Bayesian network-based well site data collector fault prediction method, system and device, and storage medium

Abstract

The invention belongs to the technical field of equipment fault prediction, and particularly relates to a Bayesian network-based fault prediction method for a well site data collector, which comprises the following steps of: acquiring a physical quantity actually measured in the well site data collector, pre-processing the physical quantity, inputting the pre-processed physical quantity into a constructed fault prediction network model, and predicting the fault of the well site data collector according to the pre-processed physical quantity. And predicting the occurrence type and the occurrence probability of the fault. A fault prediction network model construction process: selecting a characteristic fault phenomenon according to historical fault information of the well site data collector, analyzing related fault causes and related influence factors of the characteristic fault phenomenon based on a related function analysis method, and obtaining a causal relationship between the characteristic fault phenomenon and the related fault causes and the related influence factors; determining a preliminary structure of the Bayesian network; and optimizing the initial structure of the Bayesian network by using a K2 algorithm, and optimizing network parameters by using an EM algorithm to obtain a constructed fault prediction network model. The problem that an existing fault prediction method is not suitable for fault prediction of the well site data collector is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of equipment failure prediction, and in particular relates to a Bayesian network-based well site data collector failure prediction method, system, equipment and storage medium. Background Art

[0002] Fault prediction refers to the use of intelligent reasoning algorithms to evaluate the fault state of equipment at a certain moment in the future. Compared with traditional fault diagnosis, it can effectively reduce the equipment failure rate. Fault prediction itself has uncertainty, and the complex failure mechanism and working environment of the well site data collector further increase the uncertainty of the failure. After searching, the following patent documents were obtained:

[0003] The patent number is CN202211580975.7, and its name is "Electrical Fault Prediction Method and Device Based on Neuron Model". It discloses an electrical fault prediction method based on neuron model, but this method requires a huge data set to train neurons and is not suitable for the actual situation of well site data collectors.

[0004] The patent number is CN202111407910.8, and its name is "Industrial Equipment Fault Prediction Method". It discloses an industrial equipment fault prediction method that is executed on an edge computing service. However, this method relies on the network bandwidth between the device and the edge server, and has problems such as high cost and long latency.

[0005] The patent number is CN202210701223.5, and the name is "A method for predicting equipment failure based on life consumption". It discloses a method for predicting equipment failure based on life consumption, but this method is not suitable for data collectors with high integration and difficult to predict life consumption.

[0006] In summary, the above fault prediction methods are not suitable for fault prediction of well site data collectors. Summary of the invention

[0007] The purpose of the present invention is to provide a method, system, device and storage medium for predicting faults of a wellsite data collector based on a Bayesian network, which solves the problem that the current fault prediction method is not suitable for predicting faults of a wellsite data collector.

[0008] The present invention is achieved through the following technical solutions:

[0009] The present invention discloses a method for predicting a well site data collector failure based on a Bayesian network, comprising the following steps:

[0010] Obtain the physical quantities actually measured in the well site data collector, and obtain data information after preprocessing;

[0011] Input data information into the constructed fault prediction network model to predict the type and probability of fault occurrence;

[0012] The fault prediction network model is constructed based on a Bayesian network, and the construction process includes the following steps:

[0013] S1. Fault analysis: Select characteristic fault phenomena based on the historical fault information of the well site data collector, combine expert experience and prior knowledge, analyze the relevant fault causes and related influencing factors of the characteristic fault phenomena based on the correlation function analysis method, and obtain the causal relationship between the characteristic fault phenomena and the relevant fault causes and related influencing factors;

[0014] S2. Preliminary construction of Bayesian network: determine the nodes of the Bayesian network based on the relevant influencing factors, relevant fault causes and characteristic fault phenomena, and determine the preliminary structure of the Bayesian network based on the causal relationship between the characteristic fault phenomena obtained in S1 and the relevant fault causes and relevant influencing factors;

[0015] S3. Network optimization: Use the K2 algorithm to optimize the preliminary structure of the Bayesian network, and then use the EM algorithm to optimize the network parameters to obtain the constructed fault prediction network model.

[0016] Further, the preprocessing is: performing abnormality judgment on the actually measured physical quantity according to a preset threshold range, if the state of the physical quantity is normal, determining the state of the physical quantity to be 0; if the state of the physical quantity is abnormal, determining the state of the physical quantity to be 1;

[0017] The preset threshold range is a data range determined based on historical fault information of the well site data collector. If the actually measured physical quantity is not within the preset threshold range, it is determined that the state of the physical quantity is abnormal.

[0018] Further, the specific steps of S1 are as follows:

[0019] (1.1) Selecting a certain fault phenomenon of the well site data collector from the historical fault information as a characteristic fault phenomenon;

[0020] (1.2) Analyze the selected characteristic fault phenomena by combining expert experience, prior knowledge and correlation calculation results to obtain the relevant fault causes and related influencing factors corresponding to the equipment fault phenomena;

[0021] The correlation calculation is obtained by the following formula:

[0022]

[0023] In the formula, C(x 1 ,x 2 ) is the covariance, C(x 1 ,x 2 )=E[(x1 -μ 1 )(x 2 -μ 2 )];

[0024] E is the mathematical expectation, μ i =E[x i ],i=1,2;

[0025] x 1 ,x 2 All represent statistical relationship variables, referring to characteristic fault phenomena and related fault causes, and related fault causes and related influencing factors;

[0026] R represents the correlation function, -1 <R<1;

[0027] When R>0, it means x 1 and x 2 Positive correlation; when R<0, it means x 1 and x 2 Negative correlation; when R = 0, it means x 1 and x 2 Not relevant.

[0028] Furthermore, in S1, the characteristic fault phenomena selected are the collector failure to start up and the collector communication failure, and the related fault causes obtained by analysis are: hard disk failure, motherboard failure, CPU failure and interface damage;

[0029] Relevant influencing factors include: temperature, humidity, dust, voltage and human factors.

[0030] Furthermore, in S3, the specific steps of network optimization are as follows:

[0031] (3.1) Select some data from the historical fault information, pre-process the data, and obtain a data set; wherein the historical fault information includes the fault phenomenon, the related fault cause, and the related influencing factors; each fault phenomenon, the related fault cause, and the related fault influencing factor each corresponds to a node;

[0032] The K2 algorithm is used to optimize the preliminary structure of the Bayesian network in S2 to obtain a preliminary optimized Bayesian network;

[0033] Among them, the K2 algorithm uses the greedy idea to maximize the scoring function. When the scoring function takes the maximum value, the corresponding preliminary optimized Bayesian network structure is used as the optimal network structure.

[0034] (3.2) Select the temperature, humidity and voltage data of the data collector under various scenarios for a period of time, and compare them with the preset thresholds of the three data to confirm the normal and abnormal ratios of each data during this period of time, and obtain the prior probability θ of the root node 0;

[0035] (3.3) Then the EM algorithm is used to optimize the parameters of the preliminary optimized Bayesian network, as follows:

[0036] Select some fault records of the data collector and use the data preprocessing method in (3.1) to obtain the parameter optimization data set D 2 , where all node variables are X=(X 1 ,X 2 …X n ), its joint distribution probability is p(X,z;θ), and the conditional probability distribution is p(z|X,θ), where z is the implicit data in the structure and θ is the model parameter;

[0037] First, initialize the model parameters θ to θ 0 ,θ 0 is the prior probability of the root node;

[0038] Then start the following iteration:

[0039] Step E: Calculate the conditional probability expectation of the joint distribution:

[0040] Q i (z i )=p(z i |X i ,θ j )

[0041]

[0042] j is the number of iterations;

[0043] l is the likelihood function;

[0044] X i is the i-th node variable in the structure;

[0045] z i is the i-th implicit data in the structure;

[0046] θ j is the model parameter at the jth iteration;

[0047] Q is the distribution of implicit data z;

[0048] Step M: Maximize l(θ,θ j ), and we get θ j+1 :

[0049] θ j+1 =argmaxl(θ,θ j )

[0050] When θ j+1 When converged, the model parameters θ are output.

[0051] Furthermore, in (3.1), the scoring function is:

[0052]

[0053] In the formula

[0054] B s Represents the network structure;

[0055] D 1 Represents a dataset;

[0056] q i The number of parent node states representing the number of node values;

[0057] r i Represents a node variable;

[0058] N ijk Representative data set D 1 The occurrence frequency of the corresponding node is satisfied.

[0059] Further, in (3.1), the data preprocessing is: judging the abnormality of the faulty node state according to a preset threshold range, if the node state is normal, determining the node state to be 0; if the node state is abnormal, determining the node state to be 1;

[0060] The preset threshold range is a data range determined based on historical fault information of the well site data collector.

[0061] The present invention discloses a well site data collector fault prediction system based on Bayesian network, comprising:

[0062] A data acquisition module is used to obtain the physical quantities actually measured in the well site data collector;

[0063] The data preprocessing module is used to obtain data information after preprocessing the actual measured physical quantity;

[0064] Fault prediction network model, used to process data information and predict the type and probability of fault occurrence;

[0065] The fault prediction network model is constructed based on the Bayesian network, and the construction process includes the following steps:

[0066] S1. Fault analysis: Select characteristic fault phenomena based on the historical fault information of the well site data collector, combine expert experience and prior knowledge, analyze the relevant fault causes and related influencing factors of the characteristic fault phenomena based on the correlation function analysis method, and obtain the causal relationship between the characteristic fault phenomena and the relevant fault causes and related influencing factors;

[0067] S2. Preliminary construction of Bayesian network: determine the nodes of the Bayesian network based on the relevant influencing factors, relevant fault causes and characteristic fault phenomena, and determine the preliminary structure of the Bayesian network based on the causal relationship between the characteristic fault phenomena obtained in S1 and the relevant fault causes and relevant influencing factors;

[0068] S3. Network optimization: Use the K2 algorithm to optimize the preliminary structure of the Bayesian network, and then use the EM algorithm to optimize the network parameters to obtain the constructed fault prediction network model.

[0069] The present invention discloses a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the Bayesian network-based well site data collector fault prediction method are implemented.

[0070] The present invention discloses a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the Bayesian network-based well site data collector fault prediction method are implemented.

[0071] Compared with the prior art, the present invention has the following beneficial technical effects:

[0072] The present invention proposes a method for predicting the failure of a well site data collector based on a Bayesian network. The Bayesian network structure obtained by learning the K2 algorithm can mine the node causal relationship that is not considered in the process of establishing expert knowledge, and the EM algorithm can be used to obtain the optimal conditional probability table. The method can accurately predict the failure of well site equipment, remedy the equipment before failure, extend the life of the equipment, reduce losses, and has practicality and reliability. The fault prediction method proposed by the present invention can improve the reliability and accuracy of fault prediction, save and reduce the difficulty of maintenance personnel and the cost of maintenance equipment, and at the same time improve the utilization rate of data collectors, and has a good market prospect.

[0073] The present invention also discloses a well site data collector fault prediction system based on a Bayesian network, which is successfully applied to the drilling site. By combining prior knowledge with the Bayesian network for fault prediction, the probability of various faults can be intuitively observed, and the corresponding modules of the equipment can be maintained in advance, thereby reducing the failure rate of the equipment and improving the efficiency of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 is a flow chart of the method of the present invention;

[0075] Figure 2 A Bayesian network structure diagram in a specific embodiment;

[0076] Figure 3 It is a Bayesian network structure diagram after K2 algorithm optimization in a specific embodiment. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solution and advantages of the present invention more clear, the following is further described in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the embodiments described are only part of the embodiments of the present invention, not all embodiments.

[0078] The components described and shown in the drawings and embodiments of the present invention may be arranged and designed in various configurations. Therefore, the detailed description of the embodiments of the present invention provided in the following drawings is not intended to limit the scope of the claimed invention, but merely represents a selected embodiment of the present invention. Based on the drawings and embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0079] It should be noted that the terms "comprises", "includes" or any other variations are intended to cover non-exclusive inclusion, so that a process, element, method, article or equipment that includes a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to the process, element, method, article or equipment.

[0080] The present invention discloses a method for predicting a well site data collector failure based on a Bayesian network, comprising the following steps:

[0081] Obtain the physical quantities actually measured in the well site data collector, and obtain data information after preprocessing;

[0082] The data information is input into the constructed fault prediction network model to predict the type and probability of fault occurrence.

[0083] The fault prediction network model is constructed based on a Bayesian network, and the construction process includes the following steps:

[0084] S1. Fault analysis: Select characteristic fault phenomena based on the historical fault information of the well site data collector, combine expert experience and prior knowledge, analyze the relevant fault causes and related influencing factors of the characteristic fault phenomena based on the correlation function analysis method, and obtain the causal relationship between the characteristic fault phenomena and the relevant fault causes and related influencing factors;

[0085] S2. Preliminary construction of Bayesian network: determine the nodes of the Bayesian network based on the relevant influencing factors, relevant fault causes and characteristic fault phenomena, and determine the preliminary structure of the Bayesian network based on the causal relationship between the characteristic fault phenomena obtained in S1 and the relevant fault causes and relevant influencing factors;

[0086] S3. Network optimization: Use the K2 algorithm to optimize the preliminary structure of the Bayesian network, and then use the EM algorithm to optimize the network parameters to obtain the constructed fault prediction network model.

[0087] The preprocessing is: judging the abnormality of the actually measured physical quantity according to the preset threshold range, if the state of the physical quantity is normal, determining the state of the physical quantity to be 0; if the state of the physical quantity is abnormal, determining the state of the physical quantity to be 1;

[0088] The preset threshold range is a data range determined based on historical fault information of the well site data collector. If the actually measured physical quantity is not within the preset threshold range, it is determined that the state of the physical quantity is abnormal.

[0089] The physical quantities actually measured generally include temperature (T), humidity (W) and voltage (V).

[0090] The features and performances of the present invention are further described in detail below in conjunction with the embodiments.

[0091] The fault prediction network model is constructed based on a Bayesian network, and the construction process includes the following steps:

[0092] (1) Fault analysis of well site data collector:

[0093] (1.1) Based on historical fault information, the two most common and most likely fault phenomena of well site data collectors, namely, "collector cannot be turned on" and "collector communication failure", are selected as analysis targets;

[0094] (1.2) The selected fault phenomenon is analyzed by combining expert experience, prior knowledge, and correlation calculation results to obtain the influencing factors and fault causes corresponding to the equipment fault phenomenon. The correlation calculation is calculated by the following formula:

[0095]

[0096] In the formula, C(x 1 ,x 2 ) is the covariance, C(x 1 ,x 2 )=E[(x 1 -μ 1 )(x 2 -μ 2 )]

[0097] E is the mathematical expectation, μi =E[x i ],i=1,2;

[0098] x 1 ,x 2 Represents statistical relationship variables; statistical relationship variables refer to fault phenomena and fault causes, as well as fault causes and fault influencing factors.

[0099] R represents the correlation function, -1 <R<1

[0100] R>0 means x 1 and x 2 Positive correlation; R<0 means x 1 and x 2 Negative correlation; R = 0 means x 1 and x 2 Not related and have no impact on each other.

[0101] Analysis and calculation show that the related causes of the two fault phenomena of "collector cannot be turned on" and "collector communication failure" are: hard disk failure, motherboard failure, CPU failure, interface damage; related influencing factors are: temperature, humidity, dust, voltage, and human factors.

[0102] (2) Constructing a Bayesian network:

[0103] According to the correlation calculation results of step (1.2), the fault cause and fault influencing factors corresponding to the fault phenomenon are determined, and then the causal relationship between the Bayesian network nodes is determined, thereby determining Figure 2 The Bayesian network structure shown.

[0104] Since the well site data collector is only equipped with thermal resistor sensors and humidity resistor sensors, dust and human factors are difficult to measure, so temperature (T), humidity (W), voltage (V), collector failure to start (A), collector communication abnormality (B), hard disk failure (A1), motherboard failure (A2), CPU failure (A3), and interface damage (B1) are selected as nodes to build a Bayesian network.

[0105] Temperature (T) and voltage (V) will affect hard disk failure (A1), temperature (T), voltage (V) and humidity (W) will affect motherboard failure (A2), temperature (T) will affect CPU failure (A3), and temperature (T) and humidity (W) will affect interface damage (B1). Hard disk failure (A1), motherboard failure (A2) and CPU failure (A3) cause the collector to fail to start up (A), and interface damage (B1) causes collector communication abnormality (B).

[0106] (3) Network optimization:

[0107] (3.1) Select some data from the historical equipment fault information and preprocess it to obtain the data set D 1 .

[0108] The historical fault information of the equipment includes the fault phenomenon, the relevant fault cause corresponding to the fault phenomenon, and the relevant influencing factors corresponding to the relevant fault cause.

[0109] The pre-processing method is: for faulty nodes such as the collector cannot be turned on, the collector communication is abnormal, and the hard disk fails, the node status is "1" when the fault occurs, and the node status is "0" when the fault does not occur. For temperature, humidity, and voltage, according to the data, the normal temperature of the well site data collector is determined to be 15℃~30℃, the humidity is 30%~80%, and the voltage is 12V. When it does not exceed this range, the node status is determined to be "0" which is normal. When it exceeds this range, the node status is determined to be "1" which is abnormal. Only by converting text information into data information can this information be used to optimize the network.

[0110] Combined with the preprocessed data, the K2 algorithm is used to optimize the Bayesian network in step (2), and the following is obtained: Figure 3 The network structure shown.

[0111] The K2 algorithm mainly uses the greedy idea to maximize the following scoring function (2):

[0112]

[0113] In the formula

[0114]

[0115] B s - Network structure;

[0116] D 1 - Datasets;

[0117] q i ——The number of parent node states of the number of node values;

[0118] r i ——Node variables;

[0119] N ijk ——Dataset D 1 The occurrence frequency of the corresponding node is satisfied.

[0120] Because each fault phenomenon, fault cause, and fault influencing factor corresponds to a node, N ijk It is the number of times a node appears in the data set.

[0121] The scoring of the Bayesian network structure is obtained through formula (2). The purpose is to select a structure with the highest score as the optimal Bayesian network structure through calculation of each node.

[0122] (3.2) Select the temperature, humidity and voltage data of the data collector running for a period of time in various scenarios, and compare them with the preset thresholds of the three data to confirm the normal and abnormal ratios of each data during this period of time, and obtain the prior probability θ of the root node 0 ;

[0123] (3.3) Then use the EM algorithm to optimize the parameters of the optimal Bayesian network obtained in (3.1):

[0124] The EM algorithm is also called the expectation maximization algorithm. Each iteration of the algorithm is divided into an expectation step (E step) and a maximum step (M step).

[0125] The details are as follows:

[0126] Select a part of the fault records of the data collector and use the data preprocessing method in (3.1) to obtain the parameter optimization data set D 2 , where all node variables are X=(X 1 ,X 2 …X n ), its joint distribution probability is p(X,z;θ), and the conditional probability distribution is p(z|X,θ), where z is the implicit data in the structure and θ is the model parameter.

[0127] First, initialize the model parameters θ to θ 0 ,θ 0 That is, the prior probability of the root node.

[0128] Step E: Calculate the conditional probability expectation of the joint distribution:

[0129] Q i (z i )=p(z i |X i ,θ j ) (3)

[0130]

[0131] j is the number of iterations;

[0132] l is the likelihood function;

[0133] X i is the i-th node variable in the structure;

[0134] z i is the i-th implicit data in the structure;

[0135] θ j is the model parameter at the jth iteration;

[0136] Step M: Maximize l(θ,θ j ), and we get θ j+1 :

[0137] θ j+1 =argmaxl(θ,θ j ) (5)

[0138] When θ j+1 When converged, the model parameters θ are output.

[0139] The present invention also discloses a well site data collector fault prediction system based on Bayesian network, comprising:

[0140] A data acquisition module is used to obtain the physical quantities actually measured in the well site data collector;

[0141] The data preprocessing module is used to obtain data information after preprocessing the actual measured physical quantity;

[0142] The fault prediction network model is used to process data information and predict the type and probability of fault occurrence.

[0143] The well site data collector fault prediction system developed by the present invention has been successfully applied to the drilling site. By combining prior knowledge with the Bayesian network for fault prediction, the probability of various faults can be intuitively observed, and the corresponding modules of the equipment can be maintained in advance, which reduces the failure rate of the equipment and improves the efficiency of the equipment.

[0144] The actual measured temperature, humidity and voltage values ​​of the well site data collector are obtained, and after preprocessing, the data information is obtained; the data information is input into the constructed fault prediction network model, and the type and probability of occurrence of the fault are output, which directly displays the various probabilities of the collector failing to start up, CPU failure, motherboard failure, hard disk failure, interface damage, and collector communication failure. Maintenance personnel need to check in advance.

[0145] The fault prediction method proposed in the present invention combines prior knowledge with data analysis, has reliability and adaptability, can save and reduce the work difficulty of maintenance personnel and the cost of maintenance equipment, and at the same time improves the utilization rate of data collectors, and has good market prospects.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting failure of a well site data collector based on a Bayesian network, characterized in that: The process includes: Obtain the physical quantities actually measured in the well site data collector, and obtain data information after preprocessing; Input data information into the constructed fault prediction network model to predict the type and probability of fault occurrence; The fault prediction network model is constructed based on a Bayesian network, and the construction process includes the following steps: S1. Fault analysis: Select characteristic fault phenomena based on the historical fault information of the well site data collector, combine expert experience and prior knowledge, analyze the relevant fault causes and related influencing factors of the characteristic fault phenomena based on the correlation function analysis method, and obtain the causal relationship between the characteristic fault phenomena and the relevant fault causes and related influencing factors; S2. Preliminary construction of Bayesian network: determine the nodes of the Bayesian network based on the relevant influencing factors, relevant fault causes and characteristic fault phenomena, and determine the preliminary structure of the Bayesian network based on the causal relationship between the characteristic fault phenomena obtained in S1 and the relevant fault causes and relevant influencing factors; S3. Network optimization: Use the K2 algorithm to optimize the preliminary structure of the Bayesian network, and then use the EM algorithm to optimize the network parameters to obtain the constructed fault prediction network model.

2. A Bayesian network-based wellsite data collector fault prediction method according to claim 1, characterized in that: The preprocessing is: judging the abnormality of the actually measured physical quantity according to the preset threshold range, if the state of the physical quantity is normal, determining the state of the physical quantity to be 0; if the state of the physical quantity is abnormal, determining the state of the physical quantity to be 1; The preset threshold range is a data range determined based on historical fault information of the well site data collector. If the actually measured physical quantity is not within the preset threshold range, it is determined that the state of the physical quantity is abnormal.

3. The method for predicting failure of a well site data collector based on a Bayesian network according to claim 1, characterized in that: The specific steps of S1 are as follows: (1.1) Selecting a certain fault phenomenon of the well site data collector from the historical fault information as a characteristic fault phenomenon; (1.2) Analyze the selected characteristic fault phenomena by combining expert experience, prior knowledge and correlation calculation results to obtain the relevant fault causes and related influencing factors corresponding to the equipment fault phenomena; The correlation calculation is obtained by the following formula: Where C(x1,x2) is the covariance, C(x1,x2)=E[(x1-μ1)(x2-μ2)]; E is the mathematical expectation, μ i =E[x i ],i=1,2; x1 and x2 both represent statistical relationship variables, which refer to characteristic fault phenomena and related fault causes, and related fault causes and related influencing factors; R represents the correlation function, -1 <R<1; When R>0, it means that x1 and x2 are positively correlated; when R<0, it means that x1 and x2 are negatively correlated; when R=0, it means that x1 and x2 are uncorrelated.

4. The method for predicting failure of a well site data collector based on a Bayesian network according to claim 1, characterized in that: In S1, the characteristic fault phenomena selected are the collector failure to start up and the collector communication failure. The related fault causes analyzed are: hard disk failure, motherboard failure, CPU failure and interface damage; Relevant influencing factors include: temperature, humidity, dust, voltage and human factors.

5. The method for predicting failure of a well site data collector based on a Bayesian network according to claim 1, characterized in that: In S3, the specific steps of network optimization are as follows: (3.1) Select some data from the historical fault information, pre-process the data, and obtain a data set; wherein the historical fault information includes the fault phenomenon, the related fault cause, and the related influencing factors; each fault phenomenon, the related fault cause, and the related fault influencing factor each corresponds to a node; The K2 algorithm is used to optimize the preliminary structure of the Bayesian network in S2 to obtain a preliminary optimized Bayesian network; Among them, the K2 algorithm uses the greedy idea to maximize the scoring function. When the scoring function takes the maximum value, the corresponding preliminary optimized Bayesian network structure is used as the optimal network structure. (3.2) Select the temperature, humidity and voltage data of the data collector under various scenarios for a period of time, and compare them with the preset thresholds of the three data to confirm the normal and abnormal ratios of each data during this period of time, and obtain the prior probability θ0 of the root node; (3.3) Then the EM algorithm is used to optimize the parameters of the preliminary optimized Bayesian network, as follows: Select some fault records of the data collector and use the data preprocessing method in (3.1) to obtain the parameter optimization data set D2, where all node variables are X = (X1, X2…X n ), its joint distribution probability is p(X,z;θ), and the conditional probability distribution is p(z|X,θ), where z is the implicit data in the structure and θ is the model parameter; First, initialize the model parameter θ to θ0, where θ0 is the prior probability of the root node; Then start the following iteration: Step E: Calculate the conditional probability expectation of the joint distribution: Q i (z i )=p(z i |X i ,θ j ) j is the number of iterations; l is the likelihood function; X i is the i-th node variable in the structure; z i is the i-th implicit data in the structure; θ j is the model parameter at the jth iteration; Q is the distribution of implicit data z; Step M: Maximize l(θ,θ j ), and we get θ j+1 : i j+1 =argmaxl(θ,θ j ) When θ j+1 When converged, the model parameters θ are output.

6. A Bayesian network-based well site data collector fault prediction method according to claim 5, characterized in that: In (3.1), the scoring function is: In the formula B s Represents the network structure; D1 represents the data set; q i The number of parent node states representing the number of node values; r i Represents a node variable; N ijk Represents the frequency of occurrence of the corresponding node in the data set D1.

7. The method for predicting failure of a well site data collector based on a Bayesian network according to claim 5, characterized in that: In (3.1), the data preprocessing is: judging the abnormality of the faulty node state according to a preset threshold range, and if the node state is normal, determining the node state to be 0; If the node status is abnormal, the node status is determined to be 1; The preset threshold range is a data range determined based on historical fault information of the well site data collector.

8. A well site data collector fault prediction system based on Bayesian network, characterized in that: include: A data acquisition module is used to obtain the physical quantities actually measured in the well site data collector; The data preprocessing module is used to obtain data information after preprocessing the actual measured physical quantity; Fault prediction network model, used to process data information and predict the type and probability of fault occurrence; The fault prediction network model is constructed based on the Bayesian network, and the construction process includes the following steps: S1. Fault analysis: Select characteristic fault phenomena based on the historical fault information of the well site data collector, combine expert experience and prior knowledge, analyze the relevant fault causes and related influencing factors of the characteristic fault phenomena based on the correlation function analysis method, and obtain the causal relationship between the characteristic fault phenomena and the relevant fault causes and related influencing factors; S2. Preliminary construction of Bayesian network: determine the nodes of the Bayesian network based on the relevant influencing factors, relevant fault causes and characteristic fault phenomena, and determine the preliminary structure of the Bayesian network based on the causal relationship between the characteristic fault phenomena obtained in S1 and the relevant fault causes and relevant influencing factors; S3. Network optimization: Use the K2 algorithm to optimize the preliminary structure of the Bayesian network, and then use the EM algorithm to optimize the network parameters to obtain the constructed fault prediction network model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the Bayesian network-based wellsite data collector fault prediction method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the Bayesian network-based wellsite data collector fault prediction method as described in any one of claims 1 to 7 are implemented.

Images