Real-time monitoring system and method for oil pump status based on sensor network

Abstract

The present invention discloses a real-time monitoring system and method for oil pump status based on a sensor network, and relates to the technical field of oil pump status monitoring. The system collects fault labels and node status training vectors of each test oil pump at each sample collection node, constructs an oil pump adjacency matrix based on oil pump connection structure data, constructs a status monitoring graph model based on the oil pump adjacency matrix, and trains the status monitoring graph model based on the node status training vectors and fault labels of the test oil pump. The real-time status data of the oil pump group to be monitored is collected, and the real-time status data is input into the status monitoring graph model to obtain a fault prediction value vector output by the status monitoring graph model, and fault assessment and fault location are performed according to the fault prediction value vector; thus, the accuracy of fault prediction and the monitoring efficiency in the operation scenario of multiple oil pumps are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of oil pump status monitoring, and in particular to a sensor network-based real-time oil pump status monitoring system and method. Background Art

[0002] During oilfield exploitation, multiple oil pumps are usually required to work together to improve oil extraction efficiency. These oil pumps have complex connections, including series and parallel connections. The failure of a single oil pump will affect the operating status of the entire oil pump group, so it is necessary to monitor the operating status of each oil pump in real time to detect potential failures in time and avoid causing greater failures and losses.

[0003] However, the existing technology has some shortcomings in achieving real-time monitoring of the working status of the oil pump assembly:

[0004] Traditional monitoring methods cannot fully utilize the connection relationship information between oil pumps. Each oil pump is monitored independently, making it difficult to accurately evaluate the overall operating status.

[0005] The existing models cannot handle the mutual influence between oil pumps well, which affects the accuracy of fault prediction.

[0006] A Chinese patent with authorization announcement number CN111306051B discloses a probe-type condition monitoring and early warning method, device and system for an oil pump unit, the method comprising obtaining monitoring data of multiple monitoring parameters of the oil pump unit, the monitoring parameters comprising operating parameters and vibration parameters; wherein the monitoring data of the operating parameters are obtained from the SCADA subsystem of the oil pump unit, and the monitoring data of the vibration parameters are obtained from the vibration monitoring subsystem of the oil pump unit based on the probe; wherein the condition monitoring subsystem is connected to an open interface of the vibration monitoring subsystem via a probe; the operating condition of the oil pump unit is monitored using the monitoring data of the multiple monitoring parameters to obtain the operating condition monitoring result of the oil pump unit, but the method still monitors a single oil pump and fails to effectively handle the condition monitoring problem of the combined operation of multiple oil pumps.

[0007] To this end, the present invention proposes a real-time monitoring system and method for an oil pump status based on a sensor network. Summary of the invention

[0008] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a real-time monitoring system and method for oil pump status based on a sensor network to improve the accuracy of fault prediction and monitoring efficiency in the operation scenario of multiple oil pumps.

[0009] To achieve the above purpose, a real-time monitoring method for oil pump status based on sensor network is proposed, which includes the following steps:

[0010] Step 1: Collect the oil pump connection structure data of the test oil pump group in advance, and install monitoring sensors for each test oil pump;

[0011] Step 2: When the oil pump group is tested for oil delivery each time, the sensor data collected by each monitoring sensor is collected in real time; and the fault labels of each test oil pump at each sample collection node are collected; for each test oil pump, the sensor data of each sample collection node constitute a node state training vector;

[0012] Step 3: Based on the oil pump connection structure data, construct the oil pump adjacency matrix;

[0013] Step 4: Based on the oil pump adjacency matrix, a state monitoring graph model is constructed, and the state monitoring graph model is trained based on the state training vectors and fault labels of each node of the test oil pump;

[0014] Step 5: Collect the real-time status data of the oil pump group to be monitored, and input the real-time status data into the status monitoring graph model to obtain the fault prediction value vector output by the status monitoring graph model, and perform fault assessment and fault location based on the fault prediction value vector.

[0015] The method of collecting the sensor data collected by each monitoring sensor in real time during each test of the oil delivery pump group is as follows:

[0016] Each time the oil pump group is tested for oil delivery, each monitoring sensor is started.

[0017] The data collected by each monitoring sensor is labeled and differentiated, and then transmitted to the data processing background via wireless network.

[0018] The method of collecting the fault labels of each test oil pump of each sample collection node is as follows:

[0019] The duration of the sample collection cycle is determined in advance, and the time nodes every time of the collection cycle are used as sample collection nodes.

[0020] After the test oil pump group delivers oil, the operating status of each test oil pump is manually determined at each sample collection node, and the operating status is used as the fault label of the test oil pump.

[0021] The fault label is an integer from 0 to M-1, where M is the number of operating states, and each integer corresponds to an operating state.

[0022] The method of constructing the oil pump adjacency matrix based on the oil pump connection structure data is as follows:

[0023] According to the connection relationship between the test oil pumps in the oil pump connection structure data, a connection label is set for each combination of two oil pumps.

[0024] The connection tag is set as follows:

[0025] The two test oil pumps are marked as i and j respectively, and the connection label Wij is set to 0 or 1, 0 means that the jth test oil pump is not connected to the ith test oil pump, and 1 means that the jth test oil pump is connected to the ith test oil pump.

[0026] The adjacency matrix of the oil pump is N N matrix, the i-th row and j-th column in the oil pump adjacency matrix corresponds to the connection label Wij.

[0027] The construction of the state monitoring graph model based on the oil pump adjacency matrix includes the following steps:

[0028] Step 11: Define a node update function, receive the state training vector of the current oil pump node and the state training vectors of other oil pump nodes connected to it, and calculate a new node state vector.

[0029] The node update function is defined as follows:

[0030] ;

[0031] Among them, i is the number of each oil pump node in the state monitoring diagram model.

[0032] in, and are the feature vectors of the i-th oil pump node at the k-th layer and the k+1-th layer respectively.

[0033] is the feature vector of the j-th oil pump node in the k-th layer.

[0034] k is the number of iterations.

[0035] σ is a nonlinear activation function, such as the ReLU function.

[0036] W_s and b_s are the trainable weights and bias parameters of the oil pump node.

[0037] is the set of adjacent nodes of the i-th oil pump node, that is, Includes the numbers corresponding to all columns whose values ​​are 1 in the i-th row of the oil pump adjacency matrix.

[0038] α_ij is the attention weight from oil pump node j to oil pump node i, and the calculation formula is as follows:

[0039] α_ij = softmax_j(e_ij) = ;softmax_j is the softmax function;

[0040] e_ij = LeakyReLU a T × [W_h × h_i k || W_h × h_j k ]) .in, is the LeakyRelu activation function, a is the trainable attention vector, T represents the transpose, and W_h is the trainable weight matrix.

[0041] Step 12: Define the edge update function, receive the states of the two nodes connected to the current oil pump connection edge, and calculate the new edge state vector.

[0042] The edge update function is defined as:

[0043] .

[0044] in, is the edge state vector connecting nodes i and j in the k+1th layer; φ is the nonlinear activation function; W_e and b_e are the trainable weights and bias parameters of the edges between the oil pump nodes.

[0045] Step 13: Design the GNN model architecture, including the node update module and the edge update module. The node update module updates the state training vector of each oil pump node according to the node update function.

[0046] Step 14: Define a loss function that compares the final state vector output of each node with the label data to reflect the overall health of the system.

[0047] The loss function is defined as follows:

[0048] L =- ∑ i [y_i ×log( y_i ̂ )+(1-y_i)×log(1- y_i ̂ ) ] .

[0049] Where L is the loss function, is the predicted value of the fault label of the i-th node output by the model, is the fault label of the ith node.

[0050] in, The calculation formula is: ; Where W_o is the weight matrix of the output layer and b_o is the bias vector of the output layer.

[0051] The method of training the state monitoring graph model based on the state training vectors and fault labels of each node of the test oil pump is as follows:

[0052] The node state training vectors of each test oil pump collected by each sample collection node are combined into a set of model input data sets.

[0053] For each set of input data sets, the node state training vectors of each test oil pump are input into the corresponding oil pump nodes of the state monitoring graph model.

[0054] The state monitoring graph model updates parameters according to the input data set using node update functions and edge update functions, calculates the loss function, and continuously iterates and updates the loss function until the training stops when the loss function converges.

[0055] The method of performing fault assessment and fault location according to the fault prediction value vector is as follows:

[0056] If the operation state represented by the predicted value of any fault label in the fault prediction value vector is not normal, it is determined that the operation of the monitored oil pump group is abnormal.

[0057] From all the oil pumps to be monitored, find out the oil pump to be monitored whose operating status is an abnormal source, and use it as the abnormal source oil pump.

[0058] From all the oil pumps to be monitored, find out the oil pumps to be monitored whose operating status is not normal or is the source of abnormality, and perform corresponding fault detection on them respectively.

[0059] A real-time monitoring system for oil pump status based on sensor network is proposed, which includes training data collection module, graph network model training module and status monitoring module; wherein each module is connected electrically;

[0060] The training data collection module collects the oil pump connection structure data of the test oil pump group in advance, and installs a monitoring sensor for each test oil pump. During each test of the oil pump group, the sensor data collected by each monitoring sensor is collected in real time; and the fault labels of each test oil pump of each sample collection node are collected; for each test oil pump, the sensor data of each sample collection node form a node state training vector, and the oil pump adjacency matrix is ​​constructed based on the oil pump connection structure data; and the oil pump adjacency matrix, node state training vector and fault label are sent to the graph network model training module;

[0061] The graph network model training module builds a state monitoring graph model based on the oil pump adjacency matrix, trains the state monitoring graph model based on the state training vectors and fault labels of each node of the test oil pump, and sends the state monitoring graph model to the state monitoring module;

[0062] The condition monitoring module collects the real-time condition data of the oil pump group to be monitored, and inputs the real-time condition data into the condition monitoring graph model to obtain the fault prediction value vector output by the condition monitoring graph model, and performs fault assessment and fault location according to the fault prediction value vector.

[0063] An electronic device is proposed, comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0064] The processor executes the above-mentioned real-time monitoring method of the oil pump status based on the sensor network by calling the computer program stored in the memory.

[0065] A computer-readable storage medium is provided, on which a rewritable computer program is stored;

[0066] When the computer program is executed on a computer device, the computer device executes the above-mentioned method for real-time monitoring of the state of an oil pump based on a sensor network.

[0067] Compared with the prior art, the present invention has the following beneficial effects:

[0068] The present invention collects the oil pump connection structure data of the test oil pump group in advance, and installs a monitoring sensor for each test oil pump. When the test oil pump group delivers oil each time, the sensor data collected by each monitoring sensor is collected in real time; and the fault label of each test oil pump of each sample collection node is collected; for each test oil pump, each sensor data of each sample collection node constitutes a node state training vector, and an oil pump adjacency matrix is ​​constructed based on the oil pump connection structure data. Based on the oil pump adjacency matrix, a state monitoring graph model is constructed, and based on the node state training vectors and fault labels of the test oil pump, the state monitoring graph model is trained, and the real-time state data of the oil pump group to be monitored is collected, and the real-time state data is input into the state monitoring graph model to obtain the fault prediction value vector output by the state monitoring graph model, and the fault evaluation and fault location are performed according to the fault prediction value vector; the GNN model makes full use of the connection topology structure information between the pumps to improve the accuracy of fault prediction, and the fault location capability based on the graph structure can quickly and accurately locate the faulty pump, thereby improving the efficiency of fault handling. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1This is a flow chart of a method for real-time monitoring of oil pump status based on a sensor network in Example 1 of the present invention.

[0070] Figure 2 This is an example diagram of the serial connection of oil pumps in Example 1 of the present invention.

[0071] Figure 3 This is an example diagram of the parallel connection of oil pumps in Example 1 of the present invention.

[0072] Figure 4 This is a module connection diagram of the sensor network-based oil pump status real-time monitoring system in Example 2 of the present invention.

[0073] Figure 5 This is a schematic diagram of the structure of an electronic device in Example 3 of the present invention.

[0074] Figure 6 This is a schematic diagram of the computer-readable storage medium structure in Example 4 of the present invention. DETAILED DESCRIPTION

[0075] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0076] Example 1

[0077] like Figure 1 As shown, the real-time monitoring method of the oil pump state based on the sensor network includes the following steps:

[0078] Step 1: Collect the oil pump connection structure data of the test oil pump group in advance, and install monitoring sensors for each test oil pump;

[0079] Step 2: When the oil pump group is tested for oil delivery each time, the sensor data collected by each monitoring sensor is collected in real time; and the fault labels of each test oil pump at each sample collection node are collected; for each test oil pump, the sensor data of each sample collection node constitute a node state training vector;

[0080] Step 3: Based on the oil pump connection structure data, construct the oil pump adjacency matrix;

[0081] Step 4: Based on the oil pump adjacency matrix, a state monitoring graph model is constructed, and the state monitoring graph model is trained based on the training vectors and fault labels of each node state of the test oil pump;

[0082] Step 5: Collect the real-time status data of the oil pump group to be monitored, and input the real-time status data into the status monitoring graph model to obtain the fault prediction value vector output by the status monitoring graph model, and perform fault assessment and fault location based on the fault prediction value vector.

[0083] The method of pre-collecting the oil pump connection structure data of the oil pump group for testing is as follows:

[0084] Select N oil pumps as test oil pumps, and connect the test oil pumps in series or in parallel according to the oil delivery demand to form a test oil pump group; N is the number of the selected test oil pumps.

[0085] It should be noted that the series or parallel connection is an organizational form between the oil pumps to achieve a better oil delivery effect, such as Figure 2 As shown in FIG. 1 , an example diagram of oil pumps connected in series is shown. The series connection method can increase the total flow rate and total pressure head to increase the transmission distance. Figure 3 The figure shows an example of oil pumps connected in parallel. The parallel connection method can perform oil delivery operations at different locations simultaneously, thereby improving the oil delivery efficiency.

[0086] Furthermore, the method of installing a monitoring sensor for each test oil pump is as follows:

[0087] A monitoring sensor corresponding to each physical state is installed for each test oil pump.

[0088] The physical state includes but is not limited to operating data such as temperature, vibration, pressure, flow, etc. The installation position of each monitoring sensor is determined according to the actual structure of the test oil pump.

[0089] Furthermore, the method of collecting the sensor data collected by each monitoring sensor in real time during each test of the oil delivery pump group is as follows:

[0090] Each time the oil pump group is tested for oil delivery, each monitoring sensor is started.

[0091] The data collected by each monitoring sensor are labeled and differentiated, and then transmitted to the data processing background via a wireless network; the labeling and differentiation refers to labeling the data collected by each monitoring sensor with the test oil pump number and the monitoring sensor number to distinguish it from the data collected by other monitoring sensors.

[0092] Furthermore, the method of collecting the fault labels of each test oil pump of each sample collection node is as follows:

[0093] The duration of the sample collection cycle is determined in advance, and the time nodes every time of the collection cycle are used as sample collection nodes.

[0094] After the test oil pump group delivers oil, the operating status of each test oil pump is manually determined at each sample collection node, and the operating status is used as the fault label of the test oil pump.

[0095] Specifically, the operating status of the test oil pump includes but is not limited to normal, blockage, abnormal source, pump body wear, bearing damage, etc.; it can be understood that, the abnormal operating status of a single test oil pump may not be due to a fault in the test oil pump itself, but a chain reaction caused by faults in other test oil pumps connected to it. Therefore, a state is needed to screen out the source of the fault (abnormal source), and other states other than normal and abnormal sources indicate abnormalities.

[0096] The fault label is an integer from 0 to M-1, where M is the number of operating states, and each integer corresponds to an operating state. For example, when the operating state is normal, the fault label can be set to 0.

[0097] Furthermore, the method of constructing the oil pump adjacency matrix based on the oil pump connection structure data is as follows:

[0098] According to the connection relationship between the test oil pumps in the oil pump connection structure data, a connection label is set for each combination of two oil pumps.

[0099] The connection tag is set as follows:

[0100] The two test oil pumps are marked as i and j respectively, and the connection label Wij is set to 0 or 1, 0 means that the jth test oil pump is not connected to the ith test oil pump, and 1 means that the jth test oil pump is connected to the ith test oil pump.

[0101] The adjacency matrix of the oil pump is N N matrix, the i-th row and j-th column in the oil pump adjacency matrix corresponds to the connection label Wij.

[0102] It can be understood that the oil pump adjacency matrix encodes the connection relationship information between the nodes in the network in the state monitoring graph model.

[0103] Furthermore, the construction of the state monitoring graph model based on the oil pump adjacency matrix includes the following steps:

[0104] Step 11: Define a node update function, receive the state training vector of the current oil pump node and the state training vectors of other oil pump nodes connected to it, and calculate a new node state vector.

[0105] The node update function is defined as follows:

[0106] ;

[0107] Wherein, i is the number of each oil pump node in the state monitoring diagram model; it can be understood that each oil pump node corresponds to an oil pump.

[0108] in, and are the feature vectors of the i-th oil pump node at the k-th layer and the k+1-th layer respectively.

[0109] is the feature vector of the j-th oil pump node in the k-th layer.

[0110] k is the number of iterations. It can be understood that in the message passing process of GNN, k iterations of "neighbor aggregation-node update" operations are required to enable the state vector of each node to gradually contain the global graph / network topology information.

[0111] σ is a nonlinear activation function, such as the ReLU function.

[0112] W_s and b_s are the trainable weights and bias parameters of the oil pump node.

[0113] is the set of adjacent nodes of the i-th oil pump node, that is, Includes the numbers corresponding to all columns whose values ​​are 1 in the i-th row of the oil pump adjacency matrix.

[0114] α_ij is the attention weight from oil pump node j to oil pump node i, and the calculation formula is as follows:

[0115] α_ij = softmax_j(e_ij) = ;softmax_j is the softmax function;

[0116] e_ij = LeakyReLU a T × [W_h × h_i k || W_h × h_j k ]) .in, is the LeakyRelu activation function, a is the trainable attention vector, T represents the transpose, and W_h is the trainable weight matrix.

[0117] Step 12: Define the edge update function, receive the states of the two nodes connected to the current oil pump connection edge, and calculate the new edge state vector.

[0118] The edge update function is defined as follows:

[0119] .

[0120] in, is the edge state vector connecting nodes i and j in the k+1th layer; φ is the nonlinear activation function; W_e and b_e are the trainable weights and bias parameters of the edges between the oil pump nodes.

[0121] Step 13: Design the GNN model architecture, including the node update module and the edge update module, and specify that the two are executed alternately in a certain order for multiple iterations; the node update module updates the state training vector of each oil pump node according to the node update function.

[0122] Step 14: Define a loss function that compares the final state vector output of each node with the label data to reflect the overall health of the system.

[0123] The loss function is defined as follows:

[0124] L =- ∑ i [y_i ×log( y_i ̂ )+(1-y_i)×log(1- y_i ̂ ) ] .

[0125] Where L is the loss function, is the predicted value of the fault label of the i-th node output by the model, is the fault label of the ith node.

[0126] in, The calculation formula is: ; Where W_o is the weight matrix of the output layer and b_o is the bias vector of the output layer.

[0127] Furthermore, the state monitoring graph model is trained based on the training vectors and fault labels of each node state of the test oil pump as follows:

[0128] The node state training vectors of each test oil pump collected by each sample collection node are combined into a set of model input data sets.

[0129] For each set of input data sets, the node state training vectors of each test oil pump are input into the corresponding oil pump nodes of the state monitoring graph model.

[0130] The state monitoring graph model updates parameters according to the input data set using node update functions and edge update functions, calculates the loss function, and continuously iterates and updates the loss function until the training stops when the loss function converges.

[0131] Furthermore, the method of collecting the real-time status data of the oil pump group to be monitored is:

[0132] A monitoring sensor is installed for each oil pump in the oil pump group to be monitored; it is understandable that the location where the monitoring sensor is installed is consistent with the location where the monitoring sensor on each test oil pump in the test oil pump group is installed.

[0133] During the actual operation of the oil pump group, the sensor data collected by each monitoring sensor is collected in real time, and the sensor data of each monitoring sensor of each oil pump to be monitored are combined into a set of prediction input vectors in real time.

[0134] The predicted input vectors of each monitored oil pump at each moment are combined into real-time status data.

[0135] Furthermore, the method of inputting the real-time status data into the status monitoring diagram model to obtain the fault prediction value vector output by the status monitoring diagram model is:

[0136] Each prediction input vector in the real-time state data is input into the corresponding oil pump node, and the predicted value of the fault label corresponding to each node output by the state monitoring graph model is obtained to form a fault prediction value vector.

[0137] Furthermore, the method of performing fault assessment and fault location according to the fault prediction value vector is:

[0138] If the operation state represented by the predicted value of any fault label in the fault prediction value vector is not normal, it is determined that the operation of the monitored oil pump group is abnormal.

[0139] From all the oil pumps to be monitored, find out the oil pump to be monitored whose operating status is an abnormal source, and use it as the abnormal source oil pump.

[0140] From all the oil pumps to be monitored, find out the oil pumps to be monitored whose operating status is not normal or is the source of abnormality, and perform corresponding fault detection on them respectively.

[0141] Example 2

[0142] like Figure 4 As shown, the real-time monitoring system for oil pump status based on sensor network includes a training data collection module, a graph network model training module and a status monitoring module; wherein each module is electrically connected to each other.

[0143] The training data collection module collects the oil pump connection structure data of the test oil pump group in advance, and installs a monitoring sensor for each test oil pump. During each test of the oil pump group, the sensor data collected by each monitoring sensor is collected in real time; and the fault labels of each test oil pump at each sample collection node are collected; for each test oil pump, the sensor data of each sample collection node form a node state training vector, and based on the oil pump connection structure data, the oil pump adjacency matrix is ​​constructed; and the oil pump adjacency matrix, node state training vector and fault label are sent to the graph network model training module.

[0144] The graph network model training module constructs a state monitoring graph model based on the oil pump adjacency matrix, trains the state monitoring graph model based on the state training vectors and fault labels of each node of the test oil pump, and sends the state monitoring graph model to the state monitoring module.

[0145] The condition monitoring module collects the real-time condition data of the oil pump group to be monitored, and inputs the real-time condition data into the condition monitoring graph model to obtain the fault prediction value vector output by the condition monitoring graph model, and performs fault assessment and fault location according to the fault prediction value vector.

[0146] Example 3

[0147] Figure 5 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 5 As shown, according to another aspect of the present application, an electronic device 100 is also provided. The electronic device 100 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are executed by one or more processors, the above-mentioned method for real-time monitoring of oil pump status based on a sensor network can be executed.

[0148] The method or device according to the embodiment of the present application can also be used by Figure 5 The electronic device architecture shown in FIG. Figure 5 As shown, the electronic device 100 may include a bus 101, one or more CPUs 102, a ROM 103, a RAM 104, a communication port 105 connected to a network, an input / output component 106, a hard disk 107, etc. The storage device in the electronic device 100, such as the ROM 103 or the hard disk 107, may store the real-time monitoring method of the oil pump state based on the sensor network provided in the present application.

[0149] Furthermore, the electronic device 100 may further include a user interface 108. Figure 5 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 5One or more components of an electronic device are shown.

[0150] Example 4

[0151] Figure 6 Schematic diagram of a computer-readable storage medium structure provided by an embodiment of the present application. Figure 6 As shown, it is a computer-readable storage medium 200 according to an embodiment of the present application. Computer-readable instructions are stored on the computer-readable storage medium 200. When the computer-readable instructions are executed by the processor, the real-time monitoring method of the oil pump state based on the sensor network according to the embodiment of the present application described with reference to the above figures can be executed. The computer-readable storage medium 200 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0152] In addition, according to the implementation of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, the non-transitory machine-readable storage medium stores machine-readable instructions, the machine-readable instructions can be run by a processor to execute instructions corresponding to the method steps provided by the present application, and when the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0153] The methods, apparatuses, and devices of the present application may be implemented in many ways. For example, the methods, apparatuses, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0154] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0155] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0156] The above preset parameters or preset thresholds are all set by technicians in this field according to actual conditions or obtained through large-scale data simulation.

[0157] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A real-time monitoring method for oil pump status based on a sensor network, characterized in that: The following steps are involved: Step 1: Collect the oil pump connection structure data of the test oil pump group in advance, and install monitoring sensors for each test oil pump; Step 2: When the oil pump group is tested for oil delivery each time, the sensor data collected by each monitoring sensor is collected in real time; And collect the fault labels of each test oil pump at each sample collection node; For each tested oil pump, the sensor data of each sample acquisition node constitutes a node state training vector; Step 3: Based on the oil pump connection structure data, construct the oil pump adjacency matrix; Step 4: Based on the oil pump adjacency matrix, a state monitoring graph model is constructed, and the state monitoring graph model is trained based on the training vectors and fault labels of each node state of the test oil pump; Step 5: Collect the real-time status data of the oil pump group to be monitored, and input the real-time status data into the status monitoring graph model to obtain the fault prediction value vector output by the status monitoring graph model, and perform fault assessment and fault location according to the fault prediction value vector; The method of collecting the sensor data collected by each monitoring sensor in real time during each test of the oil delivery pump group is as follows: Each time the oil delivery pump group is tested for oil delivery, each monitoring sensor is started; The data collected by each monitoring sensor is labeled and differentiated, and then transmitted to the data processing background via wireless network; The fault label is an integer from 0 to M-1, where M is the number of operating states, and each integer corresponds to an operating state; The method of constructing the oil pump adjacency matrix based on the oil pump connection structure data is as follows: According to the connection relationship between the test oil pumps in the oil pump connection structure data, a connection label is set for each combination of two oil pumps; The connection tag is set as follows: The two test oil pumps are marked as i and j respectively, and the connection label Wij is set to 0 or 1, 0 means that the jth test oil pump is not connected to the ith test oil pump, and 1 means that the jth test oil pump is connected to the ith test oil pump; The adjacency matrix of the oil pump is N N matrix, the i-th row and j-th column in the oil pump adjacency matrix corresponds to the connection label Wij; The construction of the state monitoring graph model based on the oil pump adjacency matrix includes the following steps: Step 11: define a node update function, receive the state training vector of the current oil pump node and the state training vectors of other oil pump nodes connected to it, and calculate a new node state vector; Step 12: Define an edge update function, receive the states of the two nodes connected to the current oil pump connection edge, and calculate the new edge state vector; Step 13: Design the GNN model architecture, including a node update module and an edge update module. The node update module updates the state training vector of each oil pump node according to the node update function. Step 14: Define the loss function to compare the final state vector output of each node with the label data; The method of training the state monitoring graph model based on the state training vectors and fault labels of each node of the test oil pump is as follows: The node state training vectors of each test oil pump collected by each sample collection node are combined into a set of model input data sets; For each set of input data sets, the node state training vector of each test oil pump is input into the corresponding oil pump node of the state monitoring graph model; The state monitoring graph model uses the node update function and the edge update function to update the parameters according to the input data set, and calculates the loss function. The loss function is continuously updated iteratively until the loss function reaches convergence and the training is stopped. The method of performing fault assessment and fault location according to the fault prediction value vector is as follows: If the operation state represented by the predicted value of any fault label in the fault prediction value vector is not normal, it is determined that the operation of the monitored oil pump group is abnormal; Find out the oil pump to be monitored whose operating status is an abnormal source from all the oil pumps to be monitored, and use it as the abnormal source oil pump; From all the oil pumps to be monitored, find out the oil pumps to be monitored whose operating status is not normal or is the source of abnormality, and perform corresponding fault detection on them respectively.

2. The method for real-time monitoring of oil pump status based on sensor network according to claim 1 is characterized in that: The node update function is defined as follows: ； Where i is the number of each oil pump node in the state monitoring diagram model; in, and are the feature vectors of the i-th oil pump node at the k-th layer and the k+1-th layer respectively; is the feature vector of the j-th oil pump node in the k-th layer; k is the number of iterations; σ is a nonlinear activation function; W_s and b_s are the trainable weights and bias parameters of the oil pump node; is the set of adjacent nodes of the i-th oil pump node, that is, Include the numbers corresponding to all columns whose values ​​are 1 in the i-th row of the oil pump adjacency matrix; α_ij is the attention weight from oil pump node j to oil pump node i, and the calculation formula is as follows: α_ij = softmax_j(e_ij) = ;softmax_j is the softmax function; e_ij = ;in, is the LeakyRelu activation function, a is the trainable attention vector, T represents the transpose, and W_h is the trainable weight matrix.

3. The method for real-time monitoring of oil pump status based on sensor network according to claim 2 is characterized in that: The edge update function is defined as: ； in, is the edge state vector connecting nodes i and j in the k+1th layer; φ is the nonlinear activation function; W_e and b_e are the trainable weights and bias parameters of the edges between the oil pump nodes.

4. The method for real-time monitoring of oil pump status based on sensor network according to claim 3 is characterized in that: The loss function is defined as: ； Where L is the loss function, is the predicted value of the fault label of the i-th node output by the model, is the fault label of the i-th node; in, The calculation formula is: ; Where W_o is the weight matrix of the output layer and b_o is the bias vector of the output layer.

5. A real-time monitoring system for oil pump status based on a sensor network, which is used to implement the real-time monitoring method for oil pump status based on a sensor network as claimed in any one of claims 1 to 4, characterized in that: It includes a training data collection module, a graph network model training module and a state monitoring module; wherein each module is electrically connected to each other; The training data collection module collects the oil pump connection structure data of the test oil pump group in advance, and installs a monitoring sensor for each test oil pump. During each test of the oil pump group, the sensor data collected by each monitoring sensor is collected in real time; and the fault labels of each test oil pump of each sample collection node are collected; for each test oil pump, the sensor data of each sample collection node form a node state training vector, and the oil pump adjacency matrix is ​​constructed based on the oil pump connection structure data; and the oil pump adjacency matrix, node state training vector and fault label are sent to the graph network model training module; The graph network model training module builds a state monitoring graph model based on the oil pump adjacency matrix, trains the state monitoring graph model based on the state training vectors and fault labels of each node of the test oil pump, and sends the state monitoring graph model to the state monitoring module; The condition monitoring module collects the real-time condition data of the oil pump group to be monitored, and inputs the real-time condition data into the condition monitoring graph model to obtain the fault prediction value vector output by the condition monitoring graph model, and performs fault assessment and fault location according to the fault prediction value vector.

6. An electronic device, characterized in that: include: A processor and a memory, wherein: The memory stores a computer program that can be called by the processor; The processor executes the real-time monitoring method for the oil pump status based on the sensor network as described in any one of claims 1 to 4 in the background by calling the computer program stored in the memory.

7. A computer-readable storage medium, characterized in that: A rewritable computer program is stored thereon; When the computer program is executed on a computer device, the computer device executes the real-time monitoring method for the state of an oil pump based on a sensor network as described in any one of claims 1 to 4 in the background.

Images