CN114297909A - Water pump fault diagnosis method and system based on neural network - Google Patents
Water pump fault diagnosis method and system based on neural network Download PDFInfo
- Publication number
- CN114297909A CN114297909A CN202111416120.6A CN202111416120A CN114297909A CN 114297909 A CN114297909 A CN 114297909A CN 202111416120 A CN202111416120 A CN 202111416120A CN 114297909 A CN114297909 A CN 114297909A
- Authority
- CN
- China
- Prior art keywords
- water pump
- data
- neural network
- fault
- state
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention provides a water pump fault diagnosis method and system based on a neural network, which comprises the following steps: a data acquisition step: collecting water pump operation data; identifying a model: inputting water pump operation data into a water pump fault identification model to obtain a first diagnosis result output by the water pump fault identification model; a fault judgment step: and judging whether the water pump has a fault or not according to the first diagnosis result of the water pump fault recognition model. The invention can monitor the running state of the water pump in real time and judge and prompt the water pump which is likely to have faults, thereby greatly improving the safety of the water pump in the using process and improving the problems of difficult troubleshooting of the water pump and the like.
Description
Technical Field
The invention relates to the technical field of water pump running state monitoring and diagnosis, in particular to a water pump fault diagnosis method and system based on a neural network.
Background
Water pumps have fairly wide application in industrial systems. In a typical heat exchange system in the energy industry, a water pump is always in an automatic and continuous operation state, and the safety and reliability of the water pump have important significance for the stable operation of the system.
The current water pump running state monitoring and fault diagnosis mainly depends on instrument measurement and subjective experience judgment, but the water pump has numerous running parameters, complex mechanism, strong time-varying property and many coupling factors, so that complex equipment such as the water pump is difficult to describe by an accurate mathematical model.
In addition, the water pump has high safety and reliability, and fault conditions rarely occur in the life cycle, so that a fault diagnosis system of the water pump lacks sufficient fault data which can be used for analysis.
In view of the above-mentioned related technologies, the inventor considers that complex equipment such as a water pump is difficult to describe by an accurate mathematical model, and a fault diagnosis system of the water pump lacks sufficient fault data for analysis, and troubleshooting of the water pump is difficult.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a water pump fault diagnosis method and system based on a neural network.
The invention provides a water pump fault diagnosis method based on a neural network, which comprises the following steps:
a data acquisition step: collecting water pump operation data;
identifying a model: inputting water pump operation data into a water pump fault identification model to obtain a first diagnosis result output by the water pump fault identification model;
a fault judgment step: and judging whether the water pump has a fault or not according to the first diagnosis result of the water pump fault recognition model.
In one possible embodiment, in the data acquisition step, pressure sensors arranged at the upstream and downstream of the water pump pipeline are used for acquiring pipeline internal pressure data at different moments, and the head H (t) of the water pump is calculated according to the pressure data;
collecting flow data q (t) by a flow meter deployed on the water pump;
acquiring power data P (t) through an electricity meter arranged on the water pump;
acquiring frequency data f (t) through a frequency converter arranged on a water pump;
where t represents the acquisition time.
In a possible embodiment, the method further comprises a state model step;
the state model step: inputting water pump operation data into a water pump operation state model to obtain a prediction result of the water pump operation state model; determining a second diagnosis result according to the prediction result, wherein the second diagnosis result is used for indicating whether the running state of the water pump is abnormal or not;
the fault judging step comprises: and judging whether the operation of the water pump fails or not according to the first diagnosis result and the second diagnosis result.
In a possible implementation manner, in the state model step, water pump operation data are input into a water pump operation state model, a water pump frequency predicted value and a water pump power predicted value in a set time period are respectively obtained, and a variance decision coefficient and an average error percentage in the time period are calculated according to the water pump frequency predicted value and the water pump power predicted value and are used as a second diagnosis result; and if the variance decision coefficient and the average error percentage are lower than respective threshold values, determining that the water pump is abnormal.
In a possible implementation manner, in the fault judging step, if the water pump is determined to be abnormal according to both the first diagnosis result and the second diagnosis result, the water pump is considered to be in fault; otherwise, the water pump is considered to be in a normal operation state.
In one possible embodiment, the establishing of the water pump operation state model comprises the following steps:
a sample acquisition step: acquiring normal operation data of the water pump to obtain a normal operation sample set of the water pump; the water pump normal operation sample set comprises basic input parameters and basic output parameters corresponding to the basic input parameters; the basic input parameters comprise the pump lift and the pump flow of the water pump in a concentration mode when the water pump normally operates, and the basic output parameters comprise the pump power and the pump frequency in a concentration mode when the water pump normally operates;
training a state model: establishing a BP neural network, wherein basic input parameters are used as input data of the BP neural network, basic output parameters are used as marking data of corresponding basic input parameters, and output data of the BP neural network are prediction output parameters; and adjusting the network parameters of the BP neural network based on the predicted output parameters and the basic output parameters corresponding to the predicted output parameters, and determining the trained BP neural network as a water pump running state model.
In one possible embodiment, the method further comprises:
for at least one basic input parameter, performing feature expansion on the basic input parameter by using a polynomial combination to obtain a parameter after feature expansion of the basic input parameter, and taking the basic input parameter and the parameter after feature expansion as input data of the BP neural network; comparing the correlation between the basic output parameters and the input data of the BP neural network; and keeping the input data of the BP neural network with preset correlation between the basic output parameters and the input data of the BP neural network.
In one possible implementation, the establishing of the water pump fault identification model comprises the following steps:
a data set construction step: constructing a data set, wherein the data set comprises a normal sample and a fault sample, the normal sample is data when the water pump normally operates, and the fault sample is data when the water pump fails;
identification model training: establishing a convolutional neural network, wherein a data set is used as input data of the convolutional neural network, a state label is used as marking data of the input data, and the state label comprises a first state label for marking normal operation of the water pump and a second state label for marking fault of the water pump; the output data of the convolutional neural network is a prediction state, and the prediction state comprises a normal state for identifying the normal operation of the water pump and a fault state for identifying the fault of the water pump; and optimizing the convolutional neural network based on the prediction state and the state label, and determining the trained convolutional neural network as a water pump fault recognition model.
In one possible implementation, in the data set construction step, water pump operation data is acquired; dividing the water pump operation data at set time intervals, and organizing the water pump operation data into the following forms:
wherein, tkl+lRepresents the kl + l time;represents tkl+lLift at that moment;represents tkl+lThe flow rate at a moment;represents tkl+lThe water pump power at a moment;represents tkl+lThe water pump frequency at that moment; k represents the kth time; l represents the length of the time interval, and n represents a sample of the water pump operating data parameter;
and constructing a data set, and inputting data into a water pump fault identification model to be a time sequence matrix of a time period set in the data set.
In one possible embodiment, the convolutional neural network structure comprises: at least one convolution layer, a flattening layer, a forgetting layer and a full connection layer.
The water pump fault diagnosis system based on the neural network comprises data acquisition hardware, an upper computer and a computer, wherein the data acquisition hardware is connected with the upper computer;
the data acquisition hardware acquires water pump operation data and sends the water pump operation data to the upper computer;
the upper computer receives and stores the water pump operation data and sends the water pump operation data to the computer;
and the computer processes the water pump operation data.
Compared with the prior art, the invention has the following beneficial effects:
1. the water pump has quite wide application in an industrial system, and the safety and the reliability of the water pump have quite important significance for the stable operation of the system; the invention can monitor the running state of the water pump in real time and judge and prompt the water pump which is likely to have faults, thereby greatly improving the safety of the water pump in the using process and improving the problems of difficult troubleshooting of the water pump and the like;
2. according to the invention, the wireless sensor is used for acquiring lift data, the wireless sensor is convenient to deploy, wiring is not required, the connection is fast and stable, and a WiFi access control system is used;
3. the invention integrates the prediction results of the two neural network models, improves the accuracy of fault diagnosis and reduces the possibility of program misjudgment.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is a schematic diagram of a water pump fault diagnosis of the present invention;
FIG. 2 is a schematic diagram of data collected by the water pump of the present invention;
FIG. 3 is a schematic diagram of the basic process of the present invention;
FIG. 4 is a schematic structural diagram of a BP neural network according to the present invention;
FIG. 5 is a schematic diagram of the convolutional neural network of the present invention.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
The embodiment of the invention discloses a water pump fault diagnosis method based on a neural network, which comprises the following steps of: a data acquisition step: and collecting water pump operation data. Identifying a model: and inputting the water pump operation data into the water pump fault recognition model to obtain a first diagnosis result output by the water pump fault recognition model. A fault judgment step: and judging whether the water pump has a fault or not according to the first diagnosis result of the water pump fault recognition model.
As shown in fig. 2 and 3, in the data acquisition step, water pump operation data is acquired, and the water pump operation data includes: the head H (t) of the water pump, the flow Q (t) of the water pump, the power P (t) of the water pump and the frequency f (t) of the water pump.
In some embodiments, pressure sensors disposed upstream and downstream of the water pump pipeline are used for acquiring pipeline internal pressure data at different times, and the head h (t) of the water pump is calculated according to the pressure data, wherein the head is 1 m, namely the pressure is 10 kPa.
In some embodiments, the flow data q (t) is collected by a flow meter deployed on the water pump.
In some embodiments, power data p (t) is collected by an electricity meter deployed on the water pump; frequency data f (t) is collected by a transducer deployed on the water pump.
Where t represents the acquisition time.
In the specific implementation process, when the water pump data is acquired, the measuring instruments (data acquisition hardware) comprise a wireless pressure sensor, a flowmeter, an electric quantity meter and a frequency converter. Acquiring the operation data of the water pump in real time: the measuring instruments deployed on the pipeline and the equipment are connected to an upper computer, and the upper computer collects reading data of the measuring instruments changing along with time. The specific measuring instrument is arranged and collects the following physical quantities: pressure data are collected by wireless pressure sensors arranged in front of and behind the water pump pipeline, the wireless pressure sensors are wirelessly connected to an upper computer through WIFI, the lift H (t) is obtained through calculation, and the wireless pressure sensors have the advantages of being free of wiring, flexible in arrangement and stable in connection. The wireless pressure sensors may also be connected by wired means, such as using a network cable or 485 bus, or they may be connected by other wireless means (NBIOT or LORA). And a wired flowmeter arranged on the water pump collects flow data Q (t) and transmits the flow data Q (t) to an upper computer in a wired mode. The flowmeter is installed in water pump upper reaches low reaches all can, but the mounted position should satisfy that the distance of upper reaches distance water pump is no less than 5 times the straight tube end or the distance of low reaches distance water pump is no less than 3 times the straight tube section, and can not have other pipe fittings on this straight tube section. The flow meter and the upper computer can use wireless transmission or wired transmission, and the method is not limited in particular. And acquiring power data P (t) by an electricity meter arranged on the water pump and transmitting the power data P (t) to an upper computer in a wired mode. The electricity meter is installed on the distribution switch cabinet of water pump. The electric quantity meter and the upper computer can be wireless or wired.
Frequency data f (t) are collected by a frequency converter arranged on the water pump and are transmitted to an upper computer in a wired mode. The flow meter acquires flow data Q (t), the watt-hour meter acquires power data P (t), and the frequency converter acquires frequency data f (t), which are all transmitted to the upper computer in a wired mode. The converter is installed between water pump electrode and switch board, and the electricity input of switch board is to the converter usually, and the output current input of converter is to the water pump motor. The spatial position of the frequency converter has no specific requirements. The frequency converter and the upper computer are usually in wired transmission, and wireless transmission can also be used. The upper computer is responsible for collecting training data required by modeling.
In the specific implementation process, recent data (generally data from the operation time to the previous hour) is input into the model, and the operation state of the water pump is monitored in real time and subjected to fault diagnosis. And (3) analyzing the monitored recent data of the latest period of time in real time by using a data acquisition system established in the data acquisition step. The recent water pump operation data comprises the actual lift of the water pump, the actual flow of the water pump, the actual power of the water pump and the actual frequency of the water pump.
In the step of identifying the model, the operation data of the water pump is input into the water pump fault identification model, and a first diagnosis result output by the water pump fault identification model is obtained. Inputting the recent data (the actual lift of the water pump, the actual flow of the water pump, the actual power of the water pump and the actual frequency of the water pump) into the established convolutional neural network water pump fault identification model, and outputting a first diagnosis result output by the normal or abnormal water pump fault identification model by the model. The output of the convolutional neural network is a normal or abnormal label.
The method further comprises the step of state modeling: inputting water pump operation data into a water pump operation state model to obtain a prediction result of the water pump operation state model; and determining a second diagnosis result according to the prediction result, wherein the second diagnosis result is used for indicating whether the water pump running state is abnormal or not.
Specifically, the water pump operation data is input into the water pump operation state model, a water pump frequency predicted value and a water pump power predicted value in a set time period (the set time period and recent data refer to the same time period and are data from the program operation time to the previous hour) are respectively obtained, and a variance decision coefficient and an average error percentage in the time period are calculated according to the water pump frequency predicted value and the water pump power predicted value and are used as a second diagnosis result. If the variance decision coefficient and the average error percentage are lower than respective threshold values, the water pump is determined to be abnormal, otherwise, the water pump is determined to be normal.
In some embodiments, the recent data (the actual pump lift and the actual flow of the water pump) are input into the established BP neural network water pump running state model to obtain a water pump frequency predicted value and a water pump power predicted value in the time period, the water pump frequency predicted value is compared with the actual water pump frequency, the water pump power predicted value is compared with the actual water pump power, and a variance decision coefficient and an average error percentage in the time period are calculated.
One possible implementation of determining the variance-determining coefficient:
wherein R represents a variance determining coefficient; SSR, which is generally called Sum of Squares Total, represents the variable y relative to the centerThe abnormal movement of (2); SST English is called Sum of Squares Regression, and represents a variableRelative to the centerThe abnormal movement of (2);representing a predicted value of the ith sample; i refers to the number of samples in the evaluation period, usually one sample per minute, so one hour is usually 60 samples, and i usually ranges from 1 hour, i is 60.An average value representing the entire sample actual value; y isiRepresents the actual value of the ith sample, whereinAnd yiThe water pump frequency can be corresponding to the water pump power and the water pump frequency. When in useWhen the power of the water pump is predicted,and yiThe average value of the actual power of the water pump and the actual power value of the water pump are respectively. When in useWhen the frequency of the water pump is predicted,and yiThe average value of the actual frequency of the water pump and the actual frequency value of the water pump are respectively. The general evaluation method is described here, the water pump power needs to be evaluated using the formula of R2, and the water pump frequency needs to be evaluated similarly.
One possible implementation of determining the average error percentage:
wherein MAPE represents the mean absolute percentage error; actual (t) represents the actual value of the t sample; forecast (t) indicates the predicted value of the t-th sample; n represents the number of samples, wherein actual (t) and forecast (t) can represent both the water pump power and the water pump frequency. The average error percentage is not equal to the average absolute error percentage, which includes the average absolute percentage error, with the difference being whether or not absolute value calculations are involved. When actual (t) represents the actual power value of the water pump, forecast (t) represents the predicted power value of the water pump. While actual (t) represents the actual frequency value of the water pump, forecast (t) represents the predicted value of the water pump frequency. The general calculation method is introduced here, the water pump power needs to be calculated by using the formula of MAPE, and the water pump frequency also needs to be calculated in the same way.
And if the variance decision coefficient and the average error percentage are both lower than the threshold value, determining that the running state of the water pump is abnormal. The deep neural network is responsible for comparing the predicted frequency and power (the predicted value of the water pump frequency and the predicted value of the water pump power) with the actually measured frequency and power (the predicted value of the actual water pump frequency and the predicted value of the actual water pump power), and if the decision coefficient is lower than 0.85 and the average error percentage is higher than 10%, the fault is diagnosed. And.
In the invention, the sequence of the determination process of the first diagnosis result and the determination process of the second diagnosis result is not limited, the first diagnosis result can be determined first, and then the second diagnosis result can be determined; or the second diagnosis result can be determined firstly, and then the first diagnosis result is determined; determining the first diagnostic result and determining the second diagnostic result may also be performed simultaneously.
And in the fault judging step, judging according to the first diagnosis result and the second diagnosis result to determine whether the water pump is in fault during operation. Integrating the judgment results of the two models, and if the water pump is determined to be abnormal according to the first diagnosis result and the second diagnosis result, determining that the water pump has a fault; otherwise, the water pump is considered to be in a normal operation state. And (4) integrating the results of the two models, if the results are abnormal, judging that the water pump has a fault currently, and if not, judging that the water pump has no fault currently.
And carrying out BP neural network and convolutional neural network modeling.
As shown in fig. 3 and 4, the present invention provides an embodiment of establishing a model of an operation state of a water pump. The establishment of the water pump running state model comprises the following steps: (1) a sample acquisition step: and acquiring normal operation data of the water pump to obtain a normal operation sample set of the water pump. The water pump normal operation sample set comprises basic input parameters and basic output parameters corresponding to the basic input parameters. The basic input parameters comprise water pump lift (recorded as H) and water pump flow (recorded as Q) of a water pump normal operation sample set, and the basic output parameters comprise water pump power (recorded as P) and water pump frequency (recorded as f) of the water pump normal operation sample set. Obtaining n samples of the normal operation data parameters of the water pump at different moments, wherein n is a positive integer.
In some embodiments, the water pump operation state model employs a BP neural network, where BP full-text Back Propagation, chinese translation is a Back Propagation algorithm.
In some embodiments, when the input parameters and the output parameters of the water pump operation state model are obtained, the data acquisition system established in the data acquisition step obtains the normal operation data of the water pump within a set time length, and organizes the data into the following form:
wherein, tiRepresents the ith time;represents tiLift at that moment;represents tiThe flow rate at a moment;represents tiThe water pump power at a moment;represents tiTime of dayThe water pump frequency of (1). And setting the duration to be under a normal condition, carrying out value taking and calculation once per hour, wherein each value taking can acquire data from the program running time to the previous hour, and 60 samples in total.
By the method, n one-dimensional samples of the normal operation parameters of the water pump at different moments are obtained. The data acquisition step is similar to the data acquisition step of the fault diagnosis, and is not described herein again.
In some embodiments, to improve the fitting accuracy of the model, the automatic expansion of the independent variables is performed on the basic input parameters, which specifically includes: for at least one basic input parameter, feature expansion can be performed on the basic input parameter by using polynomial combination to obtain a parameter after feature expansion of the basic input parameter, and the basic input parameter and the parameter after feature expansion are both used as input data of the BP neural network.
Specifically, feature expansion is carried out on basic input parameters by using a method of logarithmic transformation and polynomial combination, and the correlation between input data (the basic input parameters and the parameters after feature expansion) of the BP neural network and basic output parameters is compared; the input data of the BP neural network that is a predetermined correlation between the basic output parameter and the input data of the BP neural network is retained (left). Only the variable features with the predetermined correlation larger than 0.5 or smaller than-0.5 among the basic input parameters and the extended parameters thereof are retained, and the others are deleted. The logarithmization is that the raw data (basic input parameters) are all taken to their natural logarithms.
And comparing the correlation between the input data (basic input parameters and expanded input parameters) and the output parameters of the BP neural network, and reserving the expanded variables with higher predetermined correlation. The correlation coefficient of each input and output is respectively evaluated by a Pearson correlation coefficient between input data and output parameters of the BP neural network, and input characteristics with the correlation coefficient more than 0.5 or less than-0.5 are reserved. The predetermined correlation is high meaning that the correlation is greater than 0.5 or less than-0.5. The output parameters are: p: power of the water pump; f: the water pump frequency.
(2) Training a state model: establishing a BP neural network, wherein basic input parameters are used as input data of the BP neural network, basic output parameters are used as marking data of corresponding basic input parameters, and output data of the BP neural network are prediction output parameters; and adjusting the network parameters of the BP neural network based on the predicted output parameters and the basic output parameters corresponding to the predicted output parameters, and determining the trained BP neural network as a water pump running state model.
Specifically, the BP neural network includes the following parameters: the number of nodes of the input layer is 6, the number of nodes of the output layer is 1, the number of nodes of the hidden layer is 10, the activation function is linear activation, and the parameters are used for describing the structure of the BP neural network model. In the model training, based on the water pump normal operation sample set obtained in the sample obtaining step, a network parameter which enables the prediction error of the output parameter to be minimum is obtained through calculation by using a back propagation algorithm, and a water pump operation state model is established according to the network parameter.
As shown in fig. 3 and 5, an embodiment of the present invention provides a method for establishing a water pump fault identification model. The method specifically comprises the following steps:
(1) a data set construction step: and constructing a data set, wherein the data set comprises a normal sample and a fault sample, the normal sample is data when the water pump normally operates, and the fault sample is data when the water pump fails.
Specifically, acquiring water pump operation data; dividing water pump operation data at set time intervals, further dividing the data at set time intervals on the basis of a one-dimensional sample form of water pump normal operation parameters at different moments, and organizing the data into a time sequence matrix in the following form:
wherein, tkl+lRepresents the kl + l time;represents tkl+lLift at that moment;represents tkl+lThe flow rate at a moment;represents tkl+lThe water pump power at a moment;represents tkl+lThe water pump frequency at that moment; k represents the kth time; l represents the length of the time interval, and the value of l is usually 20 minutes, namely 20 samples, but can also be adjusted along with the model training condition; and n represents the number of samples of the water pump operation data parameter.
In some embodiments, 1 to 4 columns (here, vertical selection, 60 samples per hour by time division, and 4 types by dimension (characteristic/physical quantity) division) are randomly selected from H, Q, P, f, random noise of no more than [ -30%, + 30% ] is added to each position of randomly selected data to form new data samples, the failure data is labeled, and the normal data is labeled on the original samples to form a data set with failure samples larger than a certain percentage.
And constructing a data set, wherein input data of the water pump fault identification model is a time sequence matrix of a time period set in the data set. The data when the water pump is in fault comprise data formed by adding random noise which does not exceed a preset range into normal operation data of the water pump and actually collected operation data when the water pump is in fault.
(2) Identification model training: establishing a convolutional neural network, wherein a data set is used as input data of the convolutional neural network, a state label is used as marking data of the input data, and the state label comprises a first state label for marking normal operation of the water pump and a second state label for marking fault of the water pump; the output data of the convolutional neural network is a prediction state, and the prediction state comprises a normal state for identifying the normal operation of the water pump and a fault state for identifying the fault of the water pump; and optimizing the convolutional neural network based on the prediction state and the state label, and determining the trained convolutional neural network as a water pump fault recognition model.
Specifically, the second state labels include a fault label of H with random noise, a fault label of Q with random noise, a fault label of P with random noise, and a fault label of f with random noise. When random noise is added to H, random noise is added to H and Q, random noise is added to H and P, random noise is added to H and f, random noise is added to H, Q and P, random noise is added to H, Q and f, random noise is added to H, P and f, and random noise is added to H, Q, P and f, the random noise corresponds to the fault tag of H with random noise in the second state tag (as long as random noise is added to H, the random noise is defined as the fault tag of H with random noise). When random noise is added to Q, random noise is added to Q and P, random noise is added to Q and f, and any one of the four conditions of adding random noise to Q, P and f corresponds to the fault label of Q with random noise in the second state label. When random noise is added into P, the random noise is added into P and f, and the random noise and the P correspond to the fault label of the P with the random noise in the second state label. And when only random noise is added into the f, a fault label corresponding to the f with random noise in the second state label is added.
In some embodiments, the convolutional neural network structure comprises: at least one convolution layer, a flattening layer, a forgetting layer and a full connection layer. The convolutional layers include a first convolutional layer (16 convolutional kernels) and a second convolutional layer (32 convolutional kernels). The full connection layer comprises a first full connection layer and a second full connection layer, and the second full connection layer is a Softmax output layer. The Softmax chinese translation is a reversible probability distribution function. The first convolution layer, the second convolution layer, the flattening layer, the forgetting layer, the first full-connection layer and the Softmax output layer are connected in sequence. The fully-connected layer is the concept of a layer inside a neural network model, each fully-connected layer also has its own independent input and output, and the input of the softmax layer is used as the input of the final model, so in the large classification of the model layer, softmax is the output layer.
The convolution layer plays a role in filtering and can extract the characteristics of the image or data; the flattening layer is used for realizing one-dimensional data and has the function of connecting a convolution layer with a forgetting layer; the forgetting layer discards specific connection to prevent overfitting; the fully connected layer functions as a classifier; the softmax layer makes the classification result probabilistic.
In some embodiments, determining the convolutional neural network structure comprises: the multilayer packaging structure comprises a first coiling layer, a second coiling layer, a flattening layer, a forgetting layer (the forgetting rate is 0.4), a first full connecting layer, a second full connecting layer, a softmax output layer and no pooling layer. The purpose of pooling is to reduce dimensionality and model complexity, and the models used herein are not so complex that pooling is not required. Each of the aforementioned layers needs to use an activation function, which uses the RELU activation function:
y=max(x,0)。
wherein x represents a model input; y represents the model output. RELU is called a Rectified Linear Unit in english, and chinese translation is a Linear rectification function, also called a modified Linear Unit. The main role of the activation function is to make the neural network model exhibit non-linear characteristics, and the function is of the form y ═ max (0, x). The convolutional neural network structure uses this functional form for each layer.
In the specific implementation process, the input parameters of the water pump fault identification model are H, Q, P, f time series matrix in a certain time period, and the output parameters are normal or fault state labels. The input parameters of the water pump fault identification model are a time sequence matrix of a certain time period in the data set. The time series matrix is a matrix (subset of the time series matrix) formed by randomly selecting a time in the data set and taking consecutive samples from the time to the time + N, where N is typically 20 minutes. The moment is generated in the model training process and is a random moment. And performing parameter optimization solution on the formed matrix based on the data set constructed in the data set construction step, wherein the solution method uses a gradient descent method, the solution target is that the loss function of the predicted value and the actual value of the model is minimum, the loss function uses Euclidean distance, namely the Euclidean distance is minimum, and finally, the convolutional neural network water pump fault identification model is established.
The data is collated into one-dimensional training data (one sample per time) and two-dimensional data (consecutive l minutes of data make up one sample). And (4) substituting the one-dimensional data into a deep neural network (BP neural network), and substituting the two-dimensional data into a convolutional neural network for training respectively.
The embodiment of the invention also provides a water pump fault diagnosis system based on the neural network, which comprises data acquisition hardware, an upper computer and a computer, wherein the data acquisition hardware is connected with the upper computer, and the upper computer is connected with the computer. The data acquisition hardware acquires water pump operation data and sends the water pump operation data to the upper computer; the upper computer receives and stores the water pump operation data and sends the water pump operation data to the computer; and the computer processes the water pump operation data. The data acquisition hardware comprises a wireless pressure sensor, a flowmeter, an electric quantity meter and a frequency converter. The wireless pressure sensor is arranged in front of and behind the water pump pipeline and connected with the upper computer. The flow meter is disposed upstream or downstream of the water pump. The flowmeter is connected with the upper computer. The electricity meter is installed on the distribution switch cabinet of water pump. The electric quantity meter is connected with the upper computer. The frequency converter is installed between water pump electrode and switch board, and the frequency converter is connected with the host computer. A BP neural network model and a convolution neural network model are built in the computer. And the upper computer inputs the data collected from the wireless pressure sensor, the flowmeter, the coulometer and the frequency converter into a BP neural network model and a convolution neural network model in the computer for processing.
The system for monitoring the running state of the water pump in real time and diagnosing the fault consists of data acquisition hardware and an artificial intelligent monitoring system. The invention relates to the field of computer body networking technology and artificial intelligence, and provides a water pump running state real-time monitoring and fault diagnosis based on a neural network. In order to verify the accuracy and reliability of the method, corresponding test data is selected and brought into the established neural network model for performance analysis. And classification evaluation indexes such as precision, accuracy, recall rate, ROC curve and AUC are introduced, and the performance of the invention is strictly evaluated and analyzed. Through the analysis of field data, the invention is reliable and effective. The classification evaluation index is used in model training and is a general method for model training.
Those skilled in the art will appreciate that, in addition to implementing the system and its various devices, modules, units provided by the present invention as pure computer readable program code, the system and its various devices, modules, units provided by the present invention can be fully implemented by logically programming method steps in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system and various devices, modules and units thereof provided by the invention can be regarded as a hardware component, and the devices, modules and units included in the system for realizing various functions can also be regarded as structures in the hardware component; means, modules, units for performing the various functions may also be regarded as structures within both software modules and hardware components for performing the method.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (11)
1. A water pump fault diagnosis method based on a neural network is characterized by comprising the following steps:
a data acquisition step: collecting water pump operation data;
identifying a model: inputting water pump operation data into a water pump fault identification model to obtain a first diagnosis result output by the water pump fault identification model;
a fault judgment step: and judging whether the water pump has a fault or not according to the first diagnosis result of the water pump fault recognition model.
2. The water pump fault diagnosis method based on the neural network as claimed in claim 1, wherein in the data acquisition step, pressure sensors disposed upstream and downstream of the water pump pipeline are used to acquire pipeline internal pressure data at different times, and the head h (t) of the water pump is calculated according to the pressure data;
collecting flow data q (t) by a flow meter deployed on the water pump;
acquiring power data P (t) through an electricity meter arranged on the water pump;
acquiring frequency data f (t) through a frequency converter arranged on a water pump;
where t represents the acquisition time.
3. The neural network-based water pump fault diagnosis method according to claim 1, further comprising a state model step;
the state model step: inputting water pump operation data into a water pump operation state model to obtain a prediction result of the water pump operation state model; determining a second diagnosis result according to the prediction result, wherein the second diagnosis result is used for indicating whether the running state of the water pump is abnormal or not;
the fault judging step comprises: and judging whether the operation of the water pump fails or not according to the first diagnosis result and the second diagnosis result.
4. The neural network-based water pump fault diagnosis method according to claim 3, wherein in the state model step, water pump operation data is input into a water pump operation state model, a water pump frequency predicted value and a water pump power predicted value within a set time period are obtained respectively, and a variance decision coefficient and an average error percentage within the time period are calculated according to the water pump frequency predicted value and the water pump power predicted value and are used as a second diagnosis result; and if the variance decision coefficient and the average error percentage are lower than respective threshold values, determining that the water pump is abnormal.
5. The water pump fault diagnosis method based on the neural network as claimed in claim 3, wherein in the fault judgment step, if it is determined that the water pump is abnormal according to both the first diagnosis result and the second diagnosis result, the water pump is considered to be faulty; otherwise, the water pump is considered to be in a normal operation state.
6. The neural network-based water pump fault diagnosis method according to claim 3, wherein the establishment of the water pump operation state model comprises the steps of:
a sample acquisition step: acquiring normal operation data of the water pump to obtain a normal operation sample set of the water pump; the water pump normal operation sample set comprises basic input parameters and basic output parameters corresponding to the basic input parameters; the basic input parameters comprise the pump lift and the pump flow of the water pump in a concentration mode when the water pump normally operates, and the basic output parameters comprise the pump power and the pump frequency in a concentration mode when the water pump normally operates;
training a state model: establishing a BP neural network, wherein basic input parameters are used as input data of the BP neural network, basic output parameters are used as marking data of corresponding basic input parameters, and output data of the BP neural network are prediction output parameters; and adjusting the network parameters of the BP neural network based on the predicted output parameters and the basic output parameters corresponding to the predicted output parameters, and determining the trained BP neural network as a water pump running state model.
7. The neural network-based water pump fault diagnosis method according to claim 6, further comprising:
for at least one basic input parameter, performing feature expansion on the basic input parameter by using a polynomial combination to obtain a parameter after feature expansion of the basic input parameter, and taking the basic input parameter and the parameter after feature expansion as input data of the BP neural network; comparing the correlation between the basic output parameters and the input data of the BP neural network; and keeping the input data of the BP neural network with preset correlation between the basic output parameters and the input data of the BP neural network.
8. The neural network-based water pump fault diagnosis method according to claim 1, wherein the establishment of the water pump fault recognition model comprises the following steps:
a data set construction step: constructing a data set, wherein the data set comprises a normal sample and a fault sample, the normal sample is data when the water pump normally operates, and the fault sample is data when the water pump fails;
identification model training: establishing a convolutional neural network, wherein a data set is used as input data of the convolutional neural network, a state label is used as marking data of the input data, and the state label comprises a first state label for marking normal operation of the water pump and a second state label for marking fault of the water pump; the output data of the convolutional neural network is a prediction state, and the prediction state comprises a normal state for identifying the normal operation of the water pump and a fault state for identifying the fault of the water pump; and optimizing the convolutional neural network based on the prediction state and the state label, and determining the trained convolutional neural network as a water pump fault recognition model.
9. The neural network-based water pump fault diagnosis method according to claim 8, wherein in the data set construction step, water pump operation data is acquired; dividing the water pump operation data at set time intervals, and organizing the water pump operation data into the following forms:
wherein, tkl+lRepresents the kl + l time;represents tkl+lLift at that moment;represents tkl+lThe flow rate at a moment;represents tkl+lThe water pump power at a moment;represents tkl+lThe water pump frequency at that moment; k represents the kth time; l represents the length of the time interval, and n represents a sample of the water pump operating data parameter;
and constructing a data set, and inputting data into a water pump fault identification model to be a time sequence matrix of a time period set in the data set.
10. The neural network-based water pump fault diagnosis method according to claim 8, wherein the convolutional neural network structure comprises: at least one convolution layer, a flattening layer, a forgetting layer and a full connection layer.
11. A water pump fault diagnosis system based on a neural network is characterized by comprising data acquisition hardware, an upper computer and a computer, wherein the data acquisition hardware is connected with the upper computer;
the data acquisition hardware acquires water pump operation data and sends the water pump operation data to the upper computer;
the upper computer receives and stores the water pump operation data and sends the water pump operation data to the computer;
and the computer processes the water pump operation data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111416120.6A CN114297909A (en) | 2021-11-25 | 2021-11-25 | Water pump fault diagnosis method and system based on neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111416120.6A CN114297909A (en) | 2021-11-25 | 2021-11-25 | Water pump fault diagnosis method and system based on neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114297909A true CN114297909A (en) | 2022-04-08 |
Family
ID=80965264
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111416120.6A Pending CN114297909A (en) | 2021-11-25 | 2021-11-25 | Water pump fault diagnosis method and system based on neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114297909A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114743356A (en) * | 2022-04-12 | 2022-07-12 | 西安热工研究院有限公司 | Intelligent monitoring and early warning system for whole operation process of steam-driven water supply pump system |
CN114893390A (en) * | 2022-07-15 | 2022-08-12 | 安徽云磬科技产业发展有限公司 | Pump equipment fault detection method based on attention and integrated learning mechanism |
CN115222164A (en) * | 2022-09-20 | 2022-10-21 | 国能大渡河大数据服务有限公司 | Water pump fault prediction method and system based on empirical coupling function |
CN116634037A (en) * | 2023-07-25 | 2023-08-22 | 水利部机电研究所 | Data transmission method and system for intelligent diagnosis of water pump |
CN116906306A (en) * | 2023-07-05 | 2023-10-20 | 山东亿宁环保科技有限公司 | Vacuum pump integrated control system based on convolutional neural network |
-
2021
- 2021-11-25 CN CN202111416120.6A patent/CN114297909A/en active Pending
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114743356A (en) * | 2022-04-12 | 2022-07-12 | 西安热工研究院有限公司 | Intelligent monitoring and early warning system for whole operation process of steam-driven water supply pump system |
CN114893390A (en) * | 2022-07-15 | 2022-08-12 | 安徽云磬科技产业发展有限公司 | Pump equipment fault detection method based on attention and integrated learning mechanism |
CN114893390B (en) * | 2022-07-15 | 2023-08-04 | 安徽云磬科技产业发展有限公司 | Pump equipment fault detection method based on attention and integrated learning mechanism |
CN115222164A (en) * | 2022-09-20 | 2022-10-21 | 国能大渡河大数据服务有限公司 | Water pump fault prediction method and system based on empirical coupling function |
CN116906306A (en) * | 2023-07-05 | 2023-10-20 | 山东亿宁环保科技有限公司 | Vacuum pump integrated control system based on convolutional neural network |
CN116906306B (en) * | 2023-07-05 | 2024-02-13 | 山东亿宁环保科技有限公司 | Vacuum pump integrated control system based on convolutional neural network |
CN116634037A (en) * | 2023-07-25 | 2023-08-22 | 水利部机电研究所 | Data transmission method and system for intelligent diagnosis of water pump |
CN116634037B (en) * | 2023-07-25 | 2023-09-29 | 水利部机电研究所 | Data transmission method and system for intelligent diagnosis of water pump |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN114297909A (en) | Water pump fault diagnosis method and system based on neural network | |
CN108604360B (en) | Facility abnormity monitoring method and system | |
CN113343633B (en) | Thermal runaway fault classification and risk prediction method and system for power lithium battery | |
CN106528975B (en) | A kind of prognostic and health management method applied to Circuits and Systems | |
CN110738274A (en) | nuclear power device fault diagnosis method based on data driving | |
US20120072029A1 (en) | Intelligent system and method for detecting and diagnosing faults in heating, ventilating and air conditioning (hvac) equipment | |
CN108304661A (en) | Diagnosis prediction method based on TDP models | |
CN109213127A (en) | A kind of HVAC system gradual failure diagnostic method based on deep learning | |
CN103631145B (en) | Multiple operating modes process monitoring method and system based on monitor control index switching | |
WO2011073613A1 (en) | A method, apparatus and computer program for diagnosing a mode of operation of a machine | |
CN110555230A (en) | rotary machine residual life prediction method based on integrated GMDH framework | |
CN110209999A (en) | A kind of mobile unit failure trend prediction method | |
CN109597396B (en) | A kind of distribution transforming on-line fault diagnosis method based on high amount of traffic and transfer learning | |
CN110375983A (en) | Failsafe valve real-time diagnosis system and diagnostic method based on time series analysis | |
Wang et al. | Fault detection and diagnosis for multiple faults of VAV terminals using self-adaptive model and layered random forest | |
CN112598144A (en) | CNN-LSTM burst fault early warning method based on correlation analysis | |
CN113757093A (en) | Fault diagnosis method for flash steam compressor unit | |
KR101505975B1 (en) | Method and system for fault dignosis of engine | |
KR20230125116A (en) | The method for fault detection without training data or diagnosis with prediction of remaining time until breakdown using clustering algorithm and statistical methods | |
CN108508860A (en) | A kind of process industry production system data monitoring method based on coupled relation | |
CN116259161B (en) | Power failure early warning system | |
CN117269340A (en) | GIS equipment intelligent fault diagnosis method and system based on gas components | |
CN114577470A (en) | Fault diagnosis method and system for fan main bearing | |
WO2017081659A1 (en) | Method for the analysis of a time series of measurements of a signal characteristic of a system | |
CN115828744A (en) | White light LED fault on-line diagnosis and service life prediction method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |