CN114741942A - Fault diagnosis device and method for ocean platform reciprocating compressor based on machine learning - Google Patents
Fault diagnosis device and method for ocean platform reciprocating compressor based on machine learning Download PDFInfo
- Publication number
- CN114741942A CN114741942A CN202111217412.7A CN202111217412A CN114741942A CN 114741942 A CN114741942 A CN 114741942A CN 202111217412 A CN202111217412 A CN 202111217412A CN 114741942 A CN114741942 A CN 114741942A
- Authority
- CN
- China
- Prior art keywords
- indicator diagram
- indicator
- reciprocating compressor
- module
- fault
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Control Of Positive-Displacement Pumps (AREA)
Abstract
The invention discloses a fault diagnosis device and a fault diagnosis method for an ocean platform reciprocating compressor based on machine learning, wherein the device comprises the following steps: the system comprises a monitoring module, a processing module, a judging module, a database module, a recording module and an early warning module, wherein the processing module generates an actual indicator diagram according to data of the monitoring module, and combines the actual indicator diagram and a normal indicator diagram curve in a stable operation state to be displayed in the same indicator diagram sample; the judging module is connected with the processing module, the convolutional neural network provides a fault diagnosis result of the reciprocating compressor according to the indicator diagram sample provided by the processing module, and the early warning module visualizes the diagnosis result and provides a health management scheme. According to the fault diagnosis device for the ocean platform reciprocating compressor based on the machine learning, the characteristic extraction and the self-learning capability of the convolutional neural network are used for carrying out indicator diagram classification and fault identification on the reciprocating compressor, and support and guarantee are provided for the predictive maintenance of the ocean platform reciprocating compressor in the future.
Description
Technical Field
The invention belongs to the field of state recognition and fault diagnosis, and particularly relates to a fault diagnosis device and a fault diagnosis method for an ocean platform reciprocating compressor based on convolutional neural network machine learning.
Background
The reciprocating compressor is important equipment for oil and gas development and transportation of an offshore platform, and the safe and reliable operation of the reciprocating compressor is very important for the operation of the whole device. Due to the complex structure, a plurality of easily damaged parts, the reciprocating compressor has high state identification difficulty and high failure rate. The pressure and temperature of the gas in the cylinder are important characteristics for fault judgment of a reciprocating compressor gas valve, a piston, the cylinder and the like. The most effective way to determine the overall health of a reciprocating gas compressor is by checking the cylinder chamber pressure distribution. The indicator diagram is constructed by using the signals of the pressure in the cylinder, the faults such as air valve fault, piston ring abrasion and the like can cause the change of the pressure in the cylinder and the thermal process, and simultaneously, the interference of the signal confusion phenomenon caused by vibration to the diagnosis result is avoided. Aiming at the air valve fault of the compressor, compared with a vibration signal, an acoustic emission signal, other pressure signals and the like, the indicator diagram diagnosis is the most accurate diagnosis with the minimum noise interference, and the same fault corresponds to the same shape change and is not influenced by working conditions, structures and the like, so that the diagnosis is more suitable for the diagnosis of variable working conditions. The testing means for obtaining the cylinder pressure signal under the condition of no damage to the cylinder wall is developed and applied, so that the indicator diagram fault diagnosis method is rapidly popularized, and the indicator diagram gradually becomes a specific parameterized diagnosis method of the reciprocating compressor.
The state of the reciprocating compressor can be judged according to the change of the indicator diagram graph and the fault can be identified, machine learning can be used for replacing manual diagnosis along with the development of artificial intelligence, image difference can be learned by self, image features can be automatically extracted, the state of the reciprocating compressor can be intelligently identified, and the fault can be diagnosed. Real-time data monitoring and acquisition are combined, various state monitoring signals of the reciprocating compressor are fused, fault feature self-extraction based on machine learning under a large amount of data can be achieved, and self-classification and diagnosis of fault states are completed.
Disclosure of Invention
The present invention is directed to solving, at least to some extent, one of the above-mentioned problems in the prior art. In view of the above, the present invention needs to provide a device and a method for diagnosing a fault of an ocean platform reciprocating compressor based on machine learning.
The fault diagnosis device for the ocean platform reciprocating compressor based on the machine learning comprises the following components: the monitoring module is connected with the reciprocating compressor to monitor the pressure in the cylinder in real time and obtain a crank angle; the processing module is connected with the monitoring module, generates an actual indicator diagram according to the data of the monitoring module, and combines the actual indicator diagram and a normal indicator diagram curve in a stable running state in the same indicator diagram sample for display; the judging module is connected with the processing module and comprises a convolutional neural network, and the convolutional neural network provides a fault diagnosis result of the reciprocating compressor according to the indicator diagram sample provided by the processing module; the database module is connected with the judging module and is an indicator diagram fault sample database for the convolutional neural network to train hyper-parameters; the recording module is connected with the judging module and the database module and is used for storing and recording the fault indicator diagram and expanding the database module; and the early warning module is connected with the judging module and used for displaying the result of the judging module to give early warning and providing a corresponding health management scheme.
According to the fault diagnosis device of the ocean platform reciprocating compressor based on the machine learning, the fault diagnosis method research of the reciprocating compressor based on the machine learning is developed on the basis of the indicator diagram of the reciprocating compressor, the convolutional neural network is proposed to be used for classifying and fault diagnosing the indicator diagram of the reciprocating compressor, meanwhile, a normal indicator curve and an actually measured indicator curve are placed in the same indicator diagram, indicator diagram classification and fault recognition of the reciprocating compressor are carried out by means of feature extraction and self-learning capacity of the convolutional neural network, and support and guarantee are provided for predictive maintenance of the ocean platform reciprocating compressor in the future.
In addition, the fault diagnosis device for the ocean platform reciprocating compressor based on machine learning according to the above embodiment of the invention may also have the following additional technical features:
according to one embodiment of the invention, the monitoring module comprises a pressure sensor connected to the reciprocating compressor for collecting dynamic pressure signals at various conditions within the reciprocating compressor cylinder and a key phase sensor for collecting key phase signals at various conditions within the compound compressor cylinder.
According to one embodiment of the invention, the database module is connected with the judgment module and is used for training the indicator diagram fault sample database of the hyperparameter by the convolutional neural network in a theoretical simulation mode and a numerical simulation mode.
The fault diagnosis method for the ocean platform reciprocating compressor based on the machine learning comprises the following steps:
step 1: obtaining real-time in-cylinder pressure and a crank angle;
step 2: an indicator diagram is generated using the dynamic pressure signal in combination with the key phase signal. Taking the indicator diagram curve in a stable running state as a normal indicator diagram curve, combining the normal indicator diagram curve with an actual indicator diagram curve generated by real-time acquisition and displaying in the same indicator diagram, and enhancing the contrast;
and step 3: and generating an indicator diagram sample database. A part of indicator diagram samples can be generated according to theoretical simulation by using the structural size of the cylinder and the rated intake and exhaust pressure, and then a part of real indicator diagram samples are acquired according to actual compressor operation and tests. Adding training labels to indicator diagram samples, wherein the indicator diagram samples in the same fault state correspond to the same label;
and 4, step 4: dividing the indicator diagram sample into a training set and a testing set, and sending the training set into a configured convolutional neural network for parameter training to finish the training of the network when certain recognition accuracy and stability are achieved;
and 5: and receiving the real indicator diagram collected in real time, inputting the actual indicator diagram into the trained convolutional neural network for outputting the classification result, and realizing the state identification and fault diagnosis of the reciprocating compressor.
According to the fault diagnosis method of the ocean platform reciprocating compressor based on the machine learning, the fault diagnosis method research of the reciprocating compressor based on the machine learning is carried out on the basis of the indicator diagram of the reciprocating compressor, the convolutional neural network is proposed to be used for classifying and fault diagnosing the indicator diagram of the reciprocating compressor, meanwhile, a normal indicator curve and an actually measured indicator curve are placed in the same indicator diagram, indicator diagram classification and fault recognition of the reciprocating compressor are carried out by means of feature extraction and self-learning capacity of the convolutional neural network, and support and guarantee are provided for predictive maintenance of the ocean platform reciprocating compressor in the future.
In addition, the method for diagnosing the fault of the ocean platform reciprocating compressor based on the machine learning according to the embodiment of the invention can also have the following additional technical characteristics:
according to one embodiment of the present invention, in step 1, a dynamic pressure signal and a key phase signal in each state in a cylinder are collected using a pressure sensor and a key phase sensor.
According to an embodiment of the present invention, the specific steps of step 2 are as follows:
step 2.1: a normalization operation is performed by dividing the dynamic pressure signal and the corresponding volume signal over a period by their maximum values.
While keeping the relative characteristic change, so as to deal with the state of the variable working condition;
step 2.2: and drawing the normalized pressure signal and volume signal into a indicator curve, taking the indicator curve in a stable running state as a normal indicator curve, combining the normal indicator curve with an actual indicator curve generated by real-time acquisition, displaying the indicator curve and the actual indicator curve in the same indicator graph, wherein the normal indicator curve is represented by a green line, the fault indicator curve or the real-time indicator curve is represented by a red line, and adding a label to generate a trainable indicator graph sample.
According to an embodiment of the present invention, in step 3, the training fault database of the convolutional neural network may be composed of a theoretical function simulation indicator diagram, a numerical simulation software simulation indicator diagram, and a fault test actual measurement indicator diagram, so as to form a fault training database of a certain scale.
According to an embodiment of the present invention, the specific steps of enhancing the training samples with the data in step 4 are as follows:
step 4.1: through the normalized indicator diagram, the pressure ratios under different working conditions are different, so that the training set is subjected to data enhancement, namely, the image is subjected to up-down fixed amount of deviation, the number of samples is increased, and the generalization capability of the sample model is enhanced.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
Fig. 1 is a schematic structural diagram of a fault diagnosis device for an ocean platform reciprocating compressor based on machine learning according to an embodiment of the invention.
FIG. 2 is a flow chart of a method for machine learning based marine platform reciprocating compressor fault diagnosis according to an embodiment of the present invention.
Fig. 3 is a flow chart of a fault diagnosis method for an ocean platform reciprocating compressor based on machine learning according to an embodiment of the invention.
Figure 4 simulates an exemplary indicator diagram of reciprocating compressor inlet valve leakage.
Figure 5 simulates an exemplary indicator diagram of reciprocating compressor inlet valve leakage.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
In the description of the present invention, it is to be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", and the like, indicate orientations and positional relationships based on those shown in the drawings, and are used only for convenience of description and simplicity of description, and do not indicate or imply that the device or element being referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, should not be considered as limiting the present invention.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or to implicitly indicate the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
In the present invention, unless otherwise expressly stated or limited, the terms "mounted," "connected," "secured," and the like are to be construed broadly and can, for example, be fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; either directly or indirectly through intervening media, either internally or in any other relationship. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
In the present invention, unless otherwise expressly stated or limited, "above" or "below" a first feature means that the first and second features are in direct contact, or that the first and second features are not in direct contact but are in contact with each other via another feature therebetween. Also, the first feature being "on," "above" and "over" the second feature includes the first feature being directly on and obliquely above the second feature, or merely indicating that the first feature is at a higher level than the second feature. A first feature being "under," "below," and "beneath" a second feature includes the first feature being directly under and obliquely below the second feature, or simply meaning that the first feature is at a lesser elevation than the second feature.
As shown in fig. 1, the marine platform reciprocating compressor fault diagnosis device according to the embodiment of the present invention comprises: the system comprises a monitoring module 10, a processing module 20, a judging module 30, a database module 40, a recording module 50 and an early warning module 60.
Specifically, the monitoring module 10 is coupled to the reciprocating compressor 70 to monitor in-cylinder pressure and obtain crank angle in real time.
The processing module 20 is connected to the monitoring module 10, and generates an actual indicator diagram according to the data of the monitoring module 10, and combines the actual indicator diagram and a normal indicator diagram curve in a steady operation state in the same indicator diagram sample for display. In other words, the indicator diagram curve in the steady state operation is taken as the normal indicator diagram curve, and the normal indicator diagram curve and the actual indicator diagram curve generated by real-time acquisition are combined in the same indicator diagram for displaying.
The judging module 30 is connected to the processing module 20, and the judging module 30 includes a convolutional neural network, and the convolutional neural network provides a fault diagnosis result of the reciprocating compressor according to the indicator diagram sample provided by the processing module 20.
The database module 40 is connected with the judging module 30, and the database module 40 is an indicator diagram fault sample database for convolutional neural network training hyper-parameters.
The recording module 50 is connected to the judging module 30 and the database module 40, and is configured to store and record the failure indicator diagram, and extend the database module 40.
The early warning module 60 is connected with the judging module 30, and sends out fault early warning according to the judging module and constructs and provides a health management scheme according to a knowledge theory.
According to the fault diagnosis device of the ocean platform reciprocating compressor based on the machine learning, the fault diagnosis method research of the reciprocating compressor based on the machine learning is developed on the basis of the indicator diagram of the reciprocating compressor, the convolutional neural network is proposed to be used for classifying and fault diagnosing the indicator diagram of the reciprocating compressor, meanwhile, a normal indicator curve and an actually measured indicator curve are placed in the same indicator diagram, indicator diagram classification and fault recognition of the reciprocating compressor are carried out by means of feature extraction and self-learning capacity of the convolutional neural network, and support and guarantee are provided for predictive maintenance of the ocean platform reciprocating compressor in the future.
According to one embodiment of the present invention, the monitoring module 10 may include a pressure sensor for acquiring a dynamic pressure signal at each state within the gas reciprocating compressor cylinder and a key phase sensor for acquiring a key phase signal at each state within the reciprocating compressor cylinder, which are connected to the reciprocating compressor 70.
Further, according to an embodiment of the present invention, the database module 40 is connected to the determining module 30, and provides the convolutional neural network to train the indicator diagram fault sample database of the hyper-parameters through simulation and emulation.
As shown in fig. 2 to 5, the method for diagnosing the fault of the ocean platform reciprocating compressor based on machine learning according to the embodiment of the present invention comprises the following steps:
step 1: obtaining real-time in-cylinder pressure and a crank angle;
step 2: an indicator diagram is generated using the dynamic pressure signal in combination with the key phase signal. Taking an indicator diagram curve in a stable running state as a normal indicator diagram curve, and combining the normal indicator diagram curve with an actual indicator diagram curve generated by real-time acquisition to display in the same indicator diagram;
and step 3: and generating an indicator diagram sample database. A part of indicator diagram samples can be generated according to theoretical simulation by using the structural size and the rated intake and exhaust pressure of the cylinder, and a part of real indicator diagram samples can be acquired according to actual compressor operation and experiments. Adding training labels to indicator diagram samples, wherein the indicator diagram samples in the same fault state correspond to the same label;
and 4, step 4: dividing the indicator diagram sample into a training set and a testing set, and sending the training set into a configured convolutional neural network for parameter training to finish the training of the network when certain recognition accuracy and stability are achieved;
and 5: and receiving the real indicator diagram collected in real time, inputting the actual indicator diagram into the trained convolutional neural network for outputting the classification result, and realizing the state identification and fault diagnosis of the reciprocating compressor.
According to one embodiment of the present invention, in step 1, a dynamic pressure signal and a key phase signal in each state in a cylinder are collected using a pressure sensor and a key phase sensor.
Step 1: and acquiring dynamic pressure signals and key phase signals in each state in the cylinder by using the pressure sensor and the key phase sensor.
Step 1.1: the key phase signal is combined to construct a volume signal corresponding to the cylinder pressure. According to the structural size of the cylinder, the method is represented by the formula:
the volume change of the cylinder corresponding to the pressure signal in one period of crankshaft rotation can be obtained, wherein V represents the corresponding volume, S represents the piston stroke, theta represents the crank angle, l represents the length of the connecting rod, B represents the inner diameter of the cylinder, and VclRepresenting clearance volume.
Step 1.2: the dynamic pressure and key phase signal under each state corresponds to a normal state and ten fault states, and the ten state faults comprise: leakage of the intake valve; leakage of the exhaust valve; piston ring leakage; the valve plate of the exhaust valve is tripped off from the valve seat; intake valve seizure/jamming; exhaust valve seizure/jamming; the clearance volume of the cylinder is overlarge; shaking the valve plate; the sectional area of the air valve or the air suction pipe is small; the cross section of the exhaust valve or the exhaust pipe channel is small. The indicator diagram for each fault condition has different characteristics, such as leakage fault of the air inlet valve, accelerated expansion process in the fault condition, retarded compression process, shortened exhaust process, increased air inlet pressure and decreased exhaust pressure. The corresponding graphical features are that compared with the normal state line, the slope of the expansion process line is increased, the slope of the compression process line is decreased, the intake pressure line is moved upwards, and the exhaust pressure line is decreased. Different faults correspond to different graph characteristics.
Step 2: and generating an indicator diagram by utilizing the dynamic pressure signal corresponding to the volume signal.
Step 2.1: a normalization operation is performed by dividing the dynamic pressure signal and the corresponding volume signal over a period by their maximum values.
While maintaining the relative characteristic change, so as to cope with the state of the variable condition.
Step 2.2: and drawing the normalized pressure signal and volume signal into indicator curves, taking the indicator curve in a stable running state as a normal indicator curve, and combining the normal indicator curve with an actual indicator curve generated by real-time acquisition in the same indicator diagram for display. The normal indicator diagram curve is represented by a green line, and the fault indicator diagram curve or the real-time indicator diagram curve is represented by a red line, so as to generate trainable indicator diagram samples.
Fig. 4 is a sample schematic diagram of an indicator diagram.
And step 3: and (3) processing the data through the steps 1 and 2 to generate indicator diagram samples, and generating an indicator diagram sample database. A part of indicator diagram samples can be generated by using the structural size and the rated intake and exhaust pressure of the cylinder according to theoretical simulation, namely, the indicator diagram samples in a normal state and a fault state are simulated by using some basic functions and random numbers according to the theoretical indicator diagram, the pressure change of an inlet and an outlet, the slope change of a compression and expansion process and the like. Under the condition of simulation or test conditions, a part of real indicator diagram samples are collected according to actual compressor operation and tests. And adding training labels to the indicator diagram samples, wherein the indicator diagram samples in the same fault state correspond to the same label.
And 4, step 4: dividing the indicator diagram sample into a training set and a testing set, and sending the training set into a configured convolutional neural network for parameter training to finish the training of the network when certain recognition accuracy and stability are achieved. The specific configuration of the convolutional network comprises the following steps:
step 4.1: through the normalized indicator diagram, the pressure ratios under different working conditions are different, so that the training set is subjected to data enhancement, namely, the image is subjected to up-down fixed amount of deviation, the number of samples is increased, and the generalization capability of the sample model is enhanced.
Step 4.2: TABLE 1 structural Table of convolutional network
The method comprises the steps of firstly sending an input picture into a convolution layer, wherein the convolution layer is mainly used for performing convolution operation on an input matrix through convolution cores to extract low-level and high-level features of input information, in a deep neural network, the convolution layer performs local perception on the image through a plurality of convolution cores when performing feature extraction on the image to extract local features, the features of the image extracted by different convolution cores are different, and finally the extracted local features are integrated to obtain the high-level features. The convolution kernel slides from left to right from top to bottom in the image, and when the convolution kernel slides a position, the convolution kernel carries out weighted summation on the pixel points in the image and the pixel points in the neighborhood of the pixel points so as to extract the image characteristics, form multi-level characteristic capture and enhance the learning and generalization capability of the whole model. The main features of the pooling layer are feature screening and feature invariance. And straightening all the indicator diagram data subjected to convolution and pooling to obtain one-dimensional data, and sending the one-dimensional data into the full-connection network. The fully connected layer processes the extracted features and plays a role of a classifier, and each node of the fully connected layer is connected with the node of the previous layer. The main role of Droupout is to randomly ignore some neurons and avoid overfitting of the data.
The bias term of the convolution kernel, the weight of the fully-connected neuron and the bias term need to be continuously trained and updated through back propagation. The loss function used is mean square error, which represents the gap between the network forward propagation reasoning result and the standard answer, and aims to find a group of parameters (w weight and b bias) so as to minimize the loss function, namely, the result accuracy is high. The mean square error is defined as follows:
wherein y isiIs the true value, y, of the ith data in a batchi' the error is expressed by a prediction value of the neural network.
The gradient of the loss function represents a vector obtained by partial derivation of each parameter by the loss function, and the direction in which the gradient of the loss function decreases is the direction in which the loss function decreases. The gradient descent method is to find the minimum value of the loss function along the descending direction of the gradient of the loss function, so as to obtain the optimal parameter, and the updating process of the gradient descent is back propagation. The gradient descent method involves the following formula:
wt+1*x+bt+1→y
lr represents the learning rate, and is a hyper-parameter, which characterizes the rate of gradient descent. If the learning rate setting is too small, the parameter update may be slow, and if the learning rate setting is too large, the parameter update may skip the minimum.
After the training set and the test set are circulated for one time, the parameters of the convolutional neural network are updated for one time, and finally the accuracy of the test set tends to be stable and meets the precision requirement.
And 5: and receiving the real indicator diagram collected in real time, inputting the actual indicator diagram into the trained convolutional neural network for outputting the classification result, and realizing the state identification and fault diagnosis of the reciprocating compressor.
Step 5.1: and storing the fault indicator diagram data and expanding the fault indicator diagram data into a fault database.
Step 5.2: and giving a fault processing pre-scheme according to the fault result of the convolutional neural network diagnosis.
Example one
Training and verification of classification and identification of the simulated indicator diagram. And (3) generating a part of indicator diagram samples according to theoretical simulation by using the structural size of the cylinder and the rated intake and exhaust pressure, namely simulating indicator diagram samples in a normal state and a fault state by using some basic functions and random numbers according to the theoretical indicator diagram, the pressure change of an inlet and an outlet, the slope change of a compression and expansion process and the like. And (3) changing different inlet and outlet pressure states, wherein the different inlet and outlet pressure states correspond to different pressure ratios, the fault characteristics of the indicator diagram curve are unchanged, the positions in the diagram are slightly different, and the test set contains the untrained pressure ratio condition of the training set so as to test the generalization capability of the model. The training and test set sample compositions are shown in table 2.
TABLE 2 simulation of training set and test set sample composition for indicator diagrams
After the corresponding labels are compiled into the samples, the samples are sent into the convolution network structure shown in the table 1 for training, the accuracy of the test set of the training is obtained after the training is finished, and after 8 times of training are repeatedly carried out, the accuracy of the test set is shown in the table 3. The 8 training results of the convolutional neural network without data enhancement are also shown in table 3.
TABLE 3 simulation indicator diagram test set accuracy
The training result shows that the data enhancement is used in the convolutional neural network aiming at the change of the variable working condition cylinder pressure ratio, so that the test accuracy is improved, and meanwhile, the effectiveness of the method in the aspects of classification and fault diagnosis of the reciprocating compressor indicator diagram is verified.
Example two
A first-stage air cylinder of a certain two-stage double-acting reciprocating compressor is selected for modeling, the actual working operation condition of the reciprocating compressor of the ocean platform is fitted, the rotating speed of the running compressor is high, Fluent software is used for simulation calculation of air valve faults, the simulated faults comprise faults of air inlet valve leakage, exhaust valve leakage and the like, and the model and the main process parameters are shown in a table 4. Fig. 5 is a sample example of a simulation generated intake valve malfunction indicator diagram. As can be seen from the example graph, the change trend of the slope of the expansion process and the compression process is the same for the change of the inlet and outlet pressure of the same fault, and the simulation through the function is closer to the result of simulating the fault.
Table 4 model and process main parameters.
| Main item | Index parameter |
| Rated intake pressure/MPa | 0.3 |
| Rated exhaust pressure/MPa | 0.85 |
| Compressor main machine speed/rpm | 1200 |
| Piston stroke/mm | 139.7 |
| Length of connecting rod/mm | 400 |
| Cylinder bore/mm | 244 |
| Clearance volume/mm3 | 263758.56 |
The method comprises the steps of simulating the leakage faults of the intake valve and the exhaust valve in different degrees respectively through simulation, generating a certain amount of fault indicator diagrams with different pressure ratios according to main parameters in a table 3 in a simulation mode to train and test a convolution network according to theory, adding one part of the simulation indicator diagrams into training concentrated training, successfully identifying the other part of the fault indicator diagrams by the trained convolution network, and enabling the test accuracy of the whole network to be more than 95%. The method comprises the steps of testing a fault diagnosis program of the reciprocating compressor based on the convolutional network by using a Fluent simulation fault indicator diagram, randomly inputting an unidentified indicator diagram, and finding that a trained convolutional network model can accurately identify and judge the fault type. In the fault identification and diagnosis of the reciprocating compressor, the trained model is directly called to output the diagnosis result, and only 0.07 second is needed from the input of the picture to the presentation of the diagnosis result.
The feasibility of the invention in the classification and fault diagnosis of the reciprocating compressor indicator diagram is verified.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
It can be understood that the invention develops the study of the reciprocating compressor fault diagnosis method based on machine learning based on the reciprocating compressor indicator diagram, proposes that the convolutional neural network is used for the classification and fault diagnosis of the reciprocating compressor indicator diagram, simultaneously places the theoretical indicator diagram curve and the measured indicator diagram in the same indicator diagram, and performs indicator diagram classification and reciprocating compressor fault identification by means of the feature extraction and self-learning capability of the convolutional neural network on the normalized indicator diagram. The ten fault states that can be identified by the invention include: leakage of the intake valve; leakage of the exhaust valve; leakage of the piston ring; the valve plate of the exhaust valve is tripped off from the valve seat; intake valve seizure/jamming; exhaust valve seizure/jamming; the clearance volume of the cylinder is too large; shaking the valve plate; the sectional area of the air valve or the air suction pipe is small; the cross section of the exhaust valve or the exhaust pipe channel is small.
In order to achieve the above purpose, the specific technical scheme of the method of the invention is as follows:
and acquiring dynamic pressure signals and key phase signals in each state in the cylinder by using the pressure sensor and the key phase sensor.
An indicator diagram is generated using the dynamic pressure signal in combination with the key phase signal. And taking the indicator diagram curve in the stable running state as a normal indicator diagram curve, and combining the normal indicator diagram curve with an actual indicator diagram curve generated by real-time acquisition to display in the same indicator diagram.
And generating an indicator diagram sample database. A part of indicator diagram samples can be generated according to theoretical simulation by using the structural size of the cylinder and the rated intake and exhaust pressure, and then a part of real indicator diagram samples are acquired according to actual compressor operation and experiments. And adding training labels to the indicator diagram samples, wherein the indicator diagram samples in the same fault state correspond to the same label.
Dividing the indicator diagram sample into a training set and a testing set, and sending the training set into a configured convolutional neural network for parameter training to finish the training of the network when certain recognition accuracy and stability are achieved.
And receiving the real indicator diagram collected in real time, inputting the actual indicator diagram into the trained convolutional neural network for outputting the classification result, and realizing the state identification and fault diagnosis of the reciprocating compressor.
Further, the invention also provides a fault diagnosis device for the reciprocating compressor of the ocean platform, which is characterized by comprising the following components:
a monitoring module: the system mainly comprises a dynamic pressure sensor and a key phase sensor, and is used for monitoring the pressure in an air cylinder in real time and obtaining a crank angle;
a processing module: and generating a indicator diagram by using the data of the monitoring module. Taking the indicator diagram curve in a stable running state as a normal indicator diagram curve, and combining the normal indicator diagram curve with an actual indicator diagram curve generated by real-time acquisition to display in the same indicator diagram;
a database module: the method comprises the steps that an indicator diagram fault sample database for a convolutional neural network to train hyper-parameters is provided through simulation, emulation and the like;
a judging module: the main body is a convolutional neural network and is used for processing indicator diagram samples provided by the module and giving fault diagnosis results of the reciprocating compressor;
the early warning module: sending out a fault early warning and processing solution according to the judging module;
a recording module: and storing and recording the fault indicator diagram for expanding the database module.
The monitoring module, the processing module, the judging module and the early warning module are sequentially linked, and the judging module is additionally connected with the database module and the recording module.
The invention also provides a computer readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the computer program can execute all the procedures of the method for classifying and diagnosing the fault of the indicator diagram of the ocean platform reciprocating compressor based on the convolutional neural network.
The method of the invention has the following characteristics and advantages:
the method comprises the steps of carrying out classification recognition on indicator diagrams by using a convolutional neural network which is stronger in the field of image recognition, carrying out feature self-extraction by using the convolutional neural network, reducing training parameters, ensuring accuracy, shortening training time and reducing a lightweight model, and simultaneously realizing multi-classification tasks of the states of the reciprocating compressor by using a softmax function in the last layer of the network.
The normal state indicator curve and the fault indicator curve are placed in one graph for comparison and identification, the convolution network receives two kinds of information simultaneously, the curve comparison characteristic is enhanced, the identification accuracy is high, and the method is suitable for fault diagnosis under the changing working condition.
The indicator diagram is simply normalized, namely:
the fault identification of the reciprocating compressor under a variable working condition state to a certain degree is realized by combining a data enhancement method while the slope of the compression and expansion processes is not changed.
The fault database used for convolutional neural network training can be obtained by theoretical function simulation, numerical simulation calculation and experimental tests, regardless of the reciprocating compressor object applied. And fault data are not easy to obtain, the database is expanded to a certain extent through theoretical function simulation and simulation, and the reliability of related data is ensured.
Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present application, and are used for illustrating the technical solutions of the present application, but not limiting the same, and the protection scope of the present invention is not limited thereto, although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope disclosed in the present application; such modifications, changes or substitutions do not depart from the spirit and scope of the present disclosure, and should be construed as being included therein. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples described in this specification can be combined and combined by those skilled in the art.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (8)
1. A fault diagnosis device for an ocean platform reciprocating compressor based on machine learning is characterized by comprising:
the monitoring module is connected with the reciprocating compressor to monitor the pressure in the cylinder in real time and obtain a crank angle;
the processing module is connected with the monitoring module, generates an actual indicator diagram according to the data of the monitoring module, and combines the actual indicator diagram and a normal indicator diagram curve in a stable running state in the same indicator diagram sample for display;
the judging module is connected with the processing module and comprises a convolutional neural network, and the convolutional neural network provides a fault diagnosis result of the reciprocating compressor according to the indicator diagram sample provided by the processing module;
the database module is connected with the judging module and is an indicator diagram fault sample database for convolutional neural network training hyper-parameters;
the recording module is connected with the judging module and the database module and is used for storing and recording the fault indicator diagram and expanding the database module;
and the early warning module is connected with the judging module and used for displaying the result of the judging module to give early warning and providing a corresponding health management scheme.
2. The machine learning based ocean platform reciprocating compressor fault diagnosis apparatus according to claim 1 wherein said monitoring module comprises a pressure sensor connected to said reciprocating compressor for collecting dynamic pressure signals at each state within said reciprocating compressor cylinder and a key phase sensor for collecting key phase signals at each state within said reciprocating compressor cylinder.
3. The machine learning-based marine platform reciprocating compressor fault diagnosis device of claim 1, wherein the database module is connected with the judgment module, and part of the sample database can be obtained through theoretical simulation and simulation calculation for the convolutional neural network to train hyper-parameters.
4. A fault diagnosis method for an ocean platform reciprocating compressor based on machine learning is characterized by comprising the following steps:
step 1: obtaining real-time in-cylinder pressure and a crank angle;
and 2, step: the dynamic pressure signal is combined with the key phase signal to generate an indicator diagram, an indicator diagram curve in a stable running state is taken as a normal indicator diagram curve, and the normal indicator diagram curve and an actual indicator diagram curve generated by real-time acquisition are combined in the same indicator diagram to be displayed, so that the contrast is enhanced;
and step 3: the method comprises the steps of generating an indicator diagram sample database, generating a part of indicator diagram samples according to theoretical simulation by using the structural size and the rated intake and exhaust pressure of a cylinder, and acquiring a part of real indicator diagram samples according to actual compressor operation and tests. Adding training labels to indicator diagram samples, wherein the indicator diagram samples in the same fault state correspond to the same label;
and 4, step 4: dividing the indicator diagram sample into a training set and a testing set, and sending the training set into a configured convolutional neural network for parameter training to finish the training of the network when certain recognition accuracy and stability are achieved;
and 5: and receiving the real indicator diagram collected in real time, inputting the actual indicator diagram into the trained convolutional neural network for outputting the classification result, and realizing the state identification and fault diagnosis of the reciprocating compressor.
5. The machine learning based ocean platform reciprocating compressor fault diagnosis method according to claim 4,
in step 1, a pressure sensor and a key phase sensor are used to acquire dynamic pressure signals and key phase signals in each state in the cylinder.
6. The method for diagnosing the fault of the reciprocating compressor of the ocean platform based on the machine learning as claimed in claim 4, wherein the specific steps of the step 2 are as follows:
step 2.1: a normalization operation is performed by dividing the dynamic pressure signal and the corresponding volume signal over a period by their maximum values.
While keeping the relative characteristic change, so as to deal with the state of the variable working condition;
step 2.2: and drawing the normalized pressure signal and volume signal into a indicator curve, taking the indicator curve in a stable running state as a normal indicator curve, combining the normal indicator curve with an actual indicator curve generated by real-time acquisition, displaying the indicator curve and the actual indicator curve in the same indicator graph, wherein the normal indicator curve is represented by a green line, the fault indicator curve or the real-time indicator curve is represented by a red line, and adding a label to generate a trainable indicator graph sample.
7. The method for diagnosing the fault of the reciprocating compressor of the ocean platform based on the machine learning as claimed in claim 4, wherein in the step 3, the training fault database of the convolutional neural network can be composed of a theoretical function simulation indicator diagram, a simulation software simulation indicator diagram and a fault test actual measurement indicator diagram, and a fault training database with a certain scale is formed.
8. The machine learning based ocean platform reciprocating compressor fault diagnosis method according to claim 4, wherein the specific steps of enhancing the training samples using data in the step 4 are as follows:
step 4.1: through the normalized indicator diagram, the pressure ratios under different working conditions are different, so that the training set is subjected to data enhancement, namely, the image is subjected to up-down fixed amount of deviation, the number of samples is increased, and the generalization capability of the sample model is enhanced.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111217412.7A CN114741942A (en) | 2021-10-19 | 2021-10-19 | Fault diagnosis device and method for ocean platform reciprocating compressor based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111217412.7A CN114741942A (en) | 2021-10-19 | 2021-10-19 | Fault diagnosis device and method for ocean platform reciprocating compressor based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114741942A true CN114741942A (en) | 2022-07-12 |
Family
ID=82274147
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111217412.7A Pending CN114741942A (en) | 2021-10-19 | 2021-10-19 | Fault diagnosis device and method for ocean platform reciprocating compressor based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114741942A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115355162A (en) * | 2022-08-08 | 2022-11-18 | 西安交通大学 | Diaphragm compressor diaphragm fault diagnosis method based on oil pressure |
| CN115596654A (en) * | 2022-09-21 | 2023-01-13 | 西安交通大学（Cn） | Reciprocating compressor fault diagnosis method and system based on state parameter learning |
-
2021
- 2021-10-19 CN CN202111217412.7A patent/CN114741942A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115355162A (en) * | 2022-08-08 | 2022-11-18 | 西安交通大学 | Diaphragm compressor diaphragm fault diagnosis method based on oil pressure |
| CN115355162B (en) * | 2022-08-08 | 2023-12-19 | 西安交通大学 | Diaphragm fault diagnosis method of diaphragm compressor based on oil pressure |
| CN115596654A (en) * | 2022-09-21 | 2023-01-13 | 西安交通大学（Cn） | Reciprocating compressor fault diagnosis method and system based on state parameter learning |
| CN115596654B (en) * | 2022-09-21 | 2023-12-22 | 西安交通大学 | Reciprocating compressor fault diagnosis method and system based on state parameter learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110399905B (en) | Method for detecting and describing wearing condition of safety helmet in construction scene | |
| CN114741942A (en) | Fault diagnosis device and method for ocean platform reciprocating compressor based on machine learning | |
| CN104712542B (en) | A kind of reciprocating compressor sensitive features based on Internet of Things are extracted and method for diagnosing faults | |
| CN108257121B (en) | Method, apparatus, storage medium and the terminal device that product defects detection model updates | |
| Wang et al. | Research on the fault monitoring method of marine diesel engines based on the manifold learning and isolation forest | |
| CN110363337B (en) | Oil measuring method and system of oil pumping unit based on data driving | |
| CN109782603A (en) | The detection method and monitoring system of rotating machinery coupling fault | |
| CN108830285A (en) | A kind of object detection method of the reinforcement study based on Faster-RCNN | |
| CN108932479A (en) | A kind of human body anomaly detection method | |
| CN108665005A (en) | A method of it is improved based on CNN image recognition performances using DCGAN | |
| CN107492098A (en) | It is a kind of based on PCA and CNN high-temperature forging surface defect in position detecting method | |
| CN109948733A (en) | More classification methods, sorter and the storage medium of alimentary tract endoscope image | |
| CN111161224A (en) | Casting internal defect grading evaluation system and method based on deep learning | |
| CN112308825B (en) | SqueezeNet-based crop leaf disease identification method | |
| CN112001110A (en) | Structural damage identification monitoring method based on vibration signal space real-time recursive graph convolutional neural network | |
| CN112950570B (en) | Crack detection method combining deep learning and dense continuous central point | |
| CN111415353A (en) | Detection structure and detection method for fastener burr defects based on ResNet58 network | |
| CN112766283B (en) | Two-phase flow pattern identification method based on multi-scale convolution network | |
| CN111428655A (en) | Scalp detection method based on deep learning | |
| CN109978868A (en) | Toy appearance quality determining method and its relevant device | |
| CN117611599B (en) | Blood vessel segmentation method and system integrating centre line diagram and contrast enhancement network | |
| CN117516939A (en) | Bearing cross-working condition fault detection method and system based on improved EfficientNetV2 | |
| Heryanto et al. | Classification of Coffee Beans Defect Using Mask Region-based Convolutional Neural Network | |
| CN116720079A (en) | Wind driven generator fault mode identification method and system based on multi-feature fusion | |
| CN116385717A (en) | Foliar disease identification method, foliar disease identification device, electronic equipment, storage medium and product |
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 |