CN117556261A - MCNN-based diaphragm pump check valve service life prediction method and system - Google Patents
MCNN-based diaphragm pump check valve service life prediction method and system Download PDFInfo
- Publication number
- CN117556261A CN117556261A CN202410022049.0A CN202410022049A CN117556261A CN 117556261 A CN117556261 A CN 117556261A CN 202410022049 A CN202410022049 A CN 202410022049A CN 117556261 A CN117556261 A CN 117556261A
- Authority
- CN
- China
- Prior art keywords
- data
- check valve
- diaphragm pump
- mcnn
- pump check
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000009826 distribution Methods 0.000 claims abstract description 17
- 230000004927 fusion Effects 0.000 claims abstract description 17
- 238000013528 artificial neural network Methods 0.000 claims abstract description 15
- 230000003042 antagnostic effect Effects 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims description 50
- 238000012360 testing method Methods 0.000 claims description 29
- 238000012545 processing Methods 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 22
- 238000007781 pre-processing Methods 0.000 claims description 14
- 238000011176 pooling Methods 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 6
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 239000012528 membrane Substances 0.000 claims description 3
- 238000005096 rolling process Methods 0.000 claims description 3
- 238000013450 outlier detection Methods 0.000 claims 1
- 238000005070 sampling Methods 0.000 claims 1
- 238000003860 storage Methods 0.000 description 9
- 210000002569 neuron Anatomy 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 239000007788 liquid Substances 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 210000002364 input neuron Anatomy 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000005086 pumping Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Reciprocating Pumps (AREA)
Abstract
The invention provides a MCNN-based diaphragm pump check valve life prediction method and a MCNN-based diaphragm pump check valve life prediction system, which are characterized in that a square Wasserstein distance is introduced to replace a JS and KL divergent antagonistic neural network, data identical to characteristic data distribution is generated through a generator, the data of the generator is optimized through a discriminator, predicted values identical to the original data set are obtained, the predicted values are input into an LSTM model to be predicted, vibration and noise are synchronously transmitted into a life prediction model, and different weights are given to two signal results according to the difference of the square Wasserstein distance of the two signals, so that fusion of the two predicted results is realized. The invention obviously enhances the accuracy of the model by means of fusion of the filling data and the predicted value, can provide more reliable prediction under the condition of data scarcity, and provides feasibility and stability for the prediction model in practical application.
Description
Technical Field
The invention belongs to the field of prediction of service life of a diaphragm pump check valve, and particularly relates to a method and a system for predicting service life of a diaphragm pump check valve based on MCNN.
Background
A diaphragm pump is a pumping device commonly used to deliver various liquids, which pushes the flow of the liquid by the up-and-down reciprocation of a diaphragm. Therefore, diaphragm pumps are widely used in industrial production, especially in the process industry, with their excellent corrosion resistance and flow controllability. In diaphragm pumps, a one-way valve is a critical component for controlling flow direction and preventing backflow of liquid. With the use of the device, the performance of the check valve is gradually degraded due to the severe working environment, and the performance of the pump is possibly reduced, even the pump is failed, so that the economic property is lost. Therefore, it becomes important to accurately predict the life of the diaphragm pump check valve. The service life of the diaphragm pump can be predicted to better help maintenance personnel to make the preparation for maintaining or replacing equipment in time. Reducing the influence on production.
In the existing studies on life prediction of diaphragm pumps, there are several problems:
a limitation that is common in current life prediction methods is that most rely on feature data extraction throughout the equipment operational cycle, combined with learning algorithms to accomplish life prediction. However, this method requires a high amount of life cycle data for large-scale devices, and obtaining enough full life cycle data samples in practical applications is difficult. However, due to the difficulty in data acquisition, such limitation may lead to erroneous judgment, which reduces the reliability and accuracy of life prediction results.
Particularly for the diaphragm pump one-way valve, there is a significant difference between the life cycle characteristic data presented by the different characteristic data. To effectively address this challenge, it is desirable to find a way to take advantage of these more significant feature data to improve the effectiveness of diaphragm pump check valve life predictions.
Disclosure of Invention
Against the problems existing in the background art. The invention provides a membrane pump check valve life prediction method and system based on MCNN (Multiple-input Convolutional Neural Networks). The invention realizes the life prediction of the diaphragm pump check valve, solves the problems that in engineering practice, the feature samples are few, the monitoring signal type is single, and the life prediction is difficult to be accurately performed, and is described in detail below:
s1, collecting data of the whole life cycle of a diaphragm pump check valve and constructing a data set, wherein the data set comprises noise signals, vibration signals and corresponding service time of the diaphragm pump check valve;
s2, preprocessing the collected diaphragm pump check valve data set, and then randomly dividing the data to obtain a training set and a testing set;
s3, inputting the divided training set into an MCNN model for training to obtain the MCNN model for generating the life prediction value of the diaphragm pump check valve;
s4, inputting the divided training set into a Wasserstein square distance-based antagonistic neural network for training, learning data distribution of original noise and vibration signals, generating filling values similar to the original data in statistical characteristics and time sequence characteristics, and seamlessly embedding the filling values into a test set according to the time sequence to realize expansion of the test set;
s5, standardizing the data test set of the one-way valve of the diaphragm pump after expansion, and ensuring that the numerical ranges of different characteristics are consistent;
s6, respectively inputting the test sets of the noise signals and the vibration signals after expansion into a trained MCNN model to obtain two groups of life prediction values, and obtaining the life prediction values of the diaphragm pump check valve after fusion according to the Wasserstein square distance and the Kalman gain formula.
Further, in the construction dataset, a life cycle dataset corresponding to two different signals is constructed by using a diaphragm pump one-way valve in a time auxiliary manner.
Further, the preprocessing includes missing value processing, abnormal value detection and processing, time sequence processing and data normalization processing for the sensor data and the workload data.
Further, the MCNN model for generating the life prediction value of the diaphragm pump check valve comprises a convolution layer, a pooling layer and a full connection layer; and (3) the data in the training set enter a full-connection layer to output a life prediction value of the diaphragm pump check valve through rolling and pooling operation.
Further, the antagonistic neural network based on the Wasserstein square distance specifically comprises: the LSTM network is used as a generator of the antagonistic neural network for generating co-distributed analog data with a time sequence.
Further, the training process is completed based on a minimum maximum cost function, and the training process of the antagonistic neural network based on the Wasserstein square distance comprises the following steps: the Wasserstein square distance is introduced as a loss function to replace JS and KL divergence, when the discriminator judges that false data generated by the generator is false, the loss function is used for adjusting weight parameters of input data in the LSTM generator by using a gradient descent method, so that the loss function in the discriminator is continuously changed towards a decreasing direction.
Further, the Wasserstein square distance is specifically:
representation->Has a joint distribution of->And->Is defined by a boundary of (2); />From which the +.f. can be sampled for each possible joint distribution gamma for the real data and the generated dummy data, respectively>Obtain->And->Samples, and calculate the distance of the pair of samples +.>And then canTo calculate the expected value of sample pair distance square +.>The method comprises the steps of carrying out a first treatment on the surface of the Although there is no overlap between the two data sample distributions, the WS distance can still calculate the relationship between the two samples, and its loss function has the characteristic of continuous smoothness, which can provide meaningful gradients for GAN networks.
Further, the fusion of the noise signal and the vibration signal is specifically:
according to the difference of Wasserstein square distance, the life prediction values with different values obtained by the noise signal and the vibration signal are respectively selected as t1 with small values and t2 with large values, and the formula is shown as follows:
and obtaining a weight e, and according to a Kalman gain formula after adjustment:
pt is the final predicted life result of the model and is output.
According to a third aspect of the specification, a system for implementing the method is disclosed, the system comprising: the system comprises a data acquisition module, a data preprocessing module, an countermeasure generation module, a prediction module and a fusion module;
the data acquisition module is used for acquiring data of the whole life cycle of the diaphragm pump check valve and constructing a data set, wherein the data set comprises noise signals, vibration signals and corresponding service time of the diaphragm pump check valve;
the data preprocessing module is used for preprocessing the collected diaphragm pump check valve data and then randomly dividing the data to obtain a training set and a testing set;
the countermeasure generation module is used for inputting the training set obtained by division into a countermeasure neural network based on Wasserstein square distance for training, learning data distribution of original noise and vibration signals, generating filling values similar to the original data in statistical characteristics and time sequence characteristics, and seamlessly embedding the filling values into a test set according to the time sequence to realize expansion of the test set;
the prediction module is used for inputting the training set obtained by division into the MCNN model for training to obtain the MCNN model for generating the life prediction value of the diaphragm pump check valve; respectively inputting the test sets of the noise signals and the vibration signals after expansion into a trained MCNN model to obtain two groups of life prediction values,
the fusion module is used for obtaining the fused predicted value of the service life of the one-way valve of the diaphragm pump according to the Wasserstein square distance and the Kalman gain formula.
According to a third aspect of the specification, a diaphragm pump check valve life prediction device based on MCNN is disclosed, and the device comprises a memory and one or more processors, wherein executable codes are stored in the memory, and when the processor executes the executable codes, the method for predicting the diaphragm pump check valve life based on the MCNN is realized.
The invention has the beneficial effects that:
the method adopts the MCNN model based on the antagonistic neural network, and effectively solves the challenge of data scarcity to the prediction model. By generating filling data similar to the original data distribution, the gap in the sample data is filled, and more information for training is provided for the model.
The method can accurately predict the different quality signals, and the prediction accuracy of the model is improved through ingenious fusion of the prediction values. Potential patterns and associations in the data are captured, resulting in more accurate and reliable results in the predictive task.
In a word, the method remarkably enhances the accuracy of the model by means of fusion of the filling data and the predicted value, can provide more reliable prediction under the condition of data scarcity, and provides feasibility and stability for the prediction model in practical application.
Drawings
FIG. 1 is a flowchart of an LSTM model for countermeasure generation provided by an embodiment of the present invention;
FIG. 2 is a diagram of an antagonistic neural network according to an embodiment of the present invention;
FIG. 3 is a diagram of a diaphragm pump check valve life prediction system according to an embodiment of the present invention;
fig. 4 is a diagram of a diaphragm pump check valve life prediction device based on an LSTM model generated based on countermeasure according to an embodiment of the present invention.
Detailed Description
The following describes the embodiments of the present invention in further detail with reference to the drawings.
With vibration and noise characteristics as input, the application provides a diaphragm pump check valve service life prediction method and system based on MCNN. We introduced wasperstein square distance instead of JS (Jensen-Shannon) and KL (Kullback-Leibler) divergence to build the antagonistic neural network. The generator is designed to generate data consistent with the profile of the feature data, and the discriminator optimizes the generator data to obtain a profile matching the original data set and feeds it as input into the depth CNN model for prediction. At the same time, vibration and noise characteristics are synchronously transferred into the life prediction model. By comparing the difference of the two Wasserstein square distances, we assign different weights to the two signal results, and achieve effective fusion of the two signal results. The method not only can effectively predict the service life of the one-way valve of the diaphragm pump, but also improves the robustness and accuracy of a prediction model in the process of fusing vibration and noise characteristics. The invention discloses a diaphragm pump check valve service life prediction method and a diaphragm pump check valve service life prediction system based on MCNN, and the method and the system comprise the following steps as shown in figure 1:
s01, signal acquisition: and collecting multidimensional sensor data of the whole life cycle of the diaphragm pump check valve, wherein the multidimensional sensor data comprise vibration and noise signals.
And the vibration signal sensor is used for respectively measuring vibration and noise signals in the whole life cycle of the diaphragm pump check valve, and analyzing the changes of the vibration and noise signals.
The vibration sensor records the vibration mode of the diaphragm pump check valve when in operation, the noise signal sensor captures sound characteristics related to the running state, and the data are used for constructing a comprehensive life cycle data set by recording the service time of the corresponding diaphragm pump check valve.
S02: data preprocessing: the acquired sensor life cycle data set is cleaned and processed, missing value processing, abnormal value detection and processing, time sequence processing and data normalization processing are carried out, and the processed data is divided into a training set and a testing set with time sequences according to a data division strategy.
And carrying out normalization processing on the acquired sensor data, and dividing the processed data into training and testing sets.
Min-max normalized conversion formula:
where max is the maximum value of the sample data and min is the minimum value of the sample data. The normalization with min-max can be the case in a numerical comparison set. The dimensionless expression is changed into the dimensionless expression, so that indexes with different units or magnitudes can be compared and weighted conveniently. After normalization, the dimensionalized dataset is changed into a scalar.
The data are randomly divided into a training set and a test set, and the dividing ratio is 0.8 and 0.2.
S03, inputting the divided training set into an MCNN model, wherein the training set input into the MCNN model is subjected to operations such as convolution pooling and the like in the model to finish training of the training set, and CNN generally comprises a convolution layer, a pooling layer and a full connection layer. The convolution operation generates an output signature by sliding a small window (convolution kernel) over the input data, performing the operations of element multiplication and summation. Convolution theory of operation:
s (i, j) is the output of the convolution result.Is an input image. />Is a convolution kernel (filter). (i, j) is the pixel coordinates in the output image. (m, n) is the weight coordinates in the convolution kernel and b is the offset.
By the pooling operation, the size of the feature map is reduced while important information is maintained, so that the computational complexity of the model is reduced. The method comprises the following specific steps:
y (i, j) represents the output of the maximum pooling, max represents the maximum operation, (i\times S) and (j\times S) are the starting position of the pooling window, (S) is the stride (stride), and the sliding step of the pooling window is controlled.
After pooling, the outputs of the rolling and pooling layers are connected to the final output by a full connection layer. The fully connected operation is expressed as:
z is the life prediction value of the output diaphragm pump check valve, W is a weight matrix,is the data input to the feature vector, i.e., output by the pooling layer, b is the bias, and L is the activation function, e.g., reLU. When training, the user is strapped>The characteristic information is characteristic information under a certain working condition in the training set, Z is service life information of the diaphragm pump check valve under the working condition in the training set, and the W weight matrix and the bias b are obtained through training.
S04, as shown in fig. 2, the divided training set is input into a generator in the GAN network, false data output by the generator is input into a discriminator for comparison, and whether a feedback adjustment generator model is needed or not is judged through the true and false output by the discriminator;
wherein the generator is an LSTM model, and the specific principle is as follows:
each LSTM neuron in the LSTM network topology has three gate functions: forgetting gate, input gate and output gate, the gating unit in the LSTM neuron can store and delete information, ensuring dependency between LSTM neuron learning and processing time series data. The input gate is responsible for selectively parsing the information of the input neuron state, the forget gate decides which information to discard in the neuron state, and the output gate decides which information to output from the neuron state. The principle is as follows:
wherein the method comprises the steps ofIs the output of the input node,/>Is the output of the input gate,/->Is the output of the forget gate, < >>Is the output of the output gate; />Is input data +.>Weights of (2); />Respectively +.>LSTM neurons;/>representing the deviation of the input node +.>Representing the deviation of the input gate +.>Deviation indicative of forgetful door->Indicating the deviation of the output gate; />Representing a sigmoid activation function,/->Representing a tanh activation function; />Is at->LSTM neuron state at time, +.>Is at->LSTM neuron state at time, +.>Representing multiplication->Indicating the final output result.
As shown in fig. 3, according to the basic framework of the GAN network, taking noise training as an example: the random noise z is input to a generator G which learns to generate from a generator LSTM modelSpoofing the discriminator to let the discriminator believe +>Is the actual data x real Is a part of (a); the discriminator D uses the real data x real And->The Wasserstein square distance between the two is used for distinguishing real data from false data, and the loss function of the training process of the existing GAN network is the minimum maximum cost function:
to minimize the loss of the generator G and maximize the loss of the discriminator D, E is expressed as a mathematical expectation of the data, subscripts X-P date(x) Is the real data sampled in the sample, subscripts Z-P z(z) Finger generating a random noise z; g (z) is data generated by the generator by noise z, and D (z)) represents a probability that this generated data is real data. D (G (z))=sigma (D (z)), D (z) being the original output of the discriminator.
When the training effect of the countermeasure network discriminator is best, the generator can have serious gradient vanishing problem, and meanwhile, the condition of unstable network training exists.
To solve the above problem, the present embodiment introduces wasperstein squared distance (WS distance) as a loss function instead of JS and KL divergence, defined as:
representation->Has a joint distribution of P data And P g Boundary of-> , />From which the +.f. can be sampled for each possible joint distribution gamma for the real data and the generated dummy data, respectively>ObtainingAnd->Samples, and calculate the distance of the pair of samples +.>Then, the expected value of sample pair distance square +.>The method comprises the steps of carrying out a first treatment on the surface of the Although there is no overlap between the two data sample distributions, the WS distance can still calculate the relationship between the two samples, and its loss function has the characteristic of continuous smoothness, which can provide meaningful gradients for GAN networks.
When the discriminator judges that the false data generated by the generator is false, the loss function is used for realizing the weight parameter of the input data in the LSTM generator by using a gradient descent methodThe adjustment is made such that the intra-discriminator loss function is continually changing in a decreasing direction. Until the discriminator model outputs a true result. The part of the false data evaluated as true is input into the MCNN model of the next step.
S05, after the MCNN model receives the test set from the original data and the false data generated by the LSTM generator, the MCNN model ensures that the numerical ranges of different features are consistent according to the respective time sequences after one-to-one correspondence, and fusion is completed in the test set, so that data filling is realized.
S06, respectively inputting test sets of different quality sensor data into the trained MCNN model, and obtaining life prediction values according to the respective test sets. The predicted values of the different values of the above quantities are selected as t1 with small values and as t2 with large values according to the difference of the Wasserstein square distance, and the formula is as follows:
and obtaining a weight e, and according to a Kalman gain formula after adjustment:
pt is the final predicted life result of the model and is output.
The method innovatively overcomes the dilemma in the absence of sufficient available training data. By introducing co-distributed data filling techniques, we have carefully designed filling on the sample data to better meet the stringent requirements of the predictive model on the number of sample data. This step not only provides more learning opportunities for the model, but also successfully simulates the distribution of samples in the absence of real samples, providing more information for training of the model.
By adopting a method for fusing predicted values of different quality signals and comprehensively analyzing a plurality of signal sources, the dynamic process of the system life evolution can be understood, so that the accuracy and the robustness of the predicted result are improved. This fusion strategy represents a significant advantage in terms of improving model performance.
According to another aspect of the embodiment of the present invention, there is also disclosed a MCNN-based diaphragm pump check valve life prediction system, as shown in fig. 3, including: the system comprises a data acquisition module, a data preprocessing module, an countermeasure generation module, a prediction module and a fusion module;
the data acquisition module is used for acquiring data of the whole life cycle of the diaphragm pump check valve and constructing a data set, wherein the data set comprises noise signals, vibration signals and corresponding service time of the diaphragm pump check valve;
the data preprocessing module is used for preprocessing the collected diaphragm pump check valve data and then randomly dividing the data to obtain a training set and a testing set;
the countermeasure generation module is used for inputting the training set obtained by division into a countermeasure neural network based on Wasserstein square distance for training, learning data distribution of original noise and vibration signals, generating filling values similar to the original data in statistical characteristics and time sequence characteristics, and seamlessly embedding the filling values into a test set according to the time sequence to realize expansion of the test set;
the prediction module is used for inputting the training set obtained by division into the MCNN model for training to obtain the MCNN model for generating the life prediction value of the diaphragm pump check valve; respectively inputting the test sets of the noise signals and the vibration signals after expansion into a trained MCNN model to obtain two groups of life prediction values,
the fusion module is used for obtaining the fused predicted value of the service life of the one-way valve of the diaphragm pump according to the Wasserstein square distance and the Kalman gain formula.
Corresponding to the embodiment of the method for predicting the service life of the diaphragm pump check valve based on the MCNN, the invention also provides the embodiment of the device for predicting the service life of the diaphragm pump check valve based on the MCNN.
Referring to fig. 4, the device for predicting the service life of the diaphragm pump check valve based on the MCNN according to the embodiment of the present invention includes a memory and one or more processors, where the memory stores executable codes, and the processors are configured to implement the method for predicting the service life of the diaphragm pump check valve based on the MCNN according to the above embodiment when executing the executable codes.
The embodiment of the MCNN-based diaphragm pump check valve life prediction device can be applied to any equipment with data processing capability, and the equipment with data processing capability can be equipment or a device such as a computer. The apparatus embodiments may be implemented by software, or may be implemented by hardware or a combination of hardware and software. Taking software implementation as an example, the device in a logic sense is formed by reading corresponding computer program instructions in a nonvolatile memory into a memory by a processor of any device with data processing capability. From the hardware level, as shown in fig. 4, a hardware structure diagram of an apparatus with optional data processing capability, where the MCNN-based diaphragm pump check valve life prediction device is located, is provided in the embodiment, except for the processor, the memory, the network interface, and the nonvolatile memory shown in fig. 4, where the apparatus with optional data processing capability is located, generally according to the actual function of the apparatus with optional data processing capability, other hardware may also be included, which will not be described herein.
The implementation process of the functions and roles of each unit in the above device is specifically shown in the implementation process of the corresponding steps in the above method, and will not be described herein again.
For the device embodiments, reference is made to the description of the method embodiments for the relevant points, since they essentially correspond to the method embodiments. The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purposes of the present invention. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
The embodiment of the invention also provides a computer readable storage medium, on which a program is stored, which when executed by a processor, implements a MCNN-based diaphragm pump check valve life prediction method in the above embodiment.
The computer readable storage medium may be an internal storage unit, such as a hard disk or a memory, of any of the data processing enabled devices described in any of the previous embodiments. The computer readable storage medium may be any external storage device that has data processing capability, such as a plug-in hard disk, a Smart Media Card (SMC), an SD Card, a Flash memory Card (Flash Card), or the like, which are provided on the device. Further, the computer readable storage medium may include both internal storage units and external storage devices of any data processing device. The computer readable storage medium is used for storing the computer program and other programs and data required by the arbitrary data processing apparatus, and may also be used for temporarily storing data that has been output or is to be output.
Those skilled in the art will appreciate that the drawings are schematic representations of only one preferred embodiment, and that the above-described embodiment numbers are merely for illustration purposes and do not represent advantages or disadvantages of the embodiments.
The above-described embodiments are intended to illustrate the present invention, not to limit it, and any modifications and variations made thereto are within the spirit of the invention and the scope of the appended claims.
Claims (10)
1. The MCNN-based diaphragm pump check valve life prediction method is characterized by comprising the following steps of:
s1, collecting data of the whole life cycle of a diaphragm pump check valve and constructing a data set, wherein the data set comprises noise signals, vibration signals and corresponding service time of the diaphragm pump check valve;
s2, preprocessing the collected diaphragm pump check valve data set, and then randomly dividing the data to obtain a training set and a testing set;
s3, inputting the divided training set into an MCNN model for training to obtain the MCNN model for generating the life prediction value of the diaphragm pump check valve;
s4, inputting the divided training set into a Wasserstein square distance-based antagonistic neural network for training, learning data distribution of original noise and vibration signals, generating filling values similar to the original data in statistical characteristics and time sequence characteristics, and seamlessly embedding the filling values into a test set according to the time sequence to realize expansion of the test set;
s5, standardizing the data test set of the one-way valve of the diaphragm pump after expansion, and ensuring that the numerical ranges of different characteristics are consistent;
s6, respectively inputting the test sets of the noise signals and the vibration signals after expansion into a trained MCNN model to obtain two groups of life prediction values, and obtaining the life prediction values of the diaphragm pump check valve after fusion according to the Wasserstein square distance and the Kalman gain formula.
2. The MCNN-based method for predicting the lifetime of a diaphragm pump check valve as in claim 1, wherein in the constructing dataset, the lifetime dataset corresponding to two different signals is constructed using a diaphragm pump check valve using time assistance.
3. The MCNN-based diaphragm pump check valve life prediction method according to claim 1, wherein the preprocessing includes missing value processing, outlier detection and processing, time series processing, and data normalization processing of the sensor data and the workload data.
4. The MCNN-based membrane pump check valve life prediction method according to claim 1, wherein the MCNN model for generating a life prediction value of a membrane pump check valve includes a convolution layer, a pooling layer and a full connection layer; and (3) the data in the training set enter a full-connection layer to output a life prediction value of the diaphragm pump check valve through rolling and pooling operation.
5. The MCNN-based diaphragm pump check valve life prediction method according to claim 1, wherein the wasperstein square distance-based antagonistic neural network specifically includes: the LSTM network is used as a generator of the antagonistic neural network for generating co-distributed analog data with a time sequence.
6. The MCNN-based diaphragm pump check valve life prediction method of claim 5, wherein the training process is completed based on a minimum cost function, and the wasperstein-squared distance-based training process for the antagonistic neural network includes: the Wasserstein square distance is introduced as a loss function to replace JS and KL divergence, when the discriminator judges that false data generated by the generator is false, the loss function is used for adjusting weight parameters of input data in the LSTM generator by using a gradient descent method, so that the loss function in the discriminator is continuously changed towards a decreasing direction.
7. The MCNN-based diaphragm pump check valve life predicting method of claim 6, wherein the wasperstein squared distance is specifically
;
Representation->Has a joint distribution of->And->Is defined by a boundary of (2); />Sampling +.>Obtain->And->Samples, and calculate the distance of the pair of samples +.>Then calculating the expected value of sample pair distance square under the joint distribution gamma>The method comprises the steps of carrying out a first treatment on the surface of the Although there is no overlap between the two data sample distributions, the WS distance can still calculate the relationship between the two samples, and its loss function has the characteristic of continuous smoothness, which can provide meaningful gradients for GAN networks.
8. The MCNN-based diaphragm pump check valve life prediction method according to claim 1, wherein the fusion of the noise signal and the vibration signal is specifically:
according to the difference of Wasserstein square distance, the life prediction values with different values obtained by the noise signal and the vibration signal are respectively selected as t1 with small values and t2 with large values, and the formula is shown as follows:
;
and obtaining a weight e, and according to a Kalman gain formula after adjustment:
;
pt is the final predicted life result of the model and is output.
9. A system for implementing the method of any one of claims 1-8, the system comprising: the system comprises a data acquisition module, a data preprocessing module, an countermeasure generation module, a prediction module and a fusion module;
the data acquisition module is used for acquiring data of the whole life cycle of the diaphragm pump check valve and constructing a data set, wherein the data set comprises noise signals, vibration signals and corresponding service time of the diaphragm pump check valve;
the data preprocessing module is used for preprocessing the collected diaphragm pump check valve data and then randomly dividing the data to obtain a training set and a testing set;
the countermeasure generation module is used for inputting the training set obtained by division into a countermeasure neural network based on Wasserstein square distance for training, learning data distribution of original noise and vibration signals, generating filling values similar to the original data in statistical characteristics and time sequence characteristics, and seamlessly embedding the filling values into a test set according to the time sequence to realize expansion of the test set;
the prediction module is used for inputting the training set obtained by division into the MCNN model for training to obtain the MCNN model for generating the life prediction value of the diaphragm pump check valve; respectively inputting the test sets of the noise signals and the vibration signals after expansion into a trained MCNN model to obtain two groups of life prediction values,
the fusion module is used for obtaining the fused predicted value of the service life of the one-way valve of the diaphragm pump according to the Wasserstein square distance and the Kalman gain formula.
10. A MCNN-based diaphragm pump check valve life prediction apparatus comprising a memory and one or more processors, the memory having executable code stored therein, wherein the processor, when executing the executable code, implements a MCNN-based diaphragm pump check valve life prediction method as claimed in any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410022049.0A CN117556261B (en) | 2024-01-08 | 2024-01-08 | MCNN-based diaphragm pump check valve service life prediction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410022049.0A CN117556261B (en) | 2024-01-08 | 2024-01-08 | MCNN-based diaphragm pump check valve service life prediction method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117556261A true CN117556261A (en) | 2024-02-13 |
| CN117556261B CN117556261B (en) | 2024-05-14 |
Family
ID=89818786
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410022049.0A Active CN117556261B (en) | 2024-01-08 | 2024-01-08 | MCNN-based diaphragm pump check valve service life prediction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117556261B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111832216A (en) * | 2020-04-14 | 2020-10-27 | 新疆大学 | Rolling bearing residual service life prediction method based on EEMD-MCNN-GRU |
| CN112149316A (en) * | 2019-11-04 | 2020-12-29 | 中国人民解放军国防科技大学 | Aero-engine residual life prediction method based on improved CNN model |
| US20210197335A1 (en) * | 2019-12-26 | 2021-07-01 | Dalian University Of Technology | Data Augmentation Method Based On Generative Adversarial Networks In Tool Condition Monitoring |
| CN114186500A (en) * | 2022-02-16 | 2022-03-15 | 华中科技大学 | Marine bearing residual life prediction method based on transfer learning and multiple time windows |
| KR20220110113A (en) * | 2021-01-29 | 2022-08-05 | 고려대학교 산학협력단 | Device and method for fault detection in time series data using 1d cnn and gan |
| CN115510612A (en) * | 2022-08-24 | 2022-12-23 | 广州大学 | Method for predicting remaining service life of lithium ion battery |
| CN115618710A (en) * | 2022-09-08 | 2023-01-17 | 济南大学 | Wind power probabilistic prediction method and system based on GAN |
| CN115828737A (en) * | 2022-11-17 | 2023-03-21 | 北京航空航天大学 | Circuit board residual life prediction method based on fault physics and data drive fusion |
| US20230153608A1 (en) * | 2021-11-15 | 2023-05-18 | East China Jiaotong University | Method for predicting remaining useful life of railway train bearing based on can-lstm |
| CN116226613A (en) * | 2023-03-10 | 2023-06-06 | 江苏大学 | Rolling bearing residual life prediction method, equipment and storage medium |
| CN117213856A (en) * | 2023-06-13 | 2023-12-12 | 贵州大学 | Bearing residual life prediction method based on Yun Bian double-data-source fusion |
-
2024
- 2024-01-08 CN CN202410022049.0A patent/CN117556261B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112149316A (en) * | 2019-11-04 | 2020-12-29 | 中国人民解放军国防科技大学 | Aero-engine residual life prediction method based on improved CNN model |
| US20210197335A1 (en) * | 2019-12-26 | 2021-07-01 | Dalian University Of Technology | Data Augmentation Method Based On Generative Adversarial Networks In Tool Condition Monitoring |
| CN111832216A (en) * | 2020-04-14 | 2020-10-27 | 新疆大学 | Rolling bearing residual service life prediction method based on EEMD-MCNN-GRU |
| KR20220110113A (en) * | 2021-01-29 | 2022-08-05 | 고려대학교 산학협력단 | Device and method for fault detection in time series data using 1d cnn and gan |
| US20230153608A1 (en) * | 2021-11-15 | 2023-05-18 | East China Jiaotong University | Method for predicting remaining useful life of railway train bearing based on can-lstm |
| CN114186500A (en) * | 2022-02-16 | 2022-03-15 | 华中科技大学 | Marine bearing residual life prediction method based on transfer learning and multiple time windows |
| CN115510612A (en) * | 2022-08-24 | 2022-12-23 | 广州大学 | Method for predicting remaining service life of lithium ion battery |
| CN115618710A (en) * | 2022-09-08 | 2023-01-17 | 济南大学 | Wind power probabilistic prediction method and system based on GAN |
| CN115828737A (en) * | 2022-11-17 | 2023-03-21 | 北京航空航天大学 | Circuit board residual life prediction method based on fault physics and data drive fusion |
| CN116226613A (en) * | 2023-03-10 | 2023-06-06 | 江苏大学 | Rolling bearing residual life prediction method, equipment and storage medium |
| CN117213856A (en) * | 2023-06-13 | 2023-12-12 | 贵州大学 | Bearing residual life prediction method based on Yun Bian double-data-source fusion |
Non-Patent Citations (4)
| Title |
|---|
| 伍建军;刘海平;廖泰健;叶祥;: "基于多传感器数据融合的破碎机辊套剩余寿命预测新方法", 机械设计与研究, no. 05, 20 October 2017 (2017-10-20) * |
| 孙鑫;孙维堂;: "基于多尺度卷积神经网络的轴承剩余寿命预测", 组合机床与自动化加工技术, no. 10, 20 October 2020 (2020-10-20) * |
| 张旺;陈磊;陈超宇;韩捷;: "Hilbert-全矢HMM轴承剩余寿命预测", 机械设计与制造, no. 03, 8 March 2020 (2020-03-08) * |
| 郭锐;赵之谦;贾鑫龙;赵静一;张生;: "基于ANFIS的外啮合齿轮泵寿命预测研究", 仪器仪表学报, no. 01, 2 March 2020 (2020-03-02) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117556261B (en) | 2024-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11334407B2 (en) | Abnormality detection system, abnormality detection method, abnormality detection program, and method for generating learned model | |
| Huang et al. | Shallow convolutional neural networks for human activity recognition using wearable sensors | |
| Gamboa | Deep learning for time-series analysis | |
| CN110427654B (en) | Landslide prediction model construction method and system based on sensitive state | |
| Zio | A study of the bootstrap method for estimating the accuracy of artificial neural networks in predicting nuclear transient processes | |
| CN112655004A (en) | Computer-implemented method, computer program product, and system for anomaly detection and/or predictive maintenance | |
| Ayodeji et al. | Causal augmented ConvNet: A temporal memory dilated convolution model for long-sequence time series prediction | |
| US11113597B2 (en) | Artificial neural network and method of training an artificial neural network with epigenetic neurogenesis | |
| CN108709745A (en) | One kind being based on enhanced LPP algorithms and the quick bearing fault recognition method of extreme learning machine | |
| CN109145516B (en) | Analog circuit fault identification method based on improved extreme learning machine | |
| Wang et al. | Learning efficient binarized object detectors with information compression | |
| CN114297036A (en) | Data processing method and device, electronic equipment and readable storage medium | |
| CN113887342A (en) | Equipment fault diagnosis method based on multi-source signals and deep learning | |
| CN114091554A (en) | Training set processing method and device | |
| CN117041017A (en) | Intelligent operation and maintenance management method and system for data center | |
| Rahadian et al. | Image encoding selection based on Pearson correlation coefficient for time series anomaly detection | |
| JP2019105871A (en) | Abnormality candidate extraction program, abnormality candidate extraction method and abnormality candidate extraction apparatus | |
| CN114547365A (en) | Image retrieval method and device | |
| Moskalenko et al. | Improving the effectiveness of training the on-board object detection system for a compact unmanned aerial vehicle | |
| Pandhiani et al. | Time series forecasting by using hybrid models for monthly streamflow data | |
| Qi et al. | Fault detection and localization in distributed systems using recurrent convolutional neural networks | |
| CN117079017A (en) | Credible small sample image identification and classification method | |
| CN117556261B (en) | MCNN-based diaphragm pump check valve service life prediction method and system | |
| CN111798518A (en) | Mechanical arm posture detection method, device and equipment and computer storage medium | |
| CN107220710B (en) | Calculation method of BN model parameters under condition of scarce sample data set |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |