Multi-working condition automatic identification method for pump equipment

Abstract

This application relates to an automatic multi-condition identification method for pump equipment. The method includes: collecting monitoring sequences of the pump equipment and obtaining a first-dimensional process detection quantity; performing feature calculation based on the process detection quantity to obtain a fixed-length feature vector in the second dimension; using a heterogeneous lightweight model to identify the operating conditions of the pump equipment based on the fixed-length feature vector, obtaining multiple operating condition results for the pump equipment; and using a preset voting mechanism to fuse the multiple operating condition results of the pump equipment to determine the actual operating condition of the pump equipment. By converting the process detection quantity into a fixed-length feature vector, this method avoids performance fluctuations in the heterogeneous lightweight model caused by inconsistent monitoring sequence lengths, thus improving the accuracy of identifying the operating condition results of the pump equipment.

Description

Technical Field

[0001] This application relates to the field of rotating machinery fault diagnosis and operating condition identification technology, and in particular to a method for automatic identification of multiple operating conditions of pump equipment. Background Technology

[0002] With the deepening of industrial intelligence and energy structure upgrading, pumps, as the core power units of fluid transportation systems, are gradually undertaking the main media transportation tasks in critical infrastructures such as power, petrochemical, and municipal industries. Their operating status directly affects system safety and energy efficiency. However, due to limitations in on-site installation conditions and sensor placement, pumps often operate under harsh conditions such as high temperature, high noise, and high vibration, making them prone to cavitation, wear, misalignment, and other faults. If these faults are not detected in time, they can lead to equipment performance degradation, downtime accidents, or even threaten production safety. Therefore, accurate identification of the operating conditions of pumps can not only identify potential fault risks in advance but also provide data support for operation and maintenance decisions, thereby effectively reducing the failure rate and improving the stability of system operation.

[0003] Currently, the identification of the operating conditions of pump equipment mainly relies on manual inspection or deep learning methods. However, manual inspection has a high error rate, and deep learning methods require a large number of samples for training. But industrial sites often only provide small sample operating condition data. Therefore, how to identify the operating conditions of pump equipment based on small sample operating condition data has become an urgent problem to be solved. Summary of the Invention

[0004] Therefore, it is necessary to provide an automatic identification method for multiple operating conditions of pump equipment to address the above-mentioned technical problems.

[0005] Firstly, this application provides a method for automatic identification of multiple operating conditions in pump equipment, including:

[0006] Collect monitoring sequences of pump-type equipment and obtain the first dimension of process detection quantity from them;

[0007] Feature calculation is performed based on the process detection quantity to obtain a fixed-length feature vector in the second dimension; the second dimension is greater than the first dimension.

[0008] The heterogeneous lightweight model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors, and the results of various operating conditions of pump equipment are obtained.

[0009] By using a pre-set voting mechanism to integrate the results of various operating conditions of pump equipment, the operating condition of the pump equipment can be determined.

[0010] In one embodiment, the preset voting mechanism includes a hard-label voting mechanism, which integrates multiple operating condition results of the pump equipment to determine the operating condition of the pump equipment, including:

[0011] The two-thirds majority vote based on the hard-label voting mechanism filters out multiple operating conditions and determines the operating condition of pump equipment based on the filtered results.

[0012] In one embodiment, the preset voting mechanism further includes a soft-label voting mechanism, which integrates multiple operating condition results of the pump equipment to determine the operating condition of the pump equipment, and also includes:

[0013] The average of multiple operating conditions is calculated based on a soft-label voting mechanism, and the operating condition of pump equipment is determined based on the average results.

[0014] In one embodiment, a two-thirds majority vote based on a hard-label voting mechanism is used to filter multiple operating condition results, and the operating condition of the pump equipment is determined based on the filtered results, including:

[0015] Based on the hard-label voting mechanism, a two-thirds majority vote is used to select the working condition result that is consistent from multiple working condition results.

[0016] Determine the operating conditions of pump equipment based on the operating results.

[0017] In one embodiment, determining the operating condition of the pump equipment based on the operating condition results includes:

[0018] When the proportion of consistent operating conditions is greater than or equal to two-thirds, the operating condition of the pump equipment shall be determined based on the consistent operating conditions.

[0019] When the proportion of consistent operating conditions is less than two-thirds, the operating condition of the pump equipment is determined based on the operating condition with the highest confidence level among the various operating condition results; the confidence level is greater than the preset confidence threshold.

[0020] In one embodiment, feature calculation based on the process detection quantity yields a fixed-length feature vector in the second dimension, including:

[0021] Motor current, motor speed, and total flow rate were selected as three sensitive channels from the process detection parameters.

[0022] The interval characteristics of the monitoring sequence in the third dimension are calculated channel by channel based on the three sensitive channels; the third dimension is smaller than the first dimension.

[0023] Based on the three sensitive channels and the interval features of the third dimension, the monitoring sequence is mapped to a fixed-length feature vector of the second dimension.

[0024] In one embodiment, the automatic identification method for multiple operating conditions of pump equipment further includes:

[0025] The process detection quantities in the first dimension are sorted in descending order using mutual information metric to obtain the ranking results;

[0026] Motor current, motor speed, and total flow rate were selected as three sensitive channels from the process monitoring parameters, including:

[0027] Based on the ranking results, motor current, motor speed, and total flow rate were selected as three sensitive channels from the process detection quantities in the first dimension.

[0028] In one embodiment, the heterogeneous lightweight model includes a radial basis function kernel support vector machine model, a one-dimensional convolutional neural network model, and a transformer model;

[0029] A radial basis function kernel support vector machine model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors and output the first operating condition result.

[0030] A one-dimensional convolutional neural network model is used to identify the operating conditions of pump equipment based on a fixed-length feature vector and output the second operating condition result.

[0031] The converter model is used to identify the operating conditions of pump equipment based on a fixed-length feature vector and output the third operating condition result.

[0032] By using a pre-defined voting mechanism to merge results from multiple operating conditions of pump equipment, the operating condition of the pump equipment is determined, including:

[0033] The results of the first, second, and third operating conditions are merged using a preset voting mechanism to determine the operating condition of the pump equipment.

[0034] In one embodiment, the automatic identification method for multiple operating conditions of pump equipment further includes:

[0035] Initialize the weights of the initial radial basis function kernel support vector machine model, the weights of the one-dimensional convolutional neural network model, and the weights of the transformer model;

[0036] Based on the monitoring sample data of pump equipment, the initialized initial radial basis function kernel support vector machine model, the initialized initial one-dimensional convolutional neural network model, and the initialized initial converter model are trained to obtain the trained radial basis function kernel support vector machine model, one-dimensional convolutional neural network model, and converter model.

[0037] Secondly, this application also provides an automatic multi-condition identification device for pump equipment, comprising:

[0038] The data acquisition module is used to acquire monitoring sequences of pump-type equipment and obtain the first-dimensional process detection quantity from them.

[0039] The calculation module is used to perform feature calculation based on the process detection quantity to obtain a fixed-length feature vector in the second dimension;

[0040] The identification module is used to identify the operating conditions of pump equipment based on a fixed-length feature vector using a heterogeneous lightweight model, and obtain various operating condition results for the pump equipment.

[0041] The fusion module is used to merge the results of multiple operating conditions of pump equipment using a preset voting mechanism to determine the operating condition of the pump equipment.

[0042] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0043] Collect monitoring sequences of pump-type equipment and obtain the first dimension of process detection quantity from them;

[0044] Feature calculation is performed based on the process detection quantity to obtain a fixed-length feature vector in the second dimension; the second dimension is greater than the first dimension.

[0045] The heterogeneous lightweight model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors, and the results of various operating conditions of pump equipment are obtained.

[0046] By using a pre-set voting mechanism to integrate the results of various operating conditions of pump equipment, the operating condition of the pump equipment can be determined.

[0047] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0048] Collect monitoring sequences of pump-type equipment and obtain the first dimension of process detection quantity from them;

[0049] Feature calculation is performed based on the process detection quantity to obtain a fixed-length feature vector in the second dimension; the second dimension is greater than the first dimension.

[0050] The heterogeneous lightweight model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors, and the results of various operating conditions of pump equipment are obtained.

[0051] By using a pre-set voting mechanism to integrate the results of various operating conditions of pump equipment, the operating condition of the pump equipment can be determined.

[0052] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0053] Collect monitoring sequences of pump-type equipment and obtain the first dimension of process detection quantity from them;

[0054] Feature calculation is performed based on the process detection quantity to obtain a fixed-length feature vector in the second dimension; the second dimension is greater than the first dimension.

[0055] The heterogeneous lightweight model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors, and the results of various operating conditions of pump equipment are obtained.

[0056] By using a pre-set voting mechanism to integrate the results of various operating conditions of pump equipment, the operating condition of the pump equipment can be determined.

[0057] The aforementioned automatic multi-condition identification method for pump equipment involves collecting monitoring sequences of the pump equipment and obtaining a first-dimensional process detection quantity. Feature calculations are then performed based on the process detection quantity to obtain a fixed-length feature vector in the second dimension. A heterogeneous lightweight model is used to identify the operating conditions of the pump equipment based on the fixed-length feature vector, resulting in multiple operating condition outcomes. Finally, a pre-defined voting mechanism is used to fuse these multiple operating condition outcomes to determine the final operating condition of the pump equipment. This method, by converting process detection quantities into fixed-length feature vectors, avoids performance fluctuations in the heterogeneous lightweight model caused by inconsistent monitoring sequence lengths, thus improving the accuracy of identifying the operating condition outcomes of the pump equipment. Furthermore, the heterogeneous lightweight model comprises multiple lightweight models, resulting in low computational resource consumption. Different models capture different operating condition features from different angles, and the multiple operating condition outcomes output by the heterogeneous lightweight model avoid the blind spots of a single model, thereby improving the accuracy of the operating condition outcome identification. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is an application environment diagram of the automatic identification method for multiple operating conditions of pump equipment in one embodiment;

[0060] Figure 2 This is one of the flowcharts illustrating an automatic multi-condition identification method for pump equipment in one embodiment;

[0061] Figure 3 Here is a heatmap of the F1 fraction in one embodiment;

[0062] Figure 4This is a second flowchart illustrating a method for automatic identification of multiple operating conditions for pump equipment in one embodiment;

[0063] Figure 5 This is the third flowchart of a method for automatic identification of multiple operating conditions of pump equipment in one embodiment;

[0064] Figure 6 This is the fourth flowchart of a method for automatic identification of multiple operating conditions of pump equipment in one embodiment;

[0065] Figure 7 This is the fifth flowchart illustrating the automatic identification method for multiple operating conditions of pump equipment in one embodiment;

[0066] Figure 8 This is a flowchart of the automatic identification method for multiple operating conditions of pump equipment in one embodiment, shown as the sixth one.

[0067] Figure 9 This is an analysis diagram of a soft and hard voting system in one embodiment;

[0068] Figure 10 This is a diagram of an automatic multi-condition identification system for pump equipment in one embodiment;

[0069] Figure 11 This is the seventh flowchart of a method for automatic identification of multiple operating conditions of pump equipment in one embodiment;

[0070] Figure 12 Here is a training loss curve for one embodiment;

[0071] Figure 13 Here is a loss curve for verification in one embodiment;

[0072] Figure 14 Here is a training set accuracy curve in one embodiment;

[0073] Figure 15 Here is a test set accuracy curve from one embodiment;

[0074] Figure 16 This is a comparison chart of model accuracy in one embodiment;

[0075] Figure 17 This is the eighth flowchart of a method for automatic identification of multiple operating conditions of pump equipment in one embodiment;

[0076] Figure 18 This is a structural block diagram of a multi-condition automatic identification device for pump equipment in one embodiment;

[0077] Figure 19 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0078] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0079] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0080] With the deepening of industrial intelligence and energy structure upgrading, pumps, as the core power units of fluid transportation systems, are gradually undertaking the main media transportation tasks in critical infrastructures such as power, petrochemical, and municipal industries. Their operating status directly affects system safety and energy efficiency. However, due to limitations in on-site installation conditions and sensor placement, pumps often operate under harsh conditions such as high temperature, high noise, and high vibration, making them prone to cavitation, wear, misalignment, and other faults. If these faults are not detected in time, they can lead to equipment performance degradation, downtime accidents, or even threaten production safety. Therefore, accurate identification of the operating conditions of pumps can not only identify potential fault risks in advance but also provide data support for operation and maintenance decisions, thereby effectively reducing the failure rate and improving the stability of system operation.

[0081] Currently, the identification of the operating conditions of pump equipment mainly relies on manual inspection or deep learning methods. However, manual inspection has a high error rate, and deep learning methods require a large number of samples for training. But industrial sites often only provide small sample operating condition data. Therefore, how to identify the operating conditions of pump equipment based on small sample operating condition data has become an urgent problem to be solved.

[0082] In view of the above-mentioned technical problems, this application provides a method for automatic identification of multiple operating conditions of pump equipment. The following embodiments will specifically illustrate the method for automatic identification of multiple operating conditions of pump equipment.

[0083] The automatic multi-condition identification method for pump equipment provided in this application embodiment can be applied to, for example... Figure 1The illustrated pump equipment multi-condition automatic identification system includes a pump (101), a distributed control system (102), an operating condition identification device (103), and a display device (104). The distributed control system (102) transmits the raw data collected from the pump (101) to the operating condition identification device (103) via a network. The operating condition identification device (103) processes the raw data, identifies the operating condition of the pump (101), and displays the identified operating condition result on the display device (104). The pump (101) can be a centrifugal pump, positive displacement pump, axial flow pump, or mixed flow pump, etc.

[0084] Those skilled in the art will understand that Figure 1 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the multi-condition automatic identification system for pump equipment to which the present application is applied. The specific identification system may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0085] In one exemplary embodiment, such as Figure 2 As shown, an automatic multi-condition identification method for pump equipment is provided, which can be applied to... Figure 1 Taking the working condition identification device in the middle as an example, the explanation includes:

[0086] S201, collect the monitoring sequence of pump equipment and obtain the first dimension of process detection quantity from it.

[0087] The process detection quantities include the motor operating voltage. Motor current Total flow Input power Motor speed The monitoring sequence is the raw physical signal that reflects the operating status of pump equipment, collected at a fixed sampling frequency by a distributed control system installed on the pump and its associated equipment.

[0088] In this embodiment, a distributed control system installed on the pump and its associated equipment is used to continuously record and acquire monitoring sequences of the pump equipment. The duration of a single start-up and shutdown process of the pump equipment varies randomly between 10 and 70 seconds. The distributed control system transmits the acquired monitoring sequences to a condition identification device. The condition identification device first filters out high-frequency signals using a sliding window, then separates harmonic components in the monitoring sequences using signal processing methods (such as wavelet packet decomposition), and eliminates interference in the acquired monitoring sequences using Kalman filtering. Finally, it extracts the fundamental frequency signal from the monitoring sequences using time-frequency domain signal processing methods to generate the first-dimensional process detection quantity, specifically including the operating voltage. Motor current Total flow Input power Motor speed .

[0089] S202, based on the process detection quantity, perform feature calculation to obtain a fixed-length feature vector of the second dimension.

[0090] The second dimension is greater than the first dimension.

[0091] In this embodiment, after obtaining the process detection quantities in the first dimension, the working condition identification device sorts the process detection quantities in descending order to retain the most discriminative information under small sample conditions, and selects a preset number of key process detection quantities (e.g., selecting the top N by numerical value or weight, where N is determined by the preset number). Subsequently, based on the channel information contained in the preset number of process detection quantities, corresponding interval features (e.g., mean, variance, peak value, or slope within the interval) are extracted from the time interval, amplitude interval, or statistical interval of each channel information. Finally, the preset number of process detection quantities and their corresponding interval features are fused. By feature concatenation or weighted combination, the original sequence data of arbitrary length can be mapped into a fixed-length feature vector with a fixed dimension, i.e., the fixed-length feature vector of the second dimension.

[0092] S203 uses a heterogeneous lightweight model to identify the operating conditions of pump equipment based on a fixed-length feature vector, and obtains various operating condition results for pump equipment.

[0093] The heterogeneous lightweight model consists of a combination of three models: a radial basis function kernel support vector machine model, a one-dimensional convolutional neural network model, and a transformer model.

[0094] In this embodiment, the operating condition identification device inputs a fixed-length feature vector of the second dimension as input data to the radial basis function kernel support vector machine (RBKM) model, the one-dimensional convolutional neural network (CNN) model, and the converter model in the heterogeneous lightweight model. The RBKM model, the one-dimensional CNN model, and the converter model respectively identify the operating condition of the pump equipment based on the fixed-length feature vector of the second dimension. Subsequently, the output results (such as probability distributions or classification labels) of the RBKM model, the one-dimensional CNN model, and the converter model are combined, and a decision fusion algorithm is used to obtain the final operating condition type based on the output results of the three models.

[0095] S204 utilizes a preset voting mechanism to integrate multiple operating condition results of pump equipment to determine the operating condition of the pump equipment.

[0096] The preset voting mechanisms include hard-label voting and soft-label voting, and users can switch between the two mechanisms with one click according to the actual needs of the scenario.

[0097] In this embodiment, to prevent a single model from causing bias in the operating condition identification results, the operating condition identification device combines a preset voting mechanism consisting of hard-label voting and soft-label voting to fuse the multiple operating condition results of pump equipment output by the three models, thereby accurately determining the actual operating condition of the equipment. Hard-label voting can be based on a majority vote using discrete category labels output by the model, while soft-label voting can calculate a weighted confidence level using the continuous probability distribution output by the model. The user can switch between the two mechanisms with a single click according to the actual scenario requirements. In the case of small samples, i.e., only 3 to 5 samples per operating condition, the operating condition identification device fuses the multiple operating condition results of pump equipment through a preset voting mechanism that integrates hard-label voting and soft-label voting, ultimately outputting the operating condition status of the pump equipment. Figure 3 It can be seen that the difference in F1 scores for various working conditions is strictly controlled to be below 0.1.

[0098] The aforementioned automatic multi-condition identification method for pump equipment involves collecting monitoring sequences of the pump equipment and obtaining a first-dimensional process detection quantity. Feature calculations are then performed based on the process detection quantity to obtain a fixed-length feature vector in the second dimension. A heterogeneous lightweight model is used to identify the operating conditions of the pump equipment based on the fixed-length feature vector, resulting in multiple operating condition outcomes. Finally, a pre-defined voting mechanism is used to fuse these multiple operating condition outcomes to determine the final operating condition of the pump equipment. This method, by converting process detection quantities into fixed-length feature vectors, avoids performance fluctuations in the heterogeneous lightweight model caused by inconsistent monitoring sequence lengths, thus improving the accuracy of identifying the operating condition outcomes of the pump equipment. Furthermore, the heterogeneous lightweight model comprises multiple lightweight models, resulting in low computational resource consumption. Different models capture different operating condition features from different angles, and the multiple operating condition outcomes output by the heterogeneous lightweight model avoid the blind spots of a single model, thereby improving the accuracy of the operating condition outcome identification.

[0099] In one exemplary embodiment, Figure 2 The specific implementation of S204 in the embodiment, "using a preset voting mechanism to fuse multiple operating condition results of pump equipment to determine the operating condition of the pump equipment," is as follows: Figure 4 As shown, this includes: using a two-thirds majority vote based on a hard-label voting mechanism to filter multiple operating conditions, and determining the operating conditions of pump equipment based on the filtered operating conditions.

[0100] In this embodiment, when using a preset voting mechanism to make the final determination of the pump equipment's operating condition, the user can choose to use either a hard-label voting mechanism or a soft-label voting mechanism based on the actual situation. If the hard-label voting mechanism is chosen, the discrete category labels output by the three models can be used to filter multiple operating condition results based on the two-thirds majority vote of the hard-label voting mechanism. The operating condition of the pump equipment is then determined based on the filtered results. For example, if both the radial basis function kernel support vector machine model and the one-dimensional convolutional neural network model determine that the pump equipment is in an off-center operating condition, but the converter model determines that the pump equipment is in an unbalanced operating condition, then the final operating condition identification device determines that the pump equipment is in an off-center operating condition.

[0101] In one exemplary embodiment, Figure 2 The specific implementation of S204 in the embodiment, "using a preset voting mechanism to fuse multiple operating condition results of pump equipment to determine the operating condition of the pump equipment," is as follows: Figure 5 As shown, it also includes: averaging the results of multiple operating conditions based on a soft-label voting mechanism, and determining the operating condition of pump equipment based on the average results.

[0102] In this embodiment of the application, when the operating condition identification device uses a preset voting mechanism to make the final determination of the operating condition of pump equipment, the user can choose to use a hard label voting mechanism or a soft label voting mechanism to make the final determination of the operating condition of pump equipment according to the actual situation. If a soft-label voting mechanism is chosen for the final determination of the operating conditions of pump equipment, the soft-label voting mechanism first needs to obtain the probability vectors output by the three models through the flexible maximum function (each vector contains the confidence level of each operating condition category), then perform an average operation on the probability values ​​at corresponding positions of the three vectors to obtain a comprehensive probability vector, and finally select the operating condition category with the highest probability as the final determination result of the operating condition identification device through the maximum index function. For example, the operating condition probability vectors of the radial basis function kernel support vector machine model, the one-dimensional convolutional neural network model, and the transformer model for pump equipment are as follows: the operating condition probability vector of the radial basis function kernel support vector machine model is [0.7, 0.1, 0.05, 0.03, 0.06, 0.04, 0.02]; the operating condition probability vector of the one-dimensional convolutional neural network model is [0.65, 0.12, 0.06, 0.04, 0.02]. The converter model's operating probability vector is [0.72, 0.08, 0.04, 0.02, 0.05, 0.06, 0.03]. By taking the average through soft voting, the comprehensive probability vector is: [(0.7+0.65+0.72) / 3, (0.1+0.12+0.08) / 3, (0.05+0.06+0.04) / 3, (0.03+0.07) / 3, (0.03+0.07) / 3, (0.07+0.08) / 3, (0.05+0.06+0.04) / 3, (0.03+0.07) / 3, (0.07 ... If 0.04+0.02) / 3,(0.06+0.07+0.05) / 3,(0.04+0.03+0.06) / 3,(0.02+0.03+0.03) / 3]=[0.69,0.1,0.05,0.03,0.06,0.04,0.03], then the maximum index function selects the "normal" working condition corresponding to the first position as the final judgment result of the working condition identification device.

[0103] In an exemplary embodiment, the above method involves "screening multiple operating condition results based on a two-thirds majority vote using a hard-label voting mechanism, and determining the operating condition of the pump equipment based on the screened results," such as... Figure 6 As shown, it includes:

[0104] S301, based on a two-thirds majority vote using a hard-label voting mechanism, selects the working condition result that is consistent from multiple working condition results.

[0105] In this embodiment, after the user selects the hard label voting mechanism as the final identification method for pump equipment, the operating condition identification device inputs three operating condition results with discrete category labels, which are output by the radial basis function kernel support vector machine model, the one-dimensional convolutional neural network model, and the transformer model, into the hard label voting mechanism. The hard label voting mechanism filters out the operating condition results that are consistent with the three operating condition results with discrete category labels according to the label type.

[0106] S302, Determine the operating conditions of pump equipment based on the operating results.

[0107] In this embodiment, the operating condition identification device determines the proportion of consistent operating condition results. If the proportion of consistent operating condition results is greater than or equal to two-thirds, the operating condition of the pump equipment is determined based on the consistent operating condition results. For example, both the radial basis function kernel support vector machine model and the one-dimensional convolutional neural network model determine that the pump equipment is in an misaligned operating condition, but the converter model determines that the pump equipment is in an unbalanced operating condition. Therefore, the operating condition identification device ultimately determines that the pump equipment is in an misaligned operating condition. If the proportion of consistent operating condition results is less than two-thirds, the operating condition identification device can determine the operating condition of the pump equipment based on the operating condition result with the highest confidence among the three operating condition results. If the highest confidence among the three operating condition results is lower than a preset confidence threshold (e.g., the preset confidence threshold is 0.6 and the highest confidence is 0.5), the operating condition identification device marks this operating condition as uncertain, and manual inspection is required to determine the operating condition of the pump equipment.

[0108] In an exemplary embodiment, the phrase "performing feature calculation based on process detection quantity to obtain a fixed-length feature vector of the second dimension" in S202 above includes, for example: Figure 7 As shown, it includes:

[0109] S401 selects motor current, motor speed, and total flow as three sensitive channels from the process detection quantities.

[0110] In this embodiment, after obtaining the process detection quantities in the first dimension, the working condition identification device, in order to retain the most discriminative information under small sample conditions, uses mutual information metric to sort the process detection quantities in the first dimension in descending order, that is, sorts the 5-dimensional process detection quantities in descending order. After obtaining the sorting result, the top three key process detection quantities are selected as sensitive channels based on the sorting result, such as the motor operating voltage. Motor current Total flow Input power Motor speed The 5-dimensional process detection quantities were sorted in descending order using mutual information metrics, and the top three were selected: motor current. Motor speed N and total flow rate Q are used as sensitive channels.

[0111] S402 calculates the third-dimensional interval features of the monitoring sequence channel by channel based on the three sensitive channels.

[0112] The third dimension is smaller than the first dimension.

[0113] In this embodiment of the application, the operating condition identification device obtains the motor current. After identifying the three sensitive channels—motor speed N and total flow rate Q—the third-dimensional interval features of the monitoring sequence on each sensitive channel are extracted. For example, for each sensitive channel, the transient impact peak, steady-state operating mean, minimum load trough, and total fluctuation range of the monitoring sequence are calculated. The transient impact peak is determined by selecting the time window of the motor startup or sudden change in operating conditions within the monitoring sequence, and filtering out the point with the largest value by traversing all data points within this window. The steady-state operating mean is determined by first identifying the stage of stable equipment operation in the monitoring sequence, and then calculating the arithmetic mean of all data points within this stage. The minimum load trough is determined by locating the time interval of low-load operation of the equipment in the monitoring sequence and finding the data point with the smallest value within it.

[0114] The total fluctuation span is calculated by directly measuring the difference between the maximum and minimum values ​​in the entire monitoring sequence.

[0115] S403 maps the monitoring sequence to a fixed-length feature vector in the second dimension based on the three sensitive channels and the interval features in the third dimension.

[0116] In this embodiment of the application, the operating equipment is based on the motor current. The system uses three sensitive channels (motor speed N and total flow Q) and four interval features (transient impact peak, steady-state average, minimum load trough, and overall fluctuation range) to map monitoring sequences of any length into a fixed-length feature vector of the second dimension. For example, if a monitoring sequence is obtained from a single start-stop cycle of 70 seconds, then using the three sensitive channels and the interval features of the third dimension, the 70-second monitoring sequence can be mapped into a fixed-length feature vector of the second dimension, i.e., a 12-dimensional fixed-length feature vector. If another monitoring sequence is obtained from a single start-stop cycle of 10 seconds, then using the three sensitive channels and the interval features of the third dimension, the 10-second monitoring sequence can also be mapped into a fixed-length feature vector of the second dimension, i.e., a 12-dimensional fixed-length feature vector. Although the two monitoring sequences have different durations, they can both be transformed into 12-dimensional fixed-length vectors through a consistent feature extraction method.

[0117] In an exemplary embodiment, the "heterogeneous lightweight model" in S203 above includes: a radial basis function kernel support vector machine model, a one-dimensional convolutional neural network model, and a transformer model; Figure 2 The specific implementation of S204 in the embodiment, "using a preset voting mechanism to fuse multiple operating condition results of pump equipment to determine the operating condition of the pump equipment," is as follows: Figure 8 As shown, this includes: using a preset voting mechanism to merge the results of the first working condition, the second working condition, and the third working condition to determine the working condition of the pump equipment.

[0118] Among them, the radial basis function kernel support vector machine model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors and output the first operating condition result; the one-dimensional convolutional neural network model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors and output the second operating condition result; and the converter model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors and output the third operating condition result.

[0119] In this embodiment, the operating condition identification device obtains three operating condition results based on three models: a first operating condition result using a radial basis function kernel support vector machine model; a second operating condition result using a one-dimensional convolutional neural network model; and a third operating condition result using a transformer model. The operating condition identification device inputs the first, second, and third operating condition results into a preset voting mechanism, and fuses these results using the preset voting mechanism to ultimately determine the operating condition of the pump equipment. The voting process is as follows: Figure 9 As shown, the automatic identification system for multiple operating conditions of pump equipment is as follows: Figure 10 As shown. This application embodiment relates to a specific method for fusing the results of a first operating condition, a second operating condition, and a third operating condition using a preset voting mechanism to ultimately determine the operating condition of pump equipment. This method is similar to the aforementioned... Figures 2-7 The methods described in any implementation are basically the same, and for details, please refer to the foregoing explanation, which will not be repeated here.

[0120] In one exemplary embodiment, the method for automatic identification of multiple operating conditions of pump equipment further includes, as follows: Figure 11 As shown:

[0121] S501 initializes the weights of the initial radial basis function kernel support vector machine model, the weights of the one-dimensional convolutional neural network model, and the weights of the transformer model.

[0122] In this embodiment, the working condition identification device uses a five-fold cross-validation method to train the initial radial basis function kernel support vector machine model, the initial one-dimensional convolutional neural network model, and the initial converter model. The monitoring sample data of the pump equipment is divided into five equal parts, four of which are used to train the initial radial basis function kernel support vector machine model, the initial one-dimensional convolutional neural network model, and the initial converter model, and one part is used to test the trained initial radial basis function kernel support vector machine model, the initial one-dimensional convolutional neural network model, and the initial converter model. Before each part starts training, the weights of the model must be initialized.

[0123] S502, based on the monitoring sample data of pump equipment, train the initialized initial radial basis function kernel support vector machine model, the initialized initial one-dimensional convolutional neural network model, and the initialized initial converter model to obtain the trained radial basis function kernel support vector machine model, one-dimensional convolutional neural network model, and converter model.

[0124] In this embodiment of the application, the working condition identification device divides the collected monitoring sample data of pump equipment (3 to 5 samples for each working condition) into 5 parts on average. Four parts are used to train the initialized initial radial basis function kernel support vector machine model, the initialized initial one-dimensional convolutional neural network model, and the initialized initial converter model. One part is used to test the trained radial basis function kernel support vector machine model, one-dimensional convolutional neural network model, and converter model. For the initialized initial radial basis function kernel support vector machine model, the parameters are configured as follows: adaptive kernel width scaling, fixed penalty coefficient of 1.0, and probabilistic output enabled. For the initialized initial one-dimensional convolutional neural network model, a 12-dimensional fixed-length feature vector is used as input. The parameters are configured as follows: sequentially passing through two one-dimensional convolutional channel expansions and batch normalization, activation by a linear rectified function, and 50% random deactivation, followed by global average pooling for dimensionality reduction, and finally mapping to seven classes through two fully connected layers. Training is performed for 100 epochs with a batch size of 8, using an Adam optimizer with a learning rate of 1e-3 and supplemented by 1e-3 weight decay. For the initialized transformer model, a 12-dimensional fixed-length feature vector is used as input. The parameters are configured as follows: linear mapping superimposed with 0.05-scale position encoding before being fed into a single-layer single-head self-attention encoder, with a hidden dimension of 16, a feedforward dimension of 16, and a deactivation rate of 0.6. The final output is classified through global pooling and two fully connected layers. The hyperparameters are consistent with the initial one-dimensional convolutional neural network model. The training loss is as follows: Figure 12 As shown, the verification loss is as follows Figure 13 As shown, the training accuracy is as follows: Figure 14 As shown, the verification accuracy is as follows: Figure 15 As shown, the model accuracy is as follows: Figure 16 As shown.

[0125] In summary, based on all the above embodiments, a method for automatic identification of multiple operating conditions of pump equipment is also provided, such as... Figure 17 As shown, the method includes:

[0126] S601, collect the monitoring sequence of pump equipment and obtain the process detection quantity of the first dimension from it;

[0127] S602 selects motor current, motor speed and total flow as three sensitive channels from process detection quantities;

[0128] S603, calculates the interval characteristics of the third dimension of the monitoring sequence channel by channel based on three sensitive channels;

[0129] S604, based on the three sensitive channels and the interval features of the third dimension, maps the monitoring sequence into a fixed-length feature vector of the second dimension;

[0130] S605 executes S606 using a radial basis function kernel support vector machine model, S607 uses a one-dimensional convolutional neural network model, and S608 uses a transformer model, outputting the results for the first working condition, the second working condition, and the third working condition, respectively; the three models are trained by S614-S615.

[0131] S606, the radial basis function kernel support vector machine model identifies the operating conditions of pump equipment based on fixed-length feature vectors and outputs the first operating condition result;

[0132] S607, a one-dimensional convolutional neural network model identifies the operating conditions of pump equipment based on a fixed-length feature vector and outputs the second operating condition result;

[0133] S608, the converter model identifies the operating conditions of pump equipment based on a fixed-length feature vector and outputs the third operating condition result;

[0134] S609, the operating condition of pump equipment is determined by a preset voting mechanism. If a hard-label voting mechanism is used, S610-S611 are executed; if a soft-label voting mechanism is used, S612 is executed.

[0135] S610, based on the hard label voting mechanism, a two-thirds majority vote is used to select the working condition result that is consistent with the first working condition result, the second working condition result and the third working condition result. When the proportion of the working condition result that is consistent with the working condition result is greater than or equal to two-thirds, the working condition of the pump equipment is determined according to the working condition result that is consistent with the working condition result.

[0136] S611, based on the hard-label voting mechanism, a two-thirds majority vote is used to select the working condition result that is consistent with the first working condition result, the second working condition result, and the third working condition result. When the proportion of the working condition result that is consistent with the first working condition result is less than two-thirds, the working condition of the pump equipment is determined according to the working condition result with the highest confidence among the multiple working condition results; the confidence is greater than the preset confidence threshold.

[0137] S612, the average of the results of the first working condition, the second working condition, and the third working condition is calculated based on the soft label voting mechanism, and the working condition of the pump equipment is determined based on the working condition results after the average calculation.

[0138] S613 yields 7 categories of working conditions;

[0139] S614, Initialize the weights of the initial radial basis function kernel support vector machine model, the weights of the one-dimensional convolutional neural network model, and the weights of the transformer model;

[0140] S615, based on the monitoring sample data of pump equipment, train the initialized initial radial basis function kernel support vector machine model, the initialized initial one-dimensional convolutional neural network model, and the initialized initial converter model to obtain the trained radial basis function kernel support vector machine model, one-dimensional convolutional neural network model, and converter model.

[0141] The methods described in each of the above steps have been described in the foregoing embodiments. For details, please refer to the foregoing descriptions. They will not be repeated here.

[0142] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0143] Based on the same inventive concept, this application also provides an automatic multi-condition identification device for pump equipment to implement the above-mentioned automatic multi-condition identification method for pump equipment. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the automatic multi-condition identification device for pump equipment provided below can be found in the limitations of the automatic multi-condition identification method for pump equipment above, and will not be repeated here.

[0144] In one exemplary embodiment, such as Figure 18 As shown, a multi-condition automatic identification device for pump equipment is provided, comprising:

[0145] The acquisition module 11 is used to acquire the monitoring sequence of pump equipment and obtain the process detection quantity of the first dimension from it;

[0146] The calculation module 12 is used to perform feature calculation based on the process detection quantity to obtain a fixed-length feature vector in the second dimension;

[0147] The identification module 13 is used to identify the operating conditions of pump equipment based on a fixed-length feature vector using a heterogeneous lightweight model, and to obtain various operating condition results of the pump equipment.

[0148] The fusion module 14 is used to fuse the results of multiple operating conditions of pump equipment using a preset voting mechanism to determine the operating condition of the pump equipment.

[0149] In one embodiment, the fusion module 14 includes:

[0150] The first fusion unit is used to filter multiple operating condition results based on a two-thirds majority vote using a hard-label voting mechanism, and to determine the operating condition of pump equipment based on the filtered operating condition results.

[0151] In one embodiment, the fusion module 14 further includes:

[0152] The second fusion unit is used to perform average calculation on the results of multiple operating conditions based on the soft label voting mechanism, and to determine the operating condition of pump equipment based on the average calculation results.

[0153] In one embodiment, the first fusion unit includes:

[0154] The filtering subunit is used to filter out the consistent working condition results from multiple working condition results based on a two-thirds majority vote of the hard label voting mechanism.

[0155] The sub-unit is determined based on the operating conditions of the pump equipment.

[0156] In one embodiment, the aforementioned determining subunit is specifically used to: determine the operating condition of the pump equipment based on the consistent operating condition results when the proportion of consistent operating condition results is greater than or equal to two-thirds; and determine the operating condition of the pump equipment based on the operating condition result with the highest confidence among multiple operating condition results when the proportion of consistent operating condition results is less than two-thirds; wherein the confidence level is greater than a preset confidence threshold.

[0157] In one embodiment, the computing module 12 includes:

[0158] The first calculation unit is used to select motor current, motor speed and total flow as three sensitive channels from the process detection quantities;

[0159] The second calculation unit is used to calculate the interval features of the third dimension of the monitoring sequence channel by channel based on the three sensitive channels; the third dimension is smaller than the first dimension.

[0160] The third calculation unit is used to map the monitoring sequence into a fixed-length feature vector of the second dimension based on the three sensitive channels and the interval features of the third dimension.

[0161] In one embodiment, the first computing unit includes:

[0162] The descending sub-unit is used to sort the process detection quantities of the first dimension in descending order using mutual information metric to obtain the sorting result;

[0163] The channel subunit is used to select motor current, motor speed, and total flow as three sensitive channels from the process detection quantities in the first dimension based on the sorting results.

[0164] In one embodiment, the heterogeneous lightweight model in the identification module 13 includes:

[0165] A radial basis function kernel support vector machine model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors and output the first operating condition result.

[0166] A one-dimensional convolutional neural network model is used to identify the operating conditions of pump equipment based on a fixed-length feature vector and output the second operating condition result.

[0167] The converter model is used to identify the operating conditions of pump equipment based on a fixed-length feature vector and output the third operating condition result.

[0168] In one embodiment, the fusion module 14 further includes:

[0169] The results of the first, second, and third operating conditions are merged using a preset voting mechanism to determine the operating condition of the pump equipment.

[0170] In one embodiment, the identification module 13 further includes:

[0171] The initial unit is used to initialize the weights of the initial radial basis function kernel support vector machine model, the weights of the one-dimensional convolutional neural network model, and the weights of the transformer model;

[0172] The training unit is used to train the initialized initial radial basis function kernel support vector machine model, the initialized initial one-dimensional convolutional neural network model, and the initialized initial converter model based on the monitoring sample data of pump equipment, so as to obtain the trained radial basis function kernel support vector machine model, one-dimensional convolutional neural network model, and converter model.

[0173] Each module in the aforementioned automatic multi-condition identification device for pump equipment 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.

[0174] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 19 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. 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 and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a multi-condition automatic identification method for pump-type equipment. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0175] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0176] Collect monitoring sequences of pump-type equipment and obtain the first dimension of process detection quantity from them;

[0177] Feature calculation is performed based on the process detection quantity to obtain a fixed-length feature vector in the second dimension; the second dimension is greater than the first dimension.

[0178] The heterogeneous lightweight model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors, and the results of various operating conditions of pump equipment are obtained.

[0179] By using a pre-set voting mechanism to integrate the results of various operating conditions of pump equipment, the operating condition of the pump equipment can be determined.

[0180] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0181] Collect monitoring sequences of pump-type equipment and obtain the first dimension of process detection quantity from them;

[0182] Feature calculation is performed based on the process detection quantity to obtain a fixed-length feature vector in the second dimension; the second dimension is greater than the first dimension.

[0183] The heterogeneous lightweight model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors, and the results of various operating conditions of pump equipment are obtained.

[0184] By using a pre-set voting mechanism to integrate the results of various operating conditions of pump equipment, the operating condition of the pump equipment can be determined.

[0185] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0186] Collect monitoring sequences of pump-type equipment and obtain the first dimension of process detection quantity from them;

[0187] Feature calculation is performed based on the process detection quantity to obtain a fixed-length feature vector in the second dimension; the second dimension is greater than the first dimension.

[0188] The heterogeneous lightweight model is used to identify the operating conditions of pump equipment based on fixed-length feature vectors, and the results of various operating conditions of pump equipment are obtained.

[0189] By using a pre-set voting mechanism to integrate the results of various operating conditions of pump equipment, the operating condition of the pump equipment can be determined.

[0190] Those skilled in the art will understand that all or part of the processes in 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, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0191] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0192] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A multi-working condition automatic identification method for pump-type equipment, characterized in that, The method includes: Collect monitoring sequences of pump-type equipment and obtain the first dimension of process detection quantity from them; Feature calculation is performed based on the process detection quantity to obtain a fixed-length feature vector with a second dimension; the second dimension is greater than the first dimension; The operating conditions of pump equipment are identified using a heterogeneous lightweight model based on the fixed-length feature vector, resulting in various operating condition results for the pump equipment. By using a pre-set voting mechanism to integrate the results of various operating conditions of pump equipment, the operating condition of the pump equipment can be determined.

2. The method according to claim 1, wherein the preset voting mechanism includes a hard-label voting mechanism, characterized in that, The process of using a preset voting mechanism to fuse multiple operating condition results of the pump equipment to determine the operating condition of the pump equipment includes: The results of the various operating conditions are filtered by a two-thirds majority vote based on a hard-label voting mechanism, and the operating condition of the pump equipment is determined based on the filtered results.

3. The method of claim 1, wherein the pre-set voting mechanism further comprises a soft label voting mechanism, and wherein the soft label voting mechanism comprises: determining a soft label for each of the plurality of candidate images based on the plurality of candidate images; and determining a final label for each of the plurality of candidate images based on the soft label for each of the plurality of candidate images. The process of using a preset voting mechanism to fuse multiple operating condition results of the pump equipment to determine the operating condition of the pump equipment includes: The average of the results of the various operating conditions is calculated based on a soft-label voting mechanism, and the operating condition of the pump equipment is determined based on the average result.

4. The method of claim 2, wherein, The two-thirds majority vote based on the hard-label voting mechanism filters the results of the various operating conditions, and determines the operating condition of the pump equipment based on the filtered results, including: Based on the two-thirds majority vote of the hard-label voting mechanism, the working condition result that is consistent with the working condition result is selected from the multiple working condition results. The operating conditions of the pump equipment are determined based on the operating results.

5. The method of claim 4, wherein, Determining the operating condition of the pump equipment based on the operating condition results includes: When the proportion of consistent operating conditions is greater than or equal to two-thirds, the operating condition of the pump equipment is determined based on the consistent operating conditions. When the proportion of consistent operating condition results is less than two-thirds, the operating condition of the pump equipment is determined based on the operating condition result with the highest confidence among the various operating condition results; the confidence level is greater than a preset confidence threshold.

6. The method of claim 1, wherein, The step of calculating features based on the process detection quantity to obtain a fixed-length feature vector of the second dimension includes: Motor current, motor speed, and total flow rate were selected as three sensitive channels from the process detection parameters. The interval feature of the third dimension of the monitoring sequence is calculated channel by channel based on the three sensitive channels; the third dimension is smaller than the first dimension; Based on the three sensitive channels and the interval features of the third dimension, the monitoring sequence is mapped to a fixed-length feature vector of the second dimension.

7. The method of claim 6, wherein, The method further includes: The process detection quantities of the first dimension are sorted in descending order using mutual information metric to obtain the ranking result; The process selects motor current, motor speed, and total flow rate as three sensitive channels from the process detection quantities, including: Based on the sorting results, the motor current, the motor speed, and the total flow rate are selected as three sensitive channels from the process detection quantities of the first dimension.

8. The method according to any one of claims 1 to 7, characterized in that, The heterogeneous lightweight model includes a radial basis function kernel support vector machine model, a one-dimensional convolutional neural network model, and a transformer model; The radial basis function kernel support vector machine model is used to identify the operating conditions of the pump equipment based on the fixed-length feature vector and output the first operating condition result. The one-dimensional convolutional neural network model is used to identify the operating conditions of the pump equipment based on the fixed-length feature vector and output a second operating condition result. The converter model is used to identify the operating conditions of the pump equipment based on the fixed-length feature vector and output a third operating condition result. The process of using a preset voting mechanism to fuse multiple operating condition results of pump equipment to determine the operating condition of the pump equipment includes: The operating conditions of the pump equipment are determined by fusing the results of the first operating condition, the second operating condition, and the third operating condition using a preset voting mechanism.

9. The method of claim 8, wherein, The method further includes: Initialize the weights of the initial radial basis function kernel support vector machine model, the initial weights of the initial one-dimensional convolutional neural network model, and the initial weights of the initial transformer model; The initial radial basis function kernel support vector machine model, the initial one-dimensional convolutional neural network model, and the initial converter model are trained based on the monitoring sample data of the pump equipment to obtain the trained radial basis function kernel support vector machine model, the one-dimensional convolutional neural network model, and the converter model.

10. A pump type equipment multi-working condition automatic identification device, characterized in that, The device includes: The data acquisition module is used to acquire monitoring sequences of pump-type equipment and obtain the first-dimensional process detection quantity from them. The calculation module is used to perform feature calculation based on the process detection quantity to obtain a fixed-length feature vector of the second dimension; The identification module is used to identify the operating conditions of pump equipment based on the fixed-length feature vector using a heterogeneous lightweight model, and obtain various operating condition results of the pump equipment. The fusion module is used to merge the results of multiple operating conditions of pump equipment using a preset voting mechanism to determine the operating condition of the pump equipment.

Images