CN112069738B - Electric steering engine residual life prediction method and system based on DBN and multilayer fuzzy LSTM - Google Patents
Electric steering engine residual life prediction method and system based on DBN and multilayer fuzzy LSTM Download PDFInfo
- Publication number
- CN112069738B CN112069738B CN202010972888.0A CN202010972888A CN112069738B CN 112069738 B CN112069738 B CN 112069738B CN 202010972888 A CN202010972888 A CN 202010972888A CN 112069738 B CN112069738 B CN 112069738B
- Authority
- CN
- China
- Prior art keywords
- steering engine
- electric steering
- residual life
- lstm
- time
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000012544 monitoring process Methods 0.000 claims abstract description 44
- 230000015556 catabolic process Effects 0.000 claims abstract description 21
- 238000006731 degradation reaction Methods 0.000 claims abstract description 21
- 238000007781 pre-processing Methods 0.000 claims abstract description 16
- 238000013135 deep learning Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 21
- 230000015654 memory Effects 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 13
- 230000005540 biological transmission Effects 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 238000010187 selection method Methods 0.000 claims description 3
- 238000013519 translation Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 abstract description 4
- 210000004027 cell Anatomy 0.000 description 13
- 238000004891 communication Methods 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000013136 deep learning model Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 230000007787 long-term memory Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 238000004804 winding Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 229910000831 Steel Inorganic materials 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
-
- 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
- G06F2119/04—Ageing analysis or optimisation against ageing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Geometry (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Steering Control In Accordance With Driving Conditions (AREA)
Abstract
The invention discloses a method and a system for predicting the residual life of an electric steering engine based on DBN and multilayer fuzzy LSTM, wherein the method comprises the following steps: acquiring real-time monitoring data of an electric steering engine; preprocessing the acquired real-time monitoring data; inputting the preprocessed data into a trained steering engine state degradation model, and outputting predicted residual life of the electric steering engine; the steering engine state degradation model extracts a characteristic rule of the preprocessed data through a deep confidence network, and then extracts time characteristics in a data sequence through a multi-layer fuzzy LSTM network; and obtaining the predicted residual life of the electric steering engine based on the characteristic rule and the time characteristic. The invention adopts the deep learning network model based on DBN and multilayer fuzzy LSTM to predict the residual life of the electric steering engine, can effectively extract the characteristic rule and the time characteristic of the sequence in the monitoring data of the multidimensional electric steering engine sensor, and improves the accuracy of the residual life prediction; the safety and reliability of the steering engine during operation are improved.
Description
Technical Field
The invention relates to the technical field of equipment residual life prediction, in particular to an electric steering engine residual life prediction method and system based on a DBN and a multilayer fuzzy LSTM.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The electric steering engine is used as core equipment of an advanced aircraft self-driving system such as a helicopter, an unmanned aerial vehicle and the like, and is an important control component in the aircraft. If the electric steering engine breaks down, particularly the main control steering engine breaks down, such as a rudder, an elevator, an aileron and the like, the aircraft can be in a runaway state, and the catastrophic results of the aircraft destruction and death can be caused seriously, so that the safety and reliability of an aircraft avionics system, a helicopter and an unmanned plane are seriously restricted. The electric steering engine has poor reliability in a plurality of airborne systems and is easy to fail. Therefore, fault monitoring and life prediction of the electric steering engine are of great significance to the flight safety of the aircraft.
Compared with a hydraulic steering engine and a pneumatic steering engine, the electric steering engine formed by the servo direct current motor and the transmission mechanism has the advantages of high precision, convenience in maintenance, small size, light weight, easiness in transmission control and the like, and is widely applied to advanced aircrafts such as unmanned aircrafts, airplanes, helicopters and spacecrafts. According to statistics of reliability of non-electronic parts, for small and medium-sized permanent magnet direct current motors used in aircrafts such as helicopters, unmanned planes and the like, brush faults account for more than 34% of motor fault probability, rotor winding faults account for about 20%, magnetic steel faults account for about 20%, and bearing faults account for 15%. Therefore, the faults of the electric steering engine mainly occur at the positions of the motor, the transmission mechanism and the position sensor, and the rotor winding, the permanent magnet, the electric brush and the rotating shaft are weak links of reliability of the steering engine.
The prior art generally adopts a post diagnosis or off-line monitoring method to monitor the fault of the steering engine, and the mode has low efficiency and can not ensure the safety and reliability of the steering engine during operation.
Disclosure of Invention
In order to solve the problems, the invention provides a method and a system for predicting the residual life of an electric steering engine based on a DBN and a multilayer fuzzy LSTM, which can monitor the running state of the steering engine in real time and predict the residual life of the steering engine on line.
In some embodiments, the following technical scheme is adopted:
an electric steering engine residual life prediction method based on DBN and multilayer fuzzy LSTM comprises the following steps:
acquiring real-time monitoring data of an electric steering engine;
preprocessing the acquired real-time monitoring data;
Inputting the preprocessed data into a trained steering engine state degradation model, and outputting predicted residual life of the electric steering engine;
Extracting a characteristic rule of the preprocessed data by the steering engine state degradation model through a deep confidence network, reducing characteristic dimension of the data, and extracting time characteristics in a data sequence through a multi-layer fuzzy LSTM network; and obtaining the predicted residual life of the electric steering engine based on the characteristic rule and the time characteristic.
In other embodiments, the following technical solutions are adopted:
an electric steering engine residual life prediction system based on a DBN and a multilayer fuzzy LSTM, comprising:
The data acquisition module is used for acquiring real-time monitoring data of the electric steering engine;
The data preprocessing module is used for preprocessing the acquired real-time monitoring data;
the residual life prediction module is used for inputting the preprocessed data into the trained steering engine state degradation model and outputting the predicted residual life of the electric steering engine;
Extracting a characteristic rule of the preprocessed data by the steering engine state degradation model through a deep confidence network, reducing characteristic dimension of the data, and extracting time characteristics in a data sequence through a multi-layer fuzzy LSTM network; and obtaining the predicted residual life of the electric steering engine based on the characteristic rule and the time characteristic.
In other embodiments, the following technical solutions are adopted:
the terminal equipment comprises a server, wherein the server comprises a memory, a processor and a computer program which is stored in the memory and can run on the processor, and the processor realizes the residual life prediction method of the electric steering engine based on the DBN and the multilayer fuzzy LSTM when executing the program.
Compared with the prior art, the invention has the beneficial effects that:
(1) The invention adopts the deep learning network model based on DBN and multilayer fuzzy LSTM to predict the residual life of the electric steering engine, can effectively extract the characteristic rule and the time characteristic of the sequence in the monitoring data of the multidimensional electric steering engine sensor, and improves the accuracy of the residual life prediction; the safety and reliability of the steering engine during operation are improved.
(2) According to the method, before the real-time monitoring data are input into the steering engine state degradation model, the missing values are filled, and the model training effect and the prediction accuracy are improved. Meanwhile, the variable with larger relevance is determined by calculating the correlation coefficient between the variables, so that the data volume is further compressed, the information of the input data is more compact, and the calculated amount of training of the deep learning model is reduced.
(3) The invention can ensure that the real-time state of each aircraft is observed and reasonably arrange the flight and maintenance tasks based on the signal transmission of the Beidou short message communication function.
Drawings
FIG. 1 is a flow chart of a method for predicting residual life of an electric steering engine based on a DBN and a multi-layer fuzzy LSTM in an embodiment of the invention;
fig. 2 is a schematic diagram of LSTM cells spread out in time series in an embodiment of the invention.
Detailed Description
It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the application. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present application. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
Example 1
In one or more embodiments, a method for predicting remaining life of an electric steering engine based on a DBN and a multi-layer fuzzy LSTM is disclosed, referring to fig. 1, comprising the steps of:
Step (1): the layout sensor acquires real-time monitoring data of the electric steering engine;
Major faults of the electric steering engine include transmission mechanism faults, motor faults and sensor faults. The current, rotation speed and vibration signals are easily available and contain a lot of steering engine status information. Thus, two current sensors are installed, monitoring the currents of motor 1 and motor 2; four vibration sensors are installed to monitor vibration signals of the motor 1, the motor 2, the transmission mechanism and the shell; three rotation speed sensors are installed to monitor the rotation speeds of the motor 1, the motor 2 and the output shaft; four temperature sensors are installed to monitor the temperatures of the motor 1, the motor 2, the transmission mechanism and the housing.
Step (2): preprocessing the acquired real-time monitoring data;
Specifically, the raw sensor monitoring data of the electric steering engine collected from the sensor monitoring system is a multi-dimensional time series, and the sensor monitoring data of each electric steering engine is denoted as X i＝[x1,x2,...,xt,...xT. Wherein T represents the maximum running time step number of the equipment, x t is n-dimensional sensor monitoring data at the moment T, i represents the current steering engine id, and steering engine data of all ids form the whole data set.
The acquired sensor monitoring data typically suffers from missing values, which can lead to discontinuities in the time frame. Moreover, the sensor monitoring system has multidimensional sensor data, and redundant information will reduce the accuracy of life prediction. Factors such as interference of external environment, load change, sensor noise and the like can influence the training effect of the model. The embodiment of the invention adopts the following method to preprocess the data.
1) Missing value handling
Sensor monitoring data often loses value due to abnormal data transmission and abnormal sensors of the electric steering engine. The abnormal data is directly input into the deep learning model, which will result in poor training effect of the model. The invention adopts local mean substitution to fill the missing value, and the specific calculation can be expressed as:
where x m is the data value of the missing point, and k is the number of missing values used for mean value replacement before and after the missing point, the present embodiment sets k to 3 to smooth the curve.
2) Sensor variable selection
From the foregoing, the collected monitoring data of the electric steering engine includes various variables, such as temperature, rotation speed, vibration, current, etc., if the state correlation between the variables and the electric steering engine is small, the fitting will be insufficient, and the model training effect will be affected. In order to screen out variables which can represent or greatly influence the state of the electric steering engine, the data size is further compressed, the information of the input data is more compact, and the calculated amount of training of the deep learning model is reduced. The present embodiment uses Pearson product-moment correlation coefficient product coefficients to calculate the correlation between each variable, and selects the valuable variables through correlation analysis to determine the final input of the model.
The pearson product moment correlation coefficient is expressed as follows:
where n is the sample size, x i and y i are the individual sample points indexed by i, And/>Is the sample mean. The pearson product moment correlation coefficient is used to measure the degree of correlation between variables, and if the correlation coefficient of a certain variable with other variables is small, this means that the variable has greater randomness, the variable should be discarded.
3) Data normalization
In order to eliminate the influence of the difference between the dimensions and the value ranges of the indexes, the standardization processing is needed, and the data is scaled according to the proportion so as to fall into a specific area, thereby being convenient for comprehensive analysis. The invention adopts zero-mean normalization (z-score standardization) to process data, and the formula is as follows:
Wherein the method comprises the steps of The mean value of the original data, sigma is the standard deviation of the original data. The mean value of the processed data is 0, and the standard deviation is 1.
Step (3): and (3) constructing a steering engine state degradation model based on the deep confidence network and the multi-layer fuzzy LSTM network, acquiring data of the history sensor of the electric steering engine, processing the data by adopting the preprocessing method in the step (2), and generating a training sample in a translation time window mode, wherein the residual life corresponding to the last moment in the time window is used as a label of the sample.
In order to meet the input requirement of the subsequent deep learning network model, the embodiment generates training samples by adopting a method of translating a time window, wherein the data input to the model each time is a two-dimensional tensor with the size of num_time_num_sensor, the num_sensor is the sensor dimension selected by the sensor variable selection method, the num_time is the size of the time window, the value of the embodiment is 30, and each time slides forwards by one time step.
The maximum time step acquired by each steering engine is T i, and then T i +1-num_time training samples can be generated.
And training the constructed steering engine state degradation model through the training sample, and outputting the residual service life of the electric steering engine.
The state characteristics of the electric steering engine are contained in multidimensional data, and the deep learning network needs to learn the complex characteristics. The sensors arranged at different positions of the electric steering engine acquire different monitoring signals, and sensor variables at different positions can be mutually influenced with sensor variables at other positions, so that inherent spatial characteristics exist between the sensors.
The steering engine state degradation model constructed in this embodiment adopts a deep-learning Deep Belief Network (DBN) to extract deep spatial features, the DBN is formed by stacking a plurality of Restricted Boltzmann Machines (RBMs), the RBMs are composed of a visual layer and a hidden layer, the number of neurons of the first visual layer is determined by the input data dimension, the weight value between any two connected neurons of the visual layer and the hidden layer is W, b and c are bias values of the visual layer and the hidden layer respectively, v is an input vector, h is an output vector, and the energy function of one RBM is expressed as follows:
When one RBM completes, its hidden layer will serve as the visual layer for the next RBM. A DBN is formed by a series connection of several RBMs. Each layer of the RBM is trained in advance as an initial weight, and then the back propagation error of the output layer fine-tunes the parameters of each layer of the network. The method adopts a DBN model with a two-layer RBM structure. The DBN uses a nonlinear structure to extract characteristics and map data from a high-dimensional space to a low-dimensional space, so that characteristic information of original data is reserved to the greatest extent, a characteristic rule which is closer to the original essence of the data is obtained, and the dimension of the data is reduced. The method adopts the ReLU function as the nonlinear activation function of the hidden layer, and when the input value is greater than 0, the derivative value is 1, so that the gradient dispersion problem can be effectively overcome.
After the characteristics of the input sensor data are extracted, the time characteristics in the data sequence are extracted through the multi-layer fuzzy LSTM network, and compared with the RNN circulating neural network, the long-term and short-term memory neural network can effectively solve the problems of gradient disappearance and gradient explosion. The LSTM cells are shown in FIG. 2.
In fig. 2, x t denotes an input at the current time, C t denotes a cell state (long-term memory cell), h t denotes a hidden layer state (short-term memory cell), f t denotes a forgetting gate unit, o t denotes an output gate unit, i t denotes an input gate unit, and σ denotes an activation function of each gate. The mathematical expression is as follows:
ft＝σ(Wfxt+W'fht-1+bf)
Candidate values representing cell states, deciding what values are to be updated through an input gate; a new candidate vector is then created by the tanh layer, which vector is added to the candidate variables.
it＝σ(Wixt+W'iht-1+bi)
The cell state before modulation to the current cell state updates the state of the cell.
The final output is determined by the cell state value and the result of the output gate together:
ot＝σ(Woxt+W'oht-1+bo)
ht＝ot*tanh(Ct)
Where W f、W'f、Wi、W'i、Wo、W'o、WC、W'C represents a weight and b f、bi、bo、bC represents a deviation.
The blurring of the blurred LSTM is performed on different layers and cells of the original LSTM model M, the concept of which is to derive the blurring weights of the LSTM pre-trained model M by usingThe stress operator or function generates a variation model M'. These stress functions have different stress intensities ρ defined and verified on the Weierstrass function. Use/>Variants of stress operators or with different ρ stress intensity sum/>The function of the stress operator stresses the weights of the LSTM cells.
Will beThe stress operator is applied to the model M as a whole, affecting the weights of the LSTM units and layers of the RUL prediction deep learning network. I (input gate), o (output gate), f (forget gate) and/>, in LSTM model M(Component node) apply/>The mathematical expression of stress, ρ stress intensity, affecting the per-layer LSTM cell weight (W f、W'f、Wi、W'i、Wo、W'o、WC、W'C) is as follows:
In the embodiment, a 4-layer LSTM network superposition mode is adopted to extract the deep abstract features, the output of the former layer is used as the input of the latter layer, and the nonlinear fitting capacity of the model is improved. The state at time t of the nth layer is expressed as follows:
In the method, in the process of the invention, The hidden layer state at the time of layer t of n-1 is shown.
And a full connection layer is connected behind the LSTM layer, final output information is extracted from the characteristic rule, and the final output is expressed as:
Wherein W q and b q are weights and deviations, and the full-connection layer finally outputs the predicted residual life of the electric steering engine.
Network parameters are set through adaptive moment estimation (Adam optimization algorithm), and in order to prevent the network from being fitted in the training process, dropout is added in the training process to inactivate a part of neurons randomly.
Step (4): inputting the preprocessed real-time monitoring data into a trained steering engine state degradation model, and outputting predicted residual life of the electric steering engine;
As an optional implementation manner, after the predicted remaining life of the electric steering engine is obtained, the remaining life of the electric steering engine is displayed in real time, and the real-time state of the electric steering engine and the predicted value of the remaining life are sent to the monitoring platform.
Example two
In one or more embodiments, an electric steering engine residual life prediction system based on a DBN and a multilayer fuzzy LSTM is disclosed, comprising:
The data acquisition module is used for acquiring real-time monitoring data of the electric steering engine;
The data preprocessing module is used for preprocessing the acquired real-time monitoring data;
the residual life prediction module is used for inputting the preprocessed data into the trained steering engine state degradation model and outputting the predicted residual life of the electric steering engine;
Extracting a characteristic rule of the preprocessed data by the steering engine state degradation model through a deep confidence network, reducing characteristic dimension of the data, and extracting time characteristics in a data sequence through a multi-layer fuzzy LSTM network; and obtaining the predicted residual life of the electric steering engine based on the characteristic rule and the time characteristic.
As an alternative embodiment, further comprising:
the display module is used for displaying the state of the electric steering engine and the predicted value of the residual life in real time;
And the data transmission module is used for sending the real-time state and the residual life prediction value of the electric steering engine to the remote monitoring platform. Specifically, the real-time state of the electric steering engine is sent to the ground monitoring and dispatching platform through the wireless communication module based on the Beidou short message communication function, the collected original data is not sent, and only the predicted real-time state and the predicted residual life value of each electric steering engine are sent. The Beidou short message communication function can be used for areas where the communication base station in an extreme environment or disaster area is destroyed and common mobile communication signals cannot cover.
It should be noted that, the specific implementation manner of the above modules is implemented by the manner disclosed in the first embodiment, which is not repeated.
In addition, the modules can be integrated on a controller and used for realizing the prediction of the residual life of the steering engine; the system of the embodiment combines the embedded Linux technology and the Jetson TX embedded platform to manufacture a residual life prediction UI interface, and integrates the modules to run on the TX2 embedded GPU platform.
Example III
In one or more embodiments, a terminal device is disclosed, including a server including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the method for predicting remaining life of an electric steering engine based on a DBN and a multilayer fuzzy LSTM in embodiment one when executing the program. For brevity, the description is omitted here.
It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general purpose processors, digital signal processors DSP, application specific integrated circuits ASIC, off-the-shelf programmable gate array FPGA or other programmable logic device, discrete gate or transistor logic devices, discrete hardware components, or the like.
A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The memory may include read only memory and random access memory and provide instructions and data to the processor, and a portion of the memory may also include non-volatile random access memory. For example, the memory may also store information of the device type.
In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in a processor or by instructions in the form of software.
The method for predicting the residual life of the electric steering engine based on the DBN and the multilayer fuzzy LSTM in the first embodiment may be directly embodied as a hardware processor for execution, or may be executed by a combination of hardware and software modules in the processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in a memory, and the processor reads the information in the memory and, in combination with its hardware, performs the steps of the above method. To avoid repetition, a detailed description is not provided herein.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.
Claims (9)
1. The method for predicting the residual life of the electric steering engine based on the DBN and the multilayer fuzzy LSTM is characterized by comprising the following steps of:
acquiring real-time monitoring data of an electric steering engine;
preprocessing the acquired real-time monitoring data;
Inputting the preprocessed data into a trained steering engine state degradation model, and outputting predicted residual life of the electric steering engine;
Extracting a characteristic rule of the preprocessed data by the steering engine state degradation model through a DBN deep confidence network, reducing characteristic dimension of the data, and extracting time characteristics in a data sequence through a multi-layer fuzzy LSTM network; based on the characteristic rule and the time characteristic, obtaining the predicted residual life of the electric steering engine;
The training process of the steering engine state degradation model comprises the following steps: acquiring historical monitoring data of an electric steering engine, and constructing a training data set; preprocessing the acquired historical monitoring data; the training sample is generated by adopting a translation time window method, and the specific method comprises the following steps: the data of each input model is a two-dimensional tensor with the size of num_time; num_sensor is the sensor dimension selected by the sensor variable selection method, num_time is the size of a time window, the value is 30, each time the sample slides forwards by one time step, and the residual life corresponding to the last moment in the time window is used as the label of the sample; inputting the training sample into a steering engine state degradation model to train and outputting the predicted residual life of the electric steering engine;
The multi-layer fuzzy LSTM network comprises the following specific methods:
Will be The stress operator is applied to the model M as a whole, thereby affecting the weights of the LSTM units and layers of the RUL predictive deep learning network, applying/>, on the input gate, output gate, forget gate and component nodes of the LSTM model MStress, ρ stress intensity affects the LSTM cell weight of each layer, and the mathematical expression is as follows:
Wherein f t denotes a forgetting gating unit, i t an input gating unit, Candidate values representing cell states, o t representing output gating cells, σ representing the activation function of each gate, W f、Wf'、Wi、Wi'、Wo、Wo'、WC、WC' representing weights, h t representing hidden layer states, b f、bi、bo、bC representing deviations, x t representing the input at the current time;
the multi-layer fuzzy LSTM network is connected with a full-connection layer at the back, and the full-connection layer outputs the predicted residual service life of the electric steering engine finally.
2. The method for predicting the residual life of an electric steering engine based on a DBN and a multi-layer fuzzy LSTM as claimed in claim 1, wherein said real-time monitoring data at least comprises: current signals and rotating speed signals of all motors in the electric steering engine, vibration signals and temperature signals of all motors and transmission shafts.
3. The method for predicting the residual life of the electric steering engine based on the DBN and the multilayer fuzzy LSTM as set forth in claim 1, wherein the preprocessing of the acquired real-time monitoring data is carried out, and the method specifically comprises the following steps:
and calculating the average value of the numerical values of the set number before and after the data loss point, and filling the missing value of the data loss point by adopting the average value.
4. The method for predicting the residual life of an electric steering engine based on a DBN and a multilayer fuzzy LSTM as claimed in claim 3, wherein the preprocessing of the acquired real-time monitoring data further comprises:
and calculating the correlation between each variable in the monitoring data by adopting the pearson moment correlation coefficient, and selecting the variable with large correlation coefficient as the variable of the input deep learning neural network.
5. The method for predicting the residual life of an electric steering engine based on a DBN and a multilayer fuzzy LSTM as claimed in claim 3, wherein the preprocessing of the acquired real-time monitoring data further comprises:
and (5) carrying out normalization processing on the data by adopting zero-mean normalization.
6. The method for predicting the residual life of the electric steering engine based on the DBN and the multilayer fuzzy LSTM according to claim 1, wherein after the predicted residual life of the electric steering engine is obtained, the residual life of the electric steering engine is displayed in real time, and the real-time state and the residual life predicted value of the electric steering engine are sent to a monitoring platform.
7. An electric steering engine residual life prediction system based on a DBN and a multilayer fuzzy LSTM, which is characterized by comprising:
The data acquisition module is used for acquiring real-time monitoring data of the electric steering engine;
The data preprocessing module is used for preprocessing the acquired real-time monitoring data;
the residual life prediction module is used for inputting the preprocessed data into the trained steering engine state degradation model and outputting the predicted residual life of the electric steering engine;
Extracting a characteristic rule of the preprocessed data by the steering engine state degradation model through a DBN deep confidence network, reducing characteristic dimension of the data, and extracting time characteristics in a data sequence through a multi-layer fuzzy LSTM network; based on the characteristic rule and the time characteristic, obtaining the predicted residual life of the electric steering engine;
The training process of the steering engine state degradation model comprises the following steps: acquiring historical monitoring data of an electric steering engine, and constructing a training data set; preprocessing the acquired historical monitoring data; the training sample is generated by adopting a translation time window method, and the specific method comprises the following steps: the data of each input model is a two-dimensional tensor with the size of num_time; num_sensor is the sensor dimension selected by the sensor variable selection method, num_time is the size of a time window, the value is 30, each time the sample slides forwards by one time step, and the residual life corresponding to the last moment in the time window is used as the label of the sample; inputting the training sample into a steering engine state degradation model to train and outputting the predicted residual life of the electric steering engine;
The multi-layer fuzzy LSTM network comprises the following specific methods:
Will be The stress operator is applied to the model M as a whole, thereby affecting the weights of the LSTM units and layers of the RUL predictive deep learning network, applying/>, on the input gate, output gate, forget gate and component nodes of the LSTM model MStress, ρ stress intensity affects the LSTM cell weight of each layer, and the mathematical expression is as follows:
Wherein f t denotes a forgetting gating unit, i t an input gating unit, Candidate values representing cell states, o t representing output gating cells, σ representing the activation function of each gate, W f、Wf'、Wi、Wi'、Wo、Wo'、WC、WC' representing weights, h t representing hidden layer states, b f、bi、bo、bC representing deviations, x t representing the input at the current time;
the multi-layer fuzzy LSTM network is connected with a full-connection layer at the back, and the full-connection layer outputs the predicted residual service life of the electric steering engine finally.
8. The DBN and multi-layer fuzzy LSTM based electric steering engine residual life prediction system of claim 7, further comprising:
the display module is used for displaying the state of the electric steering engine and the predicted value of the residual life in real time;
and the data transmission module is used for sending the real-time state and the residual life prediction value of the electric steering engine to the remote monitoring platform.
9. A terminal device comprising a server comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the DBN and multilayer fuzzy LSTM based electric steering engine residual life prediction method according to any one of claims 1-6 when executing the program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010972888.0A CN112069738B (en) | 2020-09-16 | 2020-09-16 | Electric steering engine residual life prediction method and system based on DBN and multilayer fuzzy LSTM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010972888.0A CN112069738B (en) | 2020-09-16 | 2020-09-16 | Electric steering engine residual life prediction method and system based on DBN and multilayer fuzzy LSTM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112069738A CN112069738A (en) | 2020-12-11 |
| CN112069738B true CN112069738B (en) | 2024-04-26 |
Family
ID=73696004
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010972888.0A Active CN112069738B (en) | 2020-09-16 | 2020-09-16 | Electric steering engine residual life prediction method and system based on DBN and multilayer fuzzy LSTM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112069738B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114218872B (en) * | 2021-12-28 | 2023-03-24 | 浙江大学 | DBN-LSTM semi-supervised joint model-based residual service life prediction method |
| CN117031956B (en) * | 2023-08-23 | 2024-03-19 | 无锡纬途流体科技有限公司 | Control method and system of intelligent embedded electric actuator |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109102066A (en) * | 2018-07-12 | 2018-12-28 | 山东师范大学 | Online time series data prediction technique, system and storage medium based on fuzzy reasoning |
| CN110738360A (en) * | 2019-09-27 | 2020-01-31 | 华中科技大学 | equipment residual life prediction method and system |
| KR20200023672A (en) * | 2018-08-14 | 2020-03-06 | 오토시맨틱스 주식회사 | Diagnosis method of battery using Deep Learning |
| CN111274737A (en) * | 2020-02-25 | 2020-06-12 | 山东大学 | Method and system for predicting remaining service life of mechanical equipment |
| CN111460728A (en) * | 2020-03-09 | 2020-07-28 | 华南理工大学 | Method and device for predicting residual life of industrial equipment, storage medium and equipment |
-
2020
- 2020-09-16 CN CN202010972888.0A patent/CN112069738B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109102066A (en) * | 2018-07-12 | 2018-12-28 | 山东师范大学 | Online time series data prediction technique, system and storage medium based on fuzzy reasoning |
| KR20200023672A (en) * | 2018-08-14 | 2020-03-06 | 오토시맨틱스 주식회사 | Diagnosis method of battery using Deep Learning |
| CN110738360A (en) * | 2019-09-27 | 2020-01-31 | 华中科技大学 | equipment residual life prediction method and system |
| CN111274737A (en) * | 2020-02-25 | 2020-06-12 | 山东大学 | Method and system for predicting remaining service life of mechanical equipment |
| CN111460728A (en) * | 2020-03-09 | 2020-07-28 | 华南理工大学 | Method and device for predicting residual life of industrial equipment, storage medium and equipment |
Non-Patent Citations (1)
| Title |
|---|
| Research on failure prediction using DBN and LSTM neural network;Gu Yuhai;2018 57th annual conference of Society of instrument and control engineers of Jspan;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112069738A (en) | 2020-12-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110216680B (en) | Cloud-ground cooperative fault diagnosis system and method for service robot | |
| CN112069738B (en) | Electric steering engine residual life prediction method and system based on DBN and multilayer fuzzy LSTM | |
| US20210082292A1 (en) | Unsupervised anomaly detection for autonomous vehicles | |
| CN112817296B (en) | Flight fault prediction method and device based on flight data and deep learning | |
| Li et al. | A Siamese hybrid neural network framework for few-shot fault diagnosis of fixed-wing unmanned aerial vehicles | |
| Ning et al. | Applications of deep learning in big data analytics for aircraft complex system anomaly detection | |
| Ghazali et al. | Vibration-based fault detection in drone using artificial intelligence | |
| Nasser et al. | A hybrid of convolutional neural network and long short-term memory network approach to predictive maintenance | |
| Li et al. | Civil aircraft health management research based on big data and deep learning technologies | |
| CN115510950A (en) | Aircraft telemetry data anomaly detection method and system based on time convolution network | |
| CN112560997A (en) | Fault recognition model training method, fault recognition method and related device | |
| Veshneva et al. | Increasing the safety of flights with the use of mathematical model based on status functions | |
| Yu-ming et al. | Research on intelligent manufacturing flexible production line system based on digital twin | |
| Yuan et al. | An underwater thruster fault diagnosis simulator and thrust calculation method based on fault clustering | |
| Lang et al. | Data augmentation for fault prediction of aircraft engine with generative adversarial networks | |
| CN112560252A (en) | Prediction method for residual life of aircraft engine | |
| Verstraete et al. | A deep adversarial approach based on multi-sensor fusion for remaining useful life prognostics | |
| CN114595525A (en) | Rolling bearing degradation trend prediction method based on linear regression and TCN | |
| Shin et al. | Causality-Seq2Seq model for battery anomaly detection | |
| Xu et al. | ISHM-oriented adaptive fault diagnostics for avionics based on a distributed intelligent agent system | |
| Yang et al. | Aircraft actuator fault diagnosis using deep learning based sparse representation and TSM | |
| Baptista et al. | Aircraft on-condition reliability assessment based on data-intensive analytics | |
| Zhang et al. | Simulation-to-reality UAV Fault Diagnosis with Deep Learning | |
| Darrah et al. | On-board health-state awareness to detect degradation in multirotor systems | |
| Thanaraj et al. | Hardware-in-the-loop simulation for quadrotor fault diagnosis enhancing airworthiness using OS-Fuzzy-ELM |
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 |