.

[0109] When the equipment has sufficient historical data, accurate fault time prediction can be achieved by constructing a forward-backward algorithm fault time prediction model. By utilizing rich parameter data and continuously optimizing the state transition and observation probability matrix, it ensures that the prediction model can accurately capture the potential trend of faults and provide more accurate fault prediction for equipment with richer data.

[0110] For equipment with less historical data, the trend prediction module based on the dynamic model is automatically enabled to achieve trend prediction within a shorter time series by optimizing the objective function and constraints. In the case of incomplete data, the fault trend is reasonably inferred based on the law of equipment status changes, thereby ensuring the maintenance prediction effect of equipment with insufficient data.

[0111] The multi-mode prediction method provides a unified fault prediction framework for devices with different historical data collection times. For devices with sufficient data, accurate warning information is obtained through the fault time prediction model, while for devices with limited data, potential risks are predicted through the trend prediction model. Both prediction results are uniformly processed and output by the "decision-making and control module", which improves the prediction consistency of multiple devices running together and makes the equipment status assessment more comprehensive and reliable.

[0112] On the same platform, differentiated predictions are provided for devices with different data volumes, and their early warning signals and control parameters are summarized to achieve collaborative operations of early warning and maintenance optimization. This not only ensures the complementarity of operation and maintenance of each device, but also provides adaptive maintenance solutions when the equipment status varies greatly, further improving the overall operation efficiency and fault prediction effect of the equipment group.

[0113] It should be understood by those skilled in the art that the embodiments of the present invention can be provided as methods, systems or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, disk or optical disk. These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0114] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. The electromechanical equipment maintenance prediction system based on big data is characterized by: Including data partitioning module, model building module, trend prediction module and decision-making and control module: The data partitioning module is used to collect historical data sets, analyze the historical data sets according to a preset historical data number threshold, and output a selection signal according to the analysis result, wherein the selection signal includes a rich data signal and a simple data signal; The model building module is used to receive a selection signal. If the selection signal is a rich data signal, the historical data set of the data partitioning module is obtained, the historical data set is preprocessed, the model parameters are set, and a failure time prediction model is established according to a forward-backward algorithm. The input data of the failure time prediction model is the current parameter data, and the output data is the failure time point. The current parameter data is obtained and used as the input data of the failure time prediction model, and the historical data corresponding to the failure time point and the normal state are obtained, and the output is a first parameter signal; The trend prediction module is used to receive a selection signal. If the selection signal is a simple data signal, a dynamic model is established, a time interval is set, a state change is predicted according to the dynamic model, an optimization objective function and constraint conditions are established, current parameter data is input into the objective function, future time steps and control input parameters are calculated, and the output is a second parameter signal; The decision-making and control module is used to send the first parameter signal and the second parameter signal to the control end to perform early warning and control on the electromechanical equipment.

2. The electromechanical equipment maintenance prediction system based on big data according to claim 1, characterized in that: The data partitioning module includes a data acquisition unit and a data analysis unit; The data acquisition unit is used to collect historical data sets, which include temperature data, vibration data, current data, voltage data, speed data, torque data, pressure data and corresponding real-time data under normal operating conditions, temperature data, vibration data, current data, voltage data, speed data, torque data, pressure data and corresponding real-time data under abnormal operating conditions, and initial time data of the device at an initial moment; The data analysis unit is used to analyze the historical data according to the historical data number threshold; If the number of data in the historical data set is greater than the historical data number threshold, a rich data signal is output; If the number of data in the historical data set is less than the historical data number threshold, a simple data signal is output; The rich data signal and the simple data signal are combined and output as a selection signal.

3. The electromechanical equipment maintenance prediction system based on big data according to claim 2, characterized in that: The model building module includes a first signal receiving unit, a data processing unit, a parameter setting unit, a first model building unit and a first output processing unit; The first signal receiving unit is used to receive a selection signal, and if the selection signal is a rich data signal, trigger the data processing unit, the parameter setting unit, the model building unit and the output processing unit; The data processing unit is used to preprocess the historical data set, and the preprocessing includes removing outliers and reducing noise. When removing outliers, if any data in the historical data set is removed, all data in the same instant time data and the same initial time data as the data will be removed at the same time.

4. The electromechanical equipment maintenance prediction system based on big data according to claim 3 is characterized in that: The parameter setting unit is used to set model parameters, and the model parameters include: State parameters, the state parameters including normal operating state and fault operating state; The number of observation values, the number of observation values ​​is all parameter data sets of a normal operating state and a fault operating state, the parameter data set of the normal operating state includes a normal temperature data set, a normal vibration data set, a normal current data set, a normal voltage data set, a normal speed data set, a normal torque data set and a normal pressure data set, and the parameter data set of the fault operating state includes a fault temperature data set, a fault vibration data set, a fault current data set, a fault voltage data set, a fault speed data set, a fault torque data set and a fault pressure data set; Initial state distribution, where the initial state distribution is the probability of the electromechanical equipment being in a normal operating state and a faulty operating state at the initial time data in the historical data set; A transition probability matrix, wherein the transition probability matrix is ​​the probability of the electromechanical equipment transitioning between a normal operating state and a faulty operating state; The observation probability matrix is ​​the probability that the current parameter data collected is in a certain data set in the parameter data set when the state parameter is in a normal operation state or a fault operation state.

5. The electromechanical equipment maintenance prediction system based on big data according to claim 4, characterized in that: The first model building unit is used to build a failure time prediction model using a forward-backward algorithm: After initializing the transfer probability matrix and the observation probability matrix, the forward-backward algorithm is used to continuously update the transfer probability matrix, the observation probability matrix and the initial state distribution based on the historical data set, and the log likelihood is calculated. When the log likelihood is less than the preset log likelihood threshold, convergence is stopped and the output is the failure time prediction model. The first output processing unit is used to input the current parameter data into the fault time prediction model, obtain the state parameters and transition probability matrix corresponding to the current parameter data, and if the probability of transitioning from the normal operation state to the fault operation state in the transition probability matrix is ​​greater than a preset probability threshold, then obtain the time length for the current state parameters to transfer to the fault operation state and the historical data corresponding to the normal operation state closest to the fault operation state, calculate the fault time point based on the time length, and output the fault time point and the historical data corresponding to the normal state as the first parameter signal.

6. The electromechanical equipment maintenance prediction system based on big data according to claim 5, characterized in that: The trend prediction module includes a second signal receiving unit, a second model building unit and a second output processing unit; The second signal receiving unit is used to receive a selection signal, and if the selection signal is a simple data signal, the second model building unit and the second output processing unit are triggered; The second model building unit is used to build a kinetic model, and the kinetic model is: x k+1 =Ax k +Bu k Among them, x is parameter data, u is control input data, x k+1 is the parameter data at the k+1th moment, the parameter data including temperature data, vibration data, current data, voltage data, speed data, torque data and pressure data, x k is the parameter data at the kth moment, u k is the control input data at the kth moment, A is the state matrix, and B is the input matrix; Set the time interval n, and predict the state change within a time interval according to the kinetic model. The mathematical expression is: x n =Ax n-1 +Bu n-1 Among them, x n is the parameter data at the nth moment, x n-1 is the parameter data at the n-1th moment, u n-1 is the control input data at the n-1th moment, n=1,2,3,,,n; The output is a state sequence, the state sequence is x1,x2,x3,,,x n ; Setting a constraint condition, wherein the constraint condition is a constraint range of each data in the parameter data; Establish the optimization objective function J and find the input sequence u0,u1,,,u when the objective function is the smallest n , the function expression is: Among them, x ref is the target state, the target state is the parameter data at the preset standard parameter threshold, R is the input weight matrix, Q is the error weight matrix, and T is the predicted future time step.

7. The electromechanical equipment maintenance prediction system based on big data according to claim 6, characterized in that: The second output processing unit is used to input the current parameter data into the function expression when the objective function is minimized, obtain the future time step and the input sequence, output the input sequence as a control input parameter, calculate the fault time point according to the future time step, and output the fault time point and the control input parameter as a second parameter signal.

8. The electromechanical equipment maintenance prediction system based on big data according to claim 7, characterized in that: The decision-making and control module includes an early warning processing unit and an equipment control unit; The warning processing unit is used to receive the first parameter signal and the second parameter signal, extract the fault time point in the first parameter signal and the fault time point in the second parameter signal, and output the extracted fault time point as a warning signal to send to the warning terminal; The equipment control unit is used to extract historical data from the first parameter signal and control input parameters from the second parameter signal, and output the extracted historical data and control input parameters as a target control signal to be sent to the electromechanical equipment control terminal.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the electromechanical equipment maintenance prediction system based on big data described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the electromechanical equipment maintenance prediction system based on big data described in any one of claims 1 to 8 is implemented.

Images