Vacuum equipment health diagnosis method based on deep learning and electronic equipment

Abstract

The invention relates to a vacuum equipment health diagnosis method based on deep learning and electronic equipment. The method comprises the following steps: acquiring historical data and real-time data of key operation parameters of target vacuum equipment; the obtained data is preprocessed; constructing a detection model based on stacked LSTM; inputting the preprocessed data into the trained detection model for prediction, and calculating to obtain a threshold value; and inputting the preprocessed real-time data into the trained detection model, outputting a predicted value, and performing vacuum equipment health diagnosis based on a threshold value. Compared with the prior art, the method has the advantages that the real-time diagnosis of the vacuum equipment is realized, the non-planned downtime is reduced, and the operation efficiency and reliability of the equipment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of vacuum equipment health diagnosis, and in particular to a vacuum equipment health diagnosis method and electronic equipment based on deep learning. Background Art

[0002] Equipment that simulates vacuum plays a vital role in the aerospace field. These devices are used to simulate the vacuum environment of outer space, providing indispensable experimental conditions for ground testing of spacecraft and other related equipment. The vacuum environment simulator can effectively simulate the low-pressure environment of outer space, so as to test the pressure resistance of spacecraft, the vacuum resistance of materials, and the performance of electronic equipment. The traditional monitoring system collects key parameters such as pressure, temperature, and current in real time through sensors, and triggers an alarm when these parameters exceed the preset range to remind the operator to check and handle. Although the above method can ensure the normal operation of the vacuum environment simulator to a certain extent, the traditional monitoring system can only remind the operator to handle it through an alarm after the fault occurs, and cannot predict potential faults in advance. Therefore, when the operator finds the abnormality, the equipment may have been in a faulty state for a long time, and even cause unplanned shutdown of the equipment, seriously affecting the normal experiment and test process.

[0003] The existing health diagnosis method of vacuum equipment only relies on monitoring whether certain parameters exceed the characteristic value threshold, judging based on the operator's experience, and is limited by the fact that the operator cannot always stare at the screen, so it is impossible to perform health diagnosis in a timely and effective manner.

[0004] How to achieve real-time health diagnosis of vacuum equipment has become a technical problem that needs to be solved. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a vacuum equipment health diagnosis method and electronic equipment based on deep learning.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] According to one aspect of the present invention, a vacuum equipment health diagnosis method based on deep learning is provided, the method comprising the following steps:

[0008] Step S101: Acquire historical data and real-time data of key operating parameters of target vacuum equipment;

[0009] Step S102: pre-processing the data acquired in step S101;

[0010] Step S103: constructing a detection model based on stacked LSTM;

[0011] Step S104: input the data pre-processed in step S102 into the trained detection model for prediction, and calculate the threshold value;

[0012] Step S105: Input the real-time data pre-processed in step S102 into the trained detection model, output the predicted value, and perform health diagnosis of the vacuum equipment based on the threshold value of step S104.

[0013] Preferably, the target vacuum equipment key operating parameters include: vacuum degree, pump frequency, pump current and valve switching, wherein the vacuum degree is the target parameter.

[0014] Preferably, the historical data is a combination of data from several test cycles, each test cycle being composed of several stages;

[0015] The real-time data is the real-time data generated at a certain stage of the current test cycle.

[0016] Preferably, the preprocessing of the historical data includes: dividing the historical data into several periods and several stages by a time series segmentation algorithm, clustering the data of the same stage in different periods by clustering analysis, retaining the category with the most data and its corresponding data, enhancing the data of the same stage in different periods by a combination method, logarithmically normalizing the enhanced data, and then serializing the data;

[0017] The real-time data preprocessing includes: logarithmic normalization and serialization.

[0018] More preferably, the cluster analysis dynamic time warping method calculates the distance between two time series, uses hierarchical clustering to divide the data of the same stage into two categories, and retains the category with more data.

[0019] Preferably, the stacked LSTM model includes three LSTM layers and one fully connected layer.

[0020] Preferably, the training of the detection model adopts a dynamically adjusted learning rate. If the verification loss does not decrease for m consecutive rounds, the current learning rate is multiplied by a set coefficient. The detection model adopts an early stopping strategy to interrupt training. If the verification loss does not improve within n rounds, the training is stopped.

[0021] Preferably, the process of step S104 includes:

[0022] The preprocessed data in step S102 is input into the trained detection model for prediction to obtain the error at each time point, and the error at each time point is calculated using the 3σ criterion to obtain a threshold value; wherein the threshold value is obtained by calculating the fitting error, specifically: the initial training is divided into the same stage data of each cycle, and the trained detection model is fitted to the same stage data of different cycles, and the fitting error is equal to the difference between the predicted value and the true value, and then the error set error at the same time uses the 3σ principle to calculate the threshold threshold = mean (error) + 3 * std (error), wherein mean (error) is the mean value of the error set, and std (error) is the standard deviation of the error set.

[0023] Preferably, the step S105 is specifically as follows: inputting the real-time data preprocessed in step S102 into the trained detection model to obtain the predicted value at the next moment, calculating the error between the predicted value and the true value, comparing the error with the threshold obtained in the corresponding time step S3, and if the error is greater than the threshold, issuing a warning to remind the staff that the vacuum equipment has an abnormality.

[0024] According to another aspect of the present invention, an electronic device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the method described above is implemented when the processor executes the program.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The present invention obtains historical data and real-time data of key operating parameters of vacuum equipment, performs preprocessing on them respectively, inputs the preprocessed data into a stacked LSTM model for training, predicts the training data by the trained model to obtain the error at each time point, calculates a threshold value according to the error, finally inputs the processed real-time data into the model to obtain a predicted value at the next moment, calculates the error between the predicted value and the true value, compares the error with the threshold value of the corresponding time, realizes real-time diagnosis of the vacuum equipment, and promptly notifies maintenance when an abnormality occurs, thereby greatly reducing the unplanned downtime of the equipment, reducing maintenance costs, and improving the overall operation efficiency and reliability of the vacuum equipment.

[0027] The present invention performs sequence segmentation, cluster analysis, data enhancement, normalization and serialization on historical data, and normalization and serialization on real-time data. These preprocessing methods accelerate the convergence speed of the detection model and improve the efficiency of real-time diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a flow chart of the vacuum equipment health diagnosis method in the present invention. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0030] Example 1

[0031] This embodiment relates to a vacuum equipment health diagnosis method based on deep learning, comprising the following steps:

[0032] Step S101: Acquire historical data and real-time data of key operating parameters of target vacuum equipment.

[0033] The key operating parameters of the target vacuum equipment include: vacuum degree, pump frequency, pump current, valve switching and other parameters, among which vacuum degree is the target parameter and the time interval is 1 minute.

[0034] Historical data is a combination of data from several test cycles, each of which is composed of several stages. Real-time data refers to real-time data generated at a certain stage of the current test cycle.

[0035] Step S102: pre-processing the data acquired in step S101.

[0036] The data obtained in the above step S101 is preprocessed, and the historical data is divided into several periods and several stages by a time series segmentation algorithm. The data of the same stage in different periods are clustered by cluster analysis, and the category with the most data and its corresponding data are retained. The data of the same stage in different periods are enhanced by a combination method, and the enhanced data is logarithmically normalized, and then the data is serialized. Real-time data only needs logarithmic normalization and serialization;

[0037] The time series segmentation algorithm divides the historical test data into several periods, and then divides each period into several stages of time series data, and integrates the data of the same stage in different periods.

[0038] Cluster analysis uses the Dynamic Time Warping (DTW) algorithm to calculate the distance between two time series, and uses hierarchical clustering to divide the data in the same stage into two categories, retaining the category with more data for the next step.

[0039] The logarithmic normalization expression is x norm =log 10 (x) / log 10 (x max ), where x is the original data, xmax The maximum value in the data set is used to serialize the normalized data into a format suitable for LSTM input, that is, a three-dimensional array (number of samples, time steps, feature dimensions).

[0040] Step S103: construct a detection model.

[0041] The detection model uses stacked LSTM as a model, and the data processed in step S102 is input into the model for training. The stacked LSTM model is specifically three LSTM layers and one fully connected layer. The input data time step is 20, the feature dimension is 1, and the number of neurons of the stacked LSTM is 128, 64, and 32 respectively. Because the output is the value of the next moment, the number of neurons in the fully connected layer is 1. The learning rate is dynamically adjusted during the model training process, and the initial learning rate is 0.001. If the verification loss is not reduced for m consecutive rounds (m is set to 3 in this embodiment), the current learning rate is multiplied by the set coefficient (the coefficient is set to 0.1 in this embodiment), and the early stopping strategy is adopted. If the verification loss is no longer improved within n (n is set to 6 in this embodiment) rounds, the training is interrupted, and the number of training rounds epoch=100. The model prediction step size is consistent with the training step size.

[0042] Step S104: Setting a threshold.

[0043] According to the model trained in step S103, the data processed in step S102 is predicted to obtain the error at each time point, and the error at each time point is calculated using the 3σ criterion to obtain a threshold.

[0044] The threshold is obtained by calculating the fitting error, specifically: the initial training is divided into the same stage data of each cycle, the trained model is fitted to the same stage data of different cycles, the fitting error is equal to the difference between the predicted value and the true value, and then the error set error at the same time uses the 3σ principle to calculate the threshold threshold = mean (error) + 3 * std (error), where mean (error) is the mean of the error set, and std (error) is the standard deviation of the error set.

[0045] Step S105: Detect whether an abnormality occurs.

[0046] Real-time monitoring, input the pre-processed real-time data in step S102 into the model trained in step S103 to obtain the predicted value at the next moment. Calculate the error between the predicted value and the true value, and compare the error with the threshold obtained in step S3 at the corresponding time. If the error is greater than the threshold, a warning is issued. If 5 warnings are issued in succession, it reminds the staff that the vacuum equipment has an abnormal situation and needs to be checked; if the error is less than or equal to the threshold, it is judged that the vacuum equipment is operating normally.

[0047] Previous technologies were limited to setting data limits or relying on manual experience to judge, that is, when the data exceeds the preset limit or is manually found to be abnormal. However, the present invention will predict whether the value at the next moment conforms to historical rules based on the model. Compared with previous technologies, the present invention can detect anomalies earlier based on real-time prediction values, thereby reducing unplanned downtime of equipment.

[0048] Example 2

[0049] The electronic device of the present invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0050] Multiple components in the device are connected to the I / O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as disks, optical disks, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunication networks.

[0051] The processing unit performs the various methods and processes described above, such as methods S101 to S105. For example, in some embodiments, methods S101 to S105 may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of methods S101 to S105 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S101 to S105 in any other appropriate manner (e.g., by means of firmware).

[0052] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0053] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0054] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0055] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A vacuum equipment health diagnosis method based on deep learning, characterized in that: The method comprises the following steps: Step S101: Acquire historical data and real-time data of key operating parameters of target vacuum equipment; Step S102: pre-processing the data acquired in step S101; Step S103: constructing a detection model based on stacked LSTM; Step S104: input the data pre-processed in step S102 into the trained detection model for prediction, and calculate the threshold value; Step S105: Input the real-time data pre-processed in step S102 into the trained detection model, output the predicted value, and perform health diagnosis of the vacuum equipment based on the threshold value of step S104.

2. A vacuum equipment health diagnosis method based on deep learning according to claim 1, characterized in that: The key operating parameters of the target vacuum equipment include: vacuum degree, pump frequency, pump current and valve switching, among which vacuum degree is the target parameter.

3. A vacuum equipment health diagnosis method based on deep learning according to claim 1, characterized in that: The historical data is a combination of data from several test cycles, each of which is composed of several stages; The real-time data is the real-time data generated at a certain stage of the current test cycle.

4. A vacuum equipment health diagnosis method based on deep learning according to claim 1, characterized in that: The preprocessing of the historical data includes: dividing the historical data into several periods and several stages by a time series segmentation algorithm, clustering the data of the same stage in different periods by clustering analysis, retaining the category with the most data and its corresponding data, enhancing the data of the same stage in different periods by a combination method, logarithmically normalizing the enhanced data, and then serializing the data; The real-time data preprocessing includes: logarithmic normalization and serialization.

5. A vacuum equipment health diagnosis method based on deep learning according to claim 4, characterized in that: The cluster analysis dynamic time warping method calculates the distance between two time series, uses hierarchical clustering to divide the data in the same stage into two categories, and retains the category with more data.

6. A vacuum equipment health diagnosis method based on deep learning according to claim 1, characterized in that: The stacked LSTM model includes three LSTM layers and one fully connected layer.

7. A vacuum equipment health diagnosis method based on deep learning according to claim 1, characterized in that: The training of the detection model adopts a dynamically adjusted learning rate. If the verification loss does not decrease for m consecutive rounds, the current learning rate is multiplied by the set coefficient. The detection model adopts an early stopping strategy to interrupt training. If the verification loss does not improve within n rounds, the training is stopped.

8. A vacuum equipment health diagnosis method based on deep learning according to claim 1, characterized in that: The process of step S104 includes: The preprocessed data in step S102 is input into the trained detection model for prediction to obtain the error at each time point, and the error at each time point is calculated using the 3σ criterion to obtain a threshold value; wherein the threshold value is obtained by calculating the fitting error, specifically: the initial training is divided into the same stage data of each cycle, and the trained detection model is fitted to the same stage data of different cycles, and the fitting error is equal to the difference between the predicted value and the true value, and then the error set error at the same time uses the 3σ principle to calculate the threshold threshold = mean (error) + 3 * std (error), wherein mean (error) is the mean value of the error set, and std (error) is the standard deviation of the error set.

9. A vacuum equipment health diagnosis method based on deep learning according to claim 1, characterized in that: The step S105 is specifically as follows: the real-time data pre-processed in step S102 is input into the trained detection model to obtain the predicted value at the next moment, the error between the predicted value and the true value is calculated, and the error is compared with the threshold obtained in the corresponding time step S3. If the error is greater than the threshold, a warning is issued to remind the staff that the vacuum equipment has an abnormal situation.

10. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to any one of claims 1 to 9 is implemented.

Images