Sanitation equipment fault prediction method and system based on artificial intelligence

Abstract

The invention relates to the technical field of electromechanical equipment state monitoring and intelligent operation and maintenance, in particular to a sanitation equipment fault prediction method and system based on artificial intelligence, and the method comprises the steps: obtaining and storing the operation data and equipment numbers of different equipment; preprocessing the operation data, wherein the operation data comprises a mechanical vibration amplitude, a three-phase current waveform and a hydraulic oil temperature; based on preset extraction rules, corresponding abnormal features are extracted from the operation data, and the extraction rules comprise a time-frequency domain statistical feature extraction rule, a current waveform distortion feature extraction rule and a temperature time sequence mode feature extraction rule; packaging the extracted abnormal features into a feature data set associated with the device number; and inputting the feature data sets to a pre-trained fault prediction model one by one, performing fault prediction on the feature data sets by the fault prediction model, and mapping a prediction result to an equipment number. The method has the effects of accurately predicting the fault of the environmental sanitation equipment and improving the operation and maintenance efficiency.

Description

Technical Field

[0001] This application relates to the field of electromechanical equipment condition monitoring and intelligent operation and maintenance technology, specifically an artificial intelligence-based method and system for predicting sanitation equipment failures. Background Technology

[0002] There are currently three major bottlenecks in the fault monitoring of sanitation equipment:

[0003] It is difficult to effectively integrate multi-source heterogeneous data such as vibration signals and current waveforms;

[0004] Early, subtle fault characteristics are easily masked by noise.

[0005] There is a nonlinear coupling relationship between temperature rise and mechanical vibration in hydraulic systems. Traditional threshold methods cannot establish a correlation model, resulting in a high rate of sudden equipment failure.

[0006] Therefore, improvements are needed. Summary of the Invention

[0007] To address the above technical issues, this application provides a method and system for predicting sanitation equipment failures based on artificial intelligence.

[0008] The first objective of this invention is achieved through the following technical solution:

[0009] An artificial intelligence-based method for predicting sanitation equipment malfunctions includes:

[0010] Acquire and store the operating data and device numbers of different devices;

[0011] The operating data is preprocessed, and the operating data includes mechanical vibration amplitude, three-phase current waveform, and hydraulic oil temperature.

[0012] Based on preset extraction rules, corresponding abnormal features are extracted from the preprocessed running data. The extraction rules include time-frequency domain statistical feature extraction rules, current waveform distortion feature extraction rules, and temperature time series pattern feature extraction rules.

[0013] The extracted abnormal features are packaged into a feature dataset associated with the device ID of the corresponding device;

[0014] The feature datasets are input one by one into a pre-trained fault prediction model. The fault prediction model performs fault prediction on the input feature datasets and maps the prediction results to the device numbers associated with the feature datasets.

[0015] In a preferred embodiment, the step of extracting corresponding abnormal features from preprocessed runtime data based on preset extraction rules, wherein the extraction rules include time-frequency domain statistical feature extraction rules, current waveform distortion feature extraction rules, and temperature time series pattern feature extraction rules, includes:

[0016] Extract the first anomalous feature from the mechanical vibration amplitude based on time-frequency domain statistical feature extraction rules;

[0017] Based on the current waveform distortion feature extraction rules, the corresponding second abnormal feature is extracted from the three-phase current waveform;

[0018] Based on the temperature time-series pattern feature extraction rules, the corresponding third abnormal feature is extracted from the hydraulic oil temperature.

[0019] In a preferred embodiment, the step of extracting the corresponding first anomalous feature from the mechanical vibration amplitude based on the time-frequency domain statistical feature extraction rule includes:

[0020] Wavelet packet decomposition is performed on the preprocessed mechanical vibration amplitude to calculate the energy entropy of the preset frequency band energy and the total energy;

[0021] Calculate the kurtosis and crazing factor of the mechanical vibration amplitude;

[0022] The calculated energy entropy, kurtosis, and peak factor are used as the first anomaly feature.

[0023] In a preferred embodiment, the step of extracting the corresponding second abnormal feature from the three-phase current waveform based on the current waveform distortion feature extraction rule includes:

[0024] Perform a fast Fourier transform on the preprocessed three-phase current waveform to calculate the harmonic distortion rate and total inter-harmonic distortion rate of a preset order.

[0025] Calculate the unbalance of the three-phase current waveform;

[0026] The calculated THD, TDD, and imbalance are used as the second abnormal feature.

[0027] In a preferred embodiment, the step of extracting the corresponding third abnormal feature from the hydraulic oil temperature based on the temperature time-series pattern feature extraction rule includes:

[0028] Calculate the average temperature rise rate of hydraulic oil within a preset time window; calculate the cumulative over-temperature time when the hydraulic oil temperature exceeds a preset safety threshold; calculate the standard deviation of hydraulic oil temperature fluctuation within a preset normal operating range.

[0029] The calculated average temperature rise rate, cumulative over-temperature time, and standard deviation of fluctuation are used as the third abnormal feature.

[0030] In a preferred embodiment, the step of inputting the feature dataset one by one into a pre-trained fault prediction model, wherein the fault prediction model performs fault prediction on the input feature dataset and maps the prediction results to the device numbers associated with the feature dataset includes:

[0031] The fault prediction model includes an identification layer, an adjustment layer, an association layer, and a prediction layer;

[0032] The recognition layer performs multimodal data type recognition on the first abnormal feature, the second abnormal feature, and the third abnormal feature, respectively;

[0033] The adjustment layer dynamically adjusts the weights of each modal data type based on a preset multi-head attention mechanism;

[0034] The association layer, based on dynamically adjusted weights, performs feature splicing and spatiotemporal correlation analysis on each modality of data using graph convolutional networks, and generates association features containing spatiotemporal correlation information.

[0035] The prediction layer predicts the trends of vibration amplitude change, harmonic distortion, and oil temperature rise based on the correlation features and the operational data, and outputs the prediction results.

[0036] In a preferred embodiment, the step of the prediction layer predicting the vibration amplitude change trend, harmonic distortion trend, and oil temperature rise trend based on correlation features and the operating data, and outputting the prediction results, includes:

[0037] A multivariate time series prediction model is constructed. The associated features and operating data are input, and the predicted values ​​of vibration amplitude change trend, harmonic distortion trend and oil temperature rise trend are output respectively within the future preset time window.

[0038] The predicted values ​​of vibration amplitude change trend, harmonic distortion trend, and oil temperature rise trend are fused with the first, second, and third abnormal features extracted at the current moment to form a comprehensive feature vector.

[0039] The comprehensive feature vector is input into the fault mode classifier, and the output is a fault probability vector representing the probability of occurrence of different fault types. The fault types include at least one of bearing wear, motor inter-turn short circuit, hydraulic system leakage, and oil circuit blockage.

[0040] The failure probability vector is the prediction result.

[0041] The second objective of this invention is achieved through the following technical solution:

[0042] Module 1: Acquire and store the operating data and device numbers of different devices;

[0043] The second module: preprocesses the operating data, which includes mechanical vibration amplitude, three-phase current waveform, and hydraulic oil temperature;

[0044] The third module: Based on preset extraction rules, extract corresponding abnormal features from the preprocessed running data. The extraction rules include time-frequency domain statistical feature extraction rules, current waveform distortion feature extraction rules, and temperature time series pattern feature extraction rules.

[0045] The fourth module: Packages the extracted anomaly features into a feature dataset associated with the device ID of the corresponding device;

[0046] The fifth module: Input the feature datasets one by one into the pre-trained fault prediction model. The fault prediction model performs fault prediction on the input feature datasets and maps the prediction results to the device numbers associated with the feature datasets.

[0047] The third objective of this invention is achieved through the following technical solution:

[0048] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described artificial intelligence-based sanitation equipment fault prediction method.

[0049] The fourth objective of this invention is achieved through the following technical solution:

[0050] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned artificial intelligence-based sanitation equipment fault prediction method.

[0051] In summary, this application includes at least one of the following beneficial technical effects:

[0052] This method achieves accurate prediction of sanitation equipment faults by employing multimodal data fusion and deep learning techniques. First, a sensor network is used to collect real-time operational data from the equipment, including mechanical vibration amplitude, three-phase current waveforms, and hydraulic oil temperature. This data is preprocessed to improve data quality. Next, based on time-frequency domain statistical feature extraction rules, current waveform distortion feature extraction rules, and temperature time-series pattern feature extraction rules, abnormal features reflecting the equipment's operating status are extracted from the preprocessed data. These abnormal features are then associated with the equipment number and packaged into a feature dataset, which serves as input to the fault prediction model. Finally, the prediction results are mapped to the corresponding equipment number, providing early warning information for equipment maintenance. This method significantly improves the accuracy and reliability of sanitation equipment fault prediction. By monitoring the equipment's operating status in real time and promptly identifying and warning of potential faults, this method guides maintenance personnel to take timely measures to prevent faults from occurring, thereby improving equipment operating efficiency and reliability and reducing maintenance costs. Attached Figure Description

[0053] Figure 1 This is a flowchart of an implementation of an embodiment of an artificial intelligence-based method for predicting sanitation equipment failures according to this application.

[0054] Figure 2 This is a flowchart of step S30 in an embodiment of an artificial intelligence-based method for predicting sanitation equipment faults according to this application.

[0055] Figure 3 This is a flowchart of step S301 in an embodiment of an artificial intelligence-based method for predicting sanitation equipment faults according to this application.

[0056] Figure 4 This is a flowchart of step S302 in an embodiment of an artificial intelligence-based method for predicting sanitation equipment faults according to this application.

[0057] Figure 5 This is a schematic block diagram of a computer device according to this application. Detailed Implementation

[0058] The following is in conjunction with the appendix Figures 1-5 This application will be described in further detail.

[0059] In one embodiment, such as Figure 1 As shown, this application discloses an artificial intelligence-based method for predicting sanitation equipment malfunctions, which specifically includes the following steps:

[0060] S10: Acquire and store the operating data and device number of different devices;

[0061] S20: Preprocess the operating data, which includes mechanical vibration amplitude, three-phase current waveform, and hydraulic oil temperature;

[0062] S30: Based on preset extraction rules, extract corresponding abnormal features from the preprocessed running data. The extraction rules include time-frequency domain statistical feature extraction rules, current waveform distortion feature extraction rules, and temperature time series pattern feature extraction rules.

[0063] S40: Package the extracted abnormal features into a feature dataset associated with the device number of the corresponding device;

[0064] S50: Input the feature datasets one by one into the pre-trained fault prediction model. The fault prediction model performs fault prediction on the input feature datasets and maps the prediction results to the device numbers associated with the feature datasets.

[0065] In this embodiment, multimodal data fusion and deep learning techniques are used to achieve accurate prediction of sanitation equipment faults. First, a sensor network is used to collect real-time operational data of the equipment, including mechanical vibration amplitude, three-phase current waveforms, and hydraulic oil temperature. This data is preprocessed to improve data quality. Next, based on time-frequency domain statistical feature extraction rules, current waveform distortion feature extraction rules, and temperature time-series pattern feature extraction rules, abnormal features reflecting the equipment's operating status are extracted from the preprocessed data. These abnormal features are then associated with the equipment number and packaged into a feature dataset, which serves as input to the fault prediction model. Finally, the prediction results are mapped to the corresponding equipment number, providing early warning information for equipment maintenance. This method significantly improves the accuracy and reliability of sanitation equipment fault prediction. By monitoring the equipment's operating status in real time and promptly identifying and warning of potential faults, this method can guide maintenance personnel to take timely measures to prevent faults from occurring, thereby improving equipment operating efficiency and reliability and reducing maintenance costs.

[0066] like Figure 2 As shown, step S30 includes:

[0067] S301: Extract the corresponding first abnormal feature from the mechanical vibration amplitude based on the time-frequency domain statistical feature extraction rules;

[0068] S302: Extract the corresponding second abnormal feature from the three-phase current waveform based on the current waveform distortion feature extraction rule;

[0069] S303: Extract the corresponding third abnormal feature from the hydraulic oil temperature based on the temperature time series pattern feature extraction rule.

[0070] In this embodiment, deep feature mining of mechanical vibration, current waveform and hydraulic oil temperature is achieved by extracting statistical features in the time and frequency domain, extracting current waveform distortion features and extracting temperature time series pattern features. The above feature extraction rules are applied independently to signals in different physical domains, laying a data foundation for subsequent multimodal feature fusion and fault prediction.

[0071] like Figure 3 As shown, step S301 includes:

[0072] SB1: Perform wavelet packet decomposition on the preprocessed mechanical vibration amplitude and calculate the energy entropy of the preset frequency band energy and the total energy;

[0073] SB2: Calculate the kurtosis and crescendo factor of the mechanical vibration amplitude;

[0074] SB3: The calculated energy entropy, kurtosis, and peak factor are used as the first anomaly feature.

[0075] In this embodiment, based on the time-frequency domain statistical feature extraction rules, the mechanical vibration amplitude is decomposed by wavelet packet and the energy entropy, kurtosis and peak factor are calculated. This allows for the accurate quantification of the impact energy distribution and non-Gaussian characteristics during equipment operation, and effectively captures weak vibration anomalies caused by faults such as bearing wear.

[0076] like Figure 4 As shown, step S302 includes:

[0077] SE1: Perform a fast Fourier transform on the preprocessed three-phase current waveform to calculate the harmonic distortion rate and total inter-harmonic distortion rate of the preset order.

[0078] SE2: Calculate the unbalance of the three-phase current waveform;

[0079] SE3: The calculated THD, TDD, and imbalance are used as the second abnormal feature.

[0080] In this embodiment, based on the current waveform distortion feature extraction rules, the three-phase current waveform is analyzed using fast Fourier transform to extract the harmonic distortion rate, total inter-harmonic distortion rate and three-phase imbalance, which can accurately reflect the changes in motor load and the degree of nonlinear distortion, and provide early warning indicators for faults such as inter-turn short circuits in the motor.

[0081] Step S303 includes:

[0082] SK1: Calculate the average temperature rise rate of hydraulic oil within a preset time window; SK2: Calculate the cumulative over-temperature time when the hydraulic oil temperature exceeds the preset safety threshold; SK3: Calculate the standard deviation of hydraulic oil temperature fluctuation within a preset normal operating range;

[0083] SK4: The calculated average temperature rise rate, cumulative over-temperature time, and standard deviation of fluctuation are used as the third abnormal feature.

[0084] In this embodiment, based on the temperature time series pattern feature extraction rules, by calculating the average temperature rise rate, cumulative over-temperature time and fluctuation standard deviation of hydraulic oil temperature, the thermodynamic behavior and potential fault modes of the hydraulic system can be revealed, such as the nonlinear temperature rise law caused by oil circuit blockage.

[0085] The S50 steps include:

[0086] S501: The fault prediction model includes an identification layer, an adjustment layer, an association layer, and a prediction layer;

[0087] S502: The recognition layer performs multimodal data type recognition on the first abnormal feature, the second abnormal feature and the third abnormal feature respectively;

[0088] S503: The adjustment layer dynamically adjusts the weights of each modal data type based on a preset multi-head attention mechanism;

[0089] S504: The association layer performs feature splicing and spatiotemporal correlation analysis of each modality data based on dynamically adjusted weights, and generates association features containing spatiotemporal correlation information.

[0090] S505: The prediction layer predicts the vibration amplitude change trend, harmonic distortion trend, and oil temperature rise trend based on the correlation features and the operating data, and outputs the prediction results.

[0091] In this embodiment, the fault prediction model achieves efficient fusion and in-depth analysis of multimodal data through a hierarchical architecture. Its core lies in constructing a complete inference chain from feature recognition to trend prediction. The recognition layer, as the data entry point, first labels the three heterogeneous features—vibration, current, and temperature—with modal types. It embeds vector codes to encode the physical attributes of different data sources (such as the time-frequency coupling characteristics of vibration signals, the periodicity of current waveforms, and the slow-varying characteristics of temperature time series), providing a semantic benchmark for subsequent cross-modal calculations. The adjustment layer introduces a multi-head attention mechanism, calculating the global correlation score of each modal feature through self-attention and dynamically allocating weight coefficients. For example, when a hydraulic system leaks, the weight of the temperature feature automatically increases to strengthen the thermodynamic anomaly signal, while the weight of the vibration feature decreases accordingly to suppress mechanical noise interference. The association layer uses a graph convolutional network to construct a spatiotemporal topology graph of the equipment's operating state, mapping discrete feature nodes to a spatial association graph. Through edge weights, it learns the hysteresis-synchronous coupling relationship between vibration spectrum diffusion, current harmonic propagation, and temperature gradient changes, generating high-order association features with spatiotemporal dependencies. The prediction layer is divided into two channels: the time series prediction channel uses the Transformer architecture to output three sets of predicted values ​​for vibration amplitude, harmonic distortion rate, and oil temperature for the next 10 / 30 / 60 minutes based on associated features. Its attention mechanism can capture long-term trend changes. The fault classification channel combines the predicted trend with the current features through Concat and outputs probability vectors for six typical faults such as bearing wear and motor short circuit through the improved XGBoost algorithm.

[0092] Step S505 includes:

[0093] SV1: Construct a multivariate time series prediction model, input the associated features and running data, and output the predicted values ​​of vibration amplitude change trend, harmonic distortion trend, and oil temperature rise trend for the future preset time window, respectively.

[0094] SV2: The predicted values ​​of vibration amplitude change trend, harmonic distortion trend, and oil temperature rise trend are fused with the first, second, and third abnormal features extracted at the current moment to form a comprehensive feature vector.

[0095] SV3: Input the comprehensive feature vector into the fault mode classifier and output a fault probability vector representing the probability of occurrence of different fault types. The fault types include at least one of bearing wear, motor inter-turn short circuit, hydraulic system leakage, and oil circuit blockage.

[0096] SV4: The failure probability vector is the prediction result.

[0097] In this embodiment, the fault prediction model achieves trend prediction and fault diagnosis of equipment status through a dual-channel collaborative mechanism: First, a multivariate time series prediction model is constructed. This model takes the spatiotemporal coupling characteristics output from the correlation layer and the original operating data as input. It captures the cross-modal time-series dependence of three parameters—vibration amplitude, harmonic distortion rate, and oil temperature—using a Transformer encoder, and outputs predicted values ​​for three time windows: 10 minutes, 30 minutes, and 60 minutes, respectively, using an autoregressive decoding structure. For example, the predicted value of the vibration amplitude change trend can provide an early warning of gear meshing abnormalities one hour in advance. Then... The time-series prediction results are fused with the extracted time-frequency domain statistical features, current distortion features, and temperature time-series features at the current moment. A fully connected layer is used to achieve nonlinear mapping of the high-dimensional feature space, generating a comprehensive feature vector containing historical state memory and future evolution information. This vector is then input into a fault mode classifier based on an improved LightGBM framework. Its gradient boosting framework is used to probabilistically model six typical fault modes, including bearing wear and inter-turn short circuits in motors, outputting a vector composed of the probability values ​​of each fault type. For example, a maintenance work order is triggered when the probability of leakage in the hydraulic system exceeds 0.85. This design injects knowledge into the time-series prediction results and real-time features, enabling the classifier to capture the dynamic process of fault evolution and correlate it with the current physical representation.

[0098] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0099] In one embodiment, an artificial intelligence-based sanitation equipment fault prediction system is provided, which corresponds to the artificial intelligence-based sanitation equipment fault prediction method described in the above embodiment. This artificial intelligence-based sanitation equipment fault prediction system includes:

[0100] Module 1: Acquire and store the operating data and device numbers of different devices;

[0101] The second module: preprocesses the operating data, which includes mechanical vibration amplitude, three-phase current waveform, and hydraulic oil temperature;

[0102] The third module: Based on preset extraction rules, extract corresponding abnormal features from the preprocessed running data. The extraction rules include time-frequency domain statistical feature extraction rules, current waveform distortion feature extraction rules, and temperature time series pattern feature extraction rules.

[0103] The fourth module: Packages the extracted anomaly features into a feature dataset associated with the device ID of the corresponding device;

[0104] The fifth module: Input the feature datasets one by one into the pre-trained fault prediction model. The fault prediction model performs fault prediction on the input feature datasets and maps the prediction results to the device numbers associated with the feature datasets.

[0105] Optional, also includes:

[0106] Module 6: Extracting the first anomalous feature from the mechanical vibration amplitude based on time-frequency domain statistical feature extraction rules;

[0107] Module 7: Extracting the corresponding second abnormal feature from the three-phase current waveform based on the current waveform distortion feature extraction rules;

[0108] Module 8: Extract the corresponding third abnormal feature from hydraulic oil temperature based on temperature time-series pattern feature extraction rules.

[0109] Optional, also includes:

[0110] Module 9: Perform wavelet packet decomposition on the preprocessed mechanical vibration amplitude and calculate the energy entropy of the preset frequency band energy and the total energy;

[0111] Module 10: Calculate the kurtosis and peak factor of mechanical vibration amplitude;

[0112] Module 11: The calculated energy entropy, kurtosis, and peak factor are used as the first anomaly feature.

[0113] Optional, also includes:

[0114] Module 12: Performs Fast Fourier Transform on the preprocessed three-phase current waveform to calculate the harmonic distortion rate and total interharmonic distortion rate of the preset order.

[0115] Module 13: Calculate the unbalance of three-phase current waveforms;

[0116] Module 14: The calculated THD, TDD, and imbalance are used as the second abnormal feature.

[0117] Optional, also includes:

[0118] Module 15: Calculates the average temperature rise rate of hydraulic oil within a preset time window; Module 16: Calculates the cumulative over-temperature time when the hydraulic oil temperature exceeds the preset safety threshold; Module 17: Calculates the standard deviation of hydraulic oil temperature fluctuation within a preset normal operating range.

[0119] Module 18: The calculated average temperature rise rate, cumulative over-temperature time, and standard deviation of fluctuation are used as the third abnormal feature.

[0120] Optional, also includes:

[0121] Module 19: The fault prediction model includes an identification layer, an adjustment layer, a correlation layer, and a prediction layer;

[0122] Module 20: The recognition layer performs multimodal data type recognition on the first abnormal feature, the second abnormal feature, and the third abnormal feature, respectively;

[0123] Module 21: The adjustment layer dynamically adjusts the weights of each modal data type based on a preset multi-head attention mechanism;

[0124] Module 22: The association layer, based on dynamically adjusted weights, performs feature splicing and spatiotemporal correlation analysis on each modality of data using graph convolutional networks, and generates association features containing spatiotemporal correlation information.

[0125] Modules 2 and 3: The prediction layer, based on the correlation features and the operating data, predicts the trend of vibration amplitude change, harmonic distortion trend, and oil temperature rise trend, and outputs the prediction results.

[0126] Optional, also includes:

[0127] Module 24: Construct a multivariate time series prediction model, input the associated features and operating data, and output the predicted values ​​of vibration amplitude change trend, harmonic distortion trend, and oil temperature rise trend for the future preset time window, respectively.

[0128] Module 25: The predicted values ​​of vibration amplitude change trend, harmonic distortion trend, and oil temperature rise trend are fused with the first, second, and third abnormal features extracted at the current moment to form a comprehensive feature vector.

[0129] Module 26: Input the comprehensive feature vector into the fault mode classifier and output a fault probability vector representing the probability of occurrence of different fault types. The fault types include at least one of bearing wear, motor inter-turn short circuit, hydraulic system leakage, and oil circuit blockage.

[0130] Module 27: The fault probability vector is the prediction result.

[0131] For specific limitations regarding an AI-based sanitation equipment fault prediction system, please refer to the limitations of an AI-based sanitation equipment fault prediction method described above, which will not be repeated here. The various modules in the aforementioned AI-based sanitation equipment fault prediction and control system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0132] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores device numbers. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an artificial intelligence-based method for predicting sanitation equipment faults.

[0133] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements an artificial intelligence-based method for predicting sanitation equipment failures.

[0134] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, is an artificial intelligence-based method for predicting sanitation equipment malfunctions.

[0135] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0136] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

Claims

1. A method for predicting failure of a sanitation device based on artificial intelligence, the method comprising: The method comprises the following steps: acquiring and storing running data and equipment numbers of different equipment; preprocessing the running data, which comprises mechanical vibration amplitude, three-phase current waveform and hydraulic oil temperature; extracting corresponding abnormal features from the preprocessed running data based on preset extraction rules, which comprise time-frequency domain statistical feature extraction rule, current waveform distortion feature extraction rule and temperature time series pattern feature extraction rule; packaging the extracted abnormal features into feature data sets associated with the equipment numbers of the corresponding equipment; inputting the feature data sets into a pre-trained fault prediction model, which performs fault prediction on the input feature data sets and maps the prediction results to the equipment numbers associated with the feature data sets. 2.The method of claim 1, wherein: The step of extracting corresponding abnormal features from the preprocessed running data based on preset extraction rules, which comprise time-frequency domain statistical feature extraction rule, current waveform distortion feature extraction rule and temperature time series pattern feature extraction rule, comprises: extracting corresponding first abnormal features from the mechanical vibration amplitude based on the time-frequency domain statistical feature extraction rule; extracting corresponding second abnormal features from the three-phase current waveform based on the current waveform distortion feature extraction rule; extracting corresponding third abnormal features from the hydraulic oil temperature based on the temperature time series pattern feature extraction rule. 3.The method of claim 1 or 2, wherein: The step of extracting corresponding first abnormal features from the mechanical vibration amplitude based on the time-frequency domain statistical feature extraction rule comprises: performing wavelet packet decomposition on the preprocessed mechanical vibration amplitude, calculating the energy entropy of the preset frequency band energy and the total energy; calculating the kurtosis and peak factor of the mechanical vibration amplitude; taking the calculated energy entropy, kurtosis and peak factor as the first abnormal features.

4. The method of claim 1 or 2, wherein the method further comprises: The step of extracting corresponding second abnormal features from the three-phase current waveform based on the current waveform distortion feature extraction rule comprises: performing fast Fourier transform on the preprocessed three-phase current waveform, calculating the harmonic distortion rate of the preset number of times and the total inter-harmonic distortion rate; calculating the unbalance degree of the three-phase current waveform; taking the calculated THD, TDD and unbalance degree as the second abnormal features.

5. The method of claim 1 or 2, wherein the method further comprises: The step of extracting corresponding third abnormal features from the hydraulic oil temperature based on the temperature time series pattern feature extraction rule comprises: calculating the average temperature rise rate of the hydraulic oil temperature within a preset time window; calculating the cumulative over-temperature time of the hydraulic oil temperature exceeding a preset safety threshold; calculating the fluctuation standard deviation of the hydraulic oil temperature within a preset normal working interval; taking the calculated average temperature rise rate, cumulative over-temperature time and fluctuation standard deviation as the third abnormal features. 6.The method of claim 1, wherein the method further comprises: The step of inputting the feature data sets into a pre-trained fault prediction model, which performs fault prediction on the input feature data sets and maps the prediction results to the equipment numbers associated with the feature data sets, comprises: The fault prediction model comprises an identification layer, an adjustment layer, an association layer and a prediction layer; The identification layer performs multi-modal data type identification on the first abnormal features, the second abnormal features and the third abnormal features respectively; The adjustment layer dynamically adjusts the weight of each modal data type based on a preset multi-head attention mechanism; The association layer performs feature splicing and graph convolution network spatio-temporal association analysis on each modal data based on the dynamically adjusted weight, and generates association features containing spatio-temporal association information; The prediction layer predicts the vibration amplitude change trend, harmonic distortion trend and oil temperature rise trend based on the association features and the operation data, and outputs the prediction results.

7. The method of claim 6, wherein the method further comprises: The prediction layer predicts the vibration amplitude change trend, harmonic distortion trend and oil temperature rise trend based on the association features and the operation data, and outputs the prediction results. A multivariate time series prediction model is constructed, the association features and the operation data are input, and the vibration amplitude change trend prediction value, the harmonic distortion trend prediction value and the oil temperature rise trend prediction value of a future preset time window are output respectively; The vibration amplitude change trend prediction value, the harmonic distortion trend prediction value and the oil temperature rise trend prediction value are fused with the first abnormal feature, the second abnormal feature and the third abnormal feature extracted at the current time to form a comprehensive feature vector; The comprehensive feature vector is input into a fault mode classifier to output a fault probability vector representing the probability of occurrence of different fault types, the fault types including at least one of bearing wear, motor inter-turn short circuit, hydraulic system leakage and oil line blockage; The fault probability vector is the prediction result. 8.A system for predicting failure of a sanitation device based on artificial intelligence, characterized by, It includes: The first module: acquiring and storing the operation data and equipment number of different equipment; The second module: preprocessing the operation data, which includes mechanical vibration amplitude, three-phase current waveform and hydraulic oil temperature; The third module: based on the preset extraction rule, extracting the corresponding abnormal features from the preprocessed operation data, the extraction rule including time-frequency domain statistical feature extraction rule, current waveform distortion feature extraction rule and temperature time series pattern feature extraction rule; The fourth module: packaging the extracted abnormal features into a feature data set associated with the equipment number of the corresponding equipment; The fifth module: inputting the feature data set into the pre-trained fault prediction model, the fault prediction model performing fault prediction on the input feature data set, and mapping the prediction result to the equipment number associated with the feature data set.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the artificial intelligence-based sanitation equipment fault prediction method according to claims 1-7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the steps of the artificial intelligence-based sanitation equipment fault prediction method according to claims 1-7.

Images