CN114878162B - Ship bearing lubrication state on-line monitoring system based on deep learning - Google Patents
Ship bearing lubrication state on-line monitoring system based on deep learning Download PDFInfo
- Publication number
- CN114878162B CN114878162B CN202210538087.2A CN202210538087A CN114878162B CN 114878162 B CN114878162 B CN 114878162B CN 202210538087 A CN202210538087 A CN 202210538087A CN 114878162 B CN114878162 B CN 114878162B
- Authority
- CN
- China
- Prior art keywords
- lubrication
- oil film
- bearing
- lubricating oil
- oil
- 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
- 238000005461 lubrication Methods 0.000 title claims abstract description 113
- 238000013135 deep learning Methods 0.000 title claims abstract description 18
- 238000012544 monitoring process Methods 0.000 title claims abstract description 16
- 239000003921 oil Substances 0.000 claims abstract description 104
- 239000010687 lubricating oil Substances 0.000 claims abstract description 60
- 238000013528 artificial neural network Methods 0.000 claims abstract description 22
- 238000005299 abrasion Methods 0.000 claims abstract description 13
- 230000000694 effects Effects 0.000 claims abstract description 13
- 238000005457 optimization Methods 0.000 claims abstract description 13
- 239000003795 chemical substances by application Substances 0.000 claims abstract description 4
- 238000012545 processing Methods 0.000 claims description 26
- 230000006870 function Effects 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 12
- 230000007423 decrease Effects 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 10
- 238000012549 training Methods 0.000 claims description 10
- 238000003062 neural network model Methods 0.000 claims description 7
- 230000017525 heat dissipation Effects 0.000 claims description 6
- 239000000463 material Substances 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000009434 installation Methods 0.000 claims description 4
- 230000002035 prolonged effect Effects 0.000 claims description 4
- 230000009471 action Effects 0.000 claims description 3
- 238000009530 blood pressure measurement Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 claims description 3
- 238000003745 diagnosis Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 230000008719 thickening Effects 0.000 claims description 2
- 238000010438 heat treatment Methods 0.000 abstract description 2
- 238000000034 method Methods 0.000 abstract description 2
- 230000008569 process Effects 0.000 abstract description 2
- 238000001816 cooling Methods 0.000 abstract 1
- 238000000691 measurement method Methods 0.000 abstract 1
- 230000001050 lubricating effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 239000003208 petroleum Substances 0.000 description 2
- 239000004215 Carbon black (E152) Substances 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 229930195733 hydrocarbon Natural products 0.000 description 1
- 150000002430 hydrocarbons Chemical class 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000005184 irreversible process Methods 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
Abstract
The invention discloses a deep learning-based on-line monitoring system for the lubrication state of a ship bearing, wherein the running state of a ship propulsion shafting directly determines whether the ship sails stably or not, and the poor lubrication state can aggravate the abrasion and heating of a supporting bearing, so that the whole shafting is caused to vibrate severely, the running safety of the ship is seriously affected, and serious consequences can be caused. Based on deep learning, the invention builds an LSTM-ELM neural network in consideration of the sequential influence relation of each data set on time sequence, adopts WOA algorithm to perform parameter optimization, and adopts the data acquisition aspect including oil film pressure data measured by a non-direct contact measurement method and rotational speed, temperature and viscosity data measured by a sensor to train the network through the data sets so as to achieve the effect of predicting the occurrence probability of boundary lubrication. When boundary lubrication occurs, the ship staff is warned in time to make the ship staff respond correspondingly, such as adding lubricating oil, antiwear agent or cooling properly to the bearing. The invention can obviously improve the service life of the bearing and effectively improve the stability and safety of the ship in the sailing process.
Description
Technical Field
The invention relates to the field of oil film lubrication of ship bearings, in particular to an online monitoring system for the lubrication state of a ship bearing based on deep learning.
Background
Marine transportation is one of the main logistics networks of current international trade, along with economic globalization, trade frequently goes on and off among countries, a ship is required to have higher stability in the long-distance transportation process, and whether the running state of a supporting bearing of a ship propulsion shafting directly influences the sailing stability of the ship, if a complete lubricating film cannot be formed or the lubricating oil film is thinned, the complete lubricating state is transited to a boundary lubricating state until dry friction occurs, so that abrasion and heating of the bearing are aggravated, severe vibration of the whole propulsion shafting is caused, and the stable running state of the whole ship body is influenced. Because the working environment of lubricating oil is special, the parameters of an oil film are difficult to directly measure, the lubrication states of all points of the oil film are different, and the lubrication states are related to the parameters such as oil film temperature, pressure, viscosity, bearing rotating speed and the like, so that the lubrication state of the oil film of the bearing is difficult to judge, and the lubrication failure is difficult to continuously monitor on line, therefore, the lubrication state of the bearing is monitored in real time, the abrasion of the bearing is avoided, and the bearing can work stably for a long time.
Aiming at the problem of difficult real-time monitoring, the invention combines the deep learning principle, fits a prediction model of the oil film lubrication state through experimental data, and realizes the function of predicting the oil film lubrication state by transmitting various parameters of the bearing acquired by the sensor into a neural network.
Disclosure of Invention
The invention aims to provide a ship bearing lubrication state online monitoring system based on deep learning so as to solve the technical problems.
The invention adopts the following technical scheme for realizing the purposes:
An on-line monitoring system for the lubrication state of a ship bearing based on deep learning comprises the following steps:
s1, acquiring various data during bearing operation through various sensors arranged near a bearing; the data acquisition equipment comprises an ultrasonic pressure measuring device, an oil film temperature sensor, a lubricating oil viscosity sensor, a journal rotating speed sensor, an analog-to-digital converter and an upper computer;
s2, building an LSTM-ELM neural network prediction model, searching an ELM optimal weight by adopting a WOA optimization function, processing acquired signals in real time by a threshold value, and outputting the occurrence probability of boundary lubrication;
S3, substituting parameters of oil temperature, lubricating oil viscosity, oil film pressure and rotating speed of a shaft into the trained neural network model; an LSTM-ELM neural network prediction model is built, collected parameters enter an ELM learning machine and an LSTM neural network at the same time, on one side of the ELM, compared with a traditional BP neural network, a WOA optimization algorithm is adopted, an optimal weight is found, a threshold value is adopted, L2 regularization is adopted to prevent overfitting, for a network structure, two hidden layers are adopted, the first hidden layer is a fully-connected layer of 10 neurons, an activation function is converged and calculated faster Relu times, the operation efficiency is guaranteed, the response rate of an alarm is effectively improved, the second hidden layer is a fully-connected layer of 5 neurons, the activation function is softmax, probability distribution is output, the output layer is three neurons, the occurrence probability of three lubrication states is respectively represented, the parameters are initialized by the ELM, the threshold value randomly generated by training is taken as an initial position vector of the WOA;
S4, at the LSTM side, 3 sigmoid activation functions and 2 tanh activation functions are adopted to realize the input, forgetting, updating and output of information, the number of stacked layers is 1, training is stopped when the training times reach 2000 times, and the dimension of the output result is3, so that the occurrence probability of three lubrication states is respectively represented;
s5, substituting the optimal weight found by the WOA optimizing algorithm into an ELM extreme learning machine, taking the LSTM input layer and all hidden layers as ELM input, carrying out average value processing on ELM output and LSTM neural network output, and calculating to obtain probability distribution of a lubrication state;
S6, substituting the data acquired by the processed sensor into the trained neural network model, and calculating probability distribution of an oil film lubrication state in real time, wherein the lubrication state comprises full film lubrication, boundary lubrication and mixed lubrication, when the oil film lubrication state is mainly full film lubrication, the lubrication effect is optimal, bearing abrasion basically cannot occur, the protection effect is achieved on the surfaces of the bearings and the journals, so that the state attribute lamps are green, when the oil film lubrication state is mainly mixed lubrication, although partial boundary lubrication occurs, the state attribute lamps are in an acceptable range, the state attribute lamps can be yellow, when the oil film lubrication state is mainly boundary lubrication, the bearings can generate more severe friction abrasion, the state attribute lamps can be red, and an alarm is sent to a turbine main control room;
s7, when the occurrence probability of boundary lubrication is larger than a preset probability threshold value, sending out a warning signal to a ship main control room; when the probability of oil film boundary lubrication is larger than 0.9 at a certain moment, the oil film of the bearing is basically in boundary lubrication, the oil film temperature is kept in a higher zone, serious abrasion occurs, and at the moment, by starting an alarm, a crew member adds lubricating oil, antiwear agent or properly cools, boundary lubrication is effectively avoided, the service life of the bearing is prolonged, meanwhile, a probability threshold value is properly adjusted according to actual conditions, and the probability threshold value is 0.8-0.9, so that the alarm is not too frequent, and boundary lubrication is effectively prevented.
As a further scheme of the invention, the journal rotation speed sensor in the step S1 adopts a non-contact photoelectric sensor which is arranged on one side of the bearing shell, the rotation speed sensor is arranged on the upper side of the rotating shaft, and the electro-optic sensing piece is arranged on the rotating shaft; the oil film temperature sensor and the lubricating oil viscosity sensor are arranged on the inner wall of the oil outlet; the ultrasonic pressure measuring device comprises an ultrasonic transmitting module, an ultrasonic receiving module and a data processing module, wherein the ultrasonic transmitting module and the ultrasonic receiving module are symmetrically arranged on the upper side and the lower side of the bearing, and the data processing module comprises a DSP and a control circuit.
As a further scheme of the invention, the step S1 is to measure the thicknesses d1 and d3 of the bearings at the upper and lower sides, the height d between the bearings at the upper and lower sides and the diameter d2 of the rotating shaft in the installation direction of the ultrasonic transmitting module and the receiving module before ultrasonic pressure measurement is performed; the speed of ultrasonic wave in the bearing material and the rotating shaft material can be obtained by table lookup and is set as v 1,v2.
The hydraulic pressure and the sound velocity have approximate linear relation, the linear relation is stable, and the relation formula is satisfied:
Wherein, C is the ultrasonic sound velocity in the lubricating oil, and the unit is m/s; c 0 is the ultrasonic sound velocity in lubricating oil at normal temperature and one atmosphere pressure, and the unit is m/s; p is the pressure of the lubricating oil, and the unit is Pa; k is a proportionality coefficient, wherein C 0 and K are constants, and the proportionality coefficient is obtained through experiments or table lookup; as can be seen from the formula (1), the pressure value of the pipeline can be obtained by measuring the sound velocity of ultrasonic waves in the bearing lubricating oil film;
The data processing module and the ultrasonic transmitting module transmit ultrasonic waves, the ultrasonic receiving module receives the ultrasonic waves, the total running time t 0 of the ultrasonic waves is measured, and the expression of C is obtained:
the lubricating oil pressure P expression is obtained by simplifying the combined formulas (1) and (2):
the formula (3) shows that the oil pressure of lubricating oil between bearings, namely the oil film pressure, can be obtained only by obtaining the total running time t 0 of ultrasonic waves between bearings, the data processing module is used for carrying out the calculation and converting the pressure signal into a digital signal to be sent to an upper computer for subsequent processing;
The electric signals measured by the sensor are converted into digital signals through an analog-to-digital converter, the digital signals are transmitted to an upper computer for data processing, and the analog-to-digital converter is connected to the oil film temperature sensor, the lubricating oil viscosity sensor and the journal rotation speed sensor and is connected to the upper computer by adopting RS485 communication.
As a further aspect of the present invention, the occurrence probability of the boundary lubrication of step S2 is described using the following equation:
P=f(T)+g(μ)+h(n)+j(py);
Wherein P represents the occurrence probability of boundary lubrication, T represents the temperature of the lubricating oil during the operation of the shaft, mu represents the viscosity of the lubricating oil, n represents the rotating speed of the shaft, and P y represents the oil film pressure;
For lubricating oil temperature T: the increase of the oil temperature can lead to the decrease of the viscosity of the lubricating oil, so that the local oil film is easy to damage, the lubrication fails, the bearing capacity of the bearing is reduced, and even the lubricating oil is carbonized to burn tiles; the oil temperature is reduced to increase the viscosity of the oil, thereby increasing the lubrication friction of the oil film, increasing the power consumption of the bearing, thickening the oil film, and generating machine vibration caused by the vibration of the oil film;
For viscosity μ: the viscosity of the lubricating oil at different temperatures is different, and when the viscosity is too high, the resistance is large, so that uniform lubrication is not facilitated, and the heat dissipation effect is poor; when the viscosity is too low, the axial surface cannot keep enough oil film, and the lubrication effect is poor;
for rotational speed n: the higher the rotating speed is, the higher the temperature of the bearing and the lubricating oil is, and the viscosity of the lubricating oil is reduced under the action of centrifugal force;
Oil film pressure p y: the larger the load, the larger the oil film pressure and the smaller the oil film thickness; the smaller the load is, the smaller the oil film pressure is, the larger the oil film thickness is, the heat dissipation of the bearing is affected, and the oil film pressure mainly depends on the load;
the adopted deep learning prediction model automatically fits the functional relation among the parameters according to the previous data.
As a further scheme of the invention, step S3 sets the population number N as 20 and the maximum iteration number T max as 40 for the WOA optimization algorithm, and the specific steps of the WOA optimization algorithm are as follows:
D=|CX*(t)-X(t)| (4);
X(t+1)=X*(t)-AD (5);
A=2ar1-a (6);
C=2r2 (7);
D*=|X*(t)-X(t)| (10);
Wherein T is the current iteration number, X (t) is the current solution position, X *(t) is the random optimal solution position, D is the distance between the two, r 1,r2 is a random number in (0, 1), a linearly decreases from 2 to 0, T max is the maximum iteration number, A and C are coefficients, D * is the distance from the optimal solution, b is a constant for measuring the spiral shape, l is a random number in (0, 1), p is the probability, and ELM diagnosis error rate is lower than 0.01%, and the current individual fitness value is calculated in the initialized population to obtain the optimal solution; along with the decrease of the value a, the value A fluctuates along with the decrease, the WOA is set to iterate by adopting the formula (9) when the value A is less than or equal to 1, iterate by adopting the formula (5) when the value A is more than 1, calculate the individual fitness value, find the optimal parameter and substitute the optimal parameter into the ELM.
Compared with the prior art, the invention has the following advantages: 1. the invention designs a detection system for the lubrication state of an oil film, combines a deep learning theory and provides a feasible technology for realizing the real-time monitoring of the lubrication state.
2. The invention provides an oil film lubrication state monitoring system mainly used for a ship propeller shaft bearing, which comprises non-contact measurement of oil film pressure, acquisition of oil film temperature, viscosity and rotating speed, real-time detection of the lubrication state of the ship bearing and effective improvement of the safety of shafting operation.
3. According to the invention, LSTM is selected to match ELM for predicting the lubrication state of the bearing, oil film temperature, viscosity and pressure physical quantity generated during bearing operation are sequentially related on respective time axes, data at the later moment is influenced by data at the previous moment, LSTM is selected for analyzing related time sequences, and meanwhile ELM is selected for solving the problem of inaccurate prediction possibly caused by incorrect forgetting information of LSTM. The best weight of the ELM is found by using a WOA optimization algorithm, and the threshold value can improve the performance degradation of the ELM caused by instability during training.
3. When the boundary lubrication of the bearing happens, an alarm is sent out, so that a crew is prompted to respond correspondingly, and the service life of the bearing can be remarkably prolonged.
4. Compared with the traditional regular maintenance mode, the invention can play an obvious role in prevention. Because the abrasion of the bearing is an irreversible process, the human intervention is carried out before the abrasion, and the replacement cost of the bearing can be greatly reduced.
Drawings
FIG. 1 represents the sensor distribution location of the system of the present invention.
Fig. 2 represents the physical quantity to be measured by the ultrasonic pressure measuring device of the present invention.
Fig. 3 represents a program flow chart of the present invention.
Fig. 4 represents a flow chart of the neural network in the present invention.
Detailed Description
The invention is described in further detail below with reference to the drawings and the specific examples.
A ship bearing lubrication state on-line monitoring system based on deep learning comprises the following steps: s1, acquiring various data during bearing operation through various sensors arranged near a bearing; the main equipment for collecting data comprises, but is not limited to, an ultrasonic pressure measuring device, an oil film temperature sensor, a lubricating oil viscosity sensor, a journal rotating speed sensor, an analog-to-digital converter and an upper computer.
As shown in FIG. 1, the journal rotation speed sensor adopts a non-contact photoelectric sensor, and is arranged on one side of a bearing shell, the position of the rotation speed sensor is shown as 1, and the installation position of an electro-optic sensing sheet is shown as 2; the oil film temperature sensor 3 and the lubricating oil viscosity sensor 4 are arranged on the inner wall of the oil outlet 5; the ultrasonic pressure measuring device comprises an ultrasonic transmitting module 6, an ultrasonic receiving module 7 and a data Processing module 8, wherein the ultrasonic transmitting module 6 and the ultrasonic receiving module 7 are symmetrically arranged on the upper side and the lower side of the bearing, the data Processing module comprises a DSP (DIGITAL SIGNAL Processing ) and a control circuit, and signals are processed in a digital form by a computer or special Processing equipment to obtain signal forms meeting the needs of people.
As shown in fig. 2, before ultrasonic pressure measurement is performed, thicknesses d1, d3 of the upper and lower bearings, a height d between the upper and lower bearings, and a diameter d2 of the rotation shaft are measured in the installation direction of the ultrasonic transmitting module 6 and the ultrasonic receiving module 7; the speed of ultrasonic wave in the bearing material and the rotating shaft material can be obtained by table lookup and is set as v 1,v2.
Petroleum belongs to hydrocarbon substances, the acoustic characteristics of the petroleum accord with the rule of Kneser liquid, at a certain temperature, the hydraulic pressure and the sound velocity have approximate linear relation, and when the pressure is higher and the temperature fluctuation range is not large, the linear relation is more stable, and the relation is satisfied:
Wherein, C is the ultrasonic sound velocity in the lubricating oil, and the unit is m/s; c 0 is the ultrasonic sound velocity in lubricating oil at normal temperature and one atmosphere pressure, and the unit is m/s; p is the pressure of the lubricating oil, and the unit is Pa; k is a proportionality coefficient. Wherein, C 0 and K are constants, which can be obtained by experiment or table lookup. From the formula (1), it can be seen that the pressure value of the pipeline can be obtained by measuring the sound velocity of ultrasonic waves in the bearing lubricating oil film.
The data processing module 8 and the ultrasonic control module 6 transmit ultrasonic waves, the ultrasonic receiving module 7 receives the ultrasonic waves, the total ultrasonic running time t 0 is measured, and the expression of C can be obtained:
And (3) combining the formula (1), the formula (2) and simplifying to obtain an expression of the lubricating oil pressure P:
From the formula (3), the oil pressure of lubricating oil between bearings, namely the oil film pressure, can be obtained only by obtaining the total running time t 0 of ultrasonic waves between bearings. The data processing module 8 performs the calculation and converts the pressure signal into a digital signal to be sent to an upper computer for subsequent processing.
The electric signal measured by the sensor is converted into a digital signal through an analog-to-digital converter and then transmitted to an upper computer for data processing. The analog-to-digital converter is connected with the oil film temperature sensor, the lubricating oil viscosity sensor and the journal rotating speed sensor, and is connected with the upper computer by adopting RS485 communication.
S2, building an LSTM-ELM neural network prediction model, searching an ELM optimal weight by adopting a WOA optimization function, processing acquired signals in real time by a threshold value, and outputting the occurrence probability of boundary lubrication.
The probability of occurrence of boundary lubrication can be described using the following equation:
P=f(T)+g(μ)+h(n)+j(py);
wherein P represents the occurrence probability of boundary lubrication, T represents the temperature of the lubricating oil during the operation of the shaft, mu represents the viscosity of the lubricating oil, n represents the rotational speed of the shaft, and P y represents the oil film pressure.
For lubricating oil temperature T: the increase of the oil temperature can lead to the decrease of the viscosity of the lubricating oil, so that the local oil film is easy to damage, the lubrication fails, the bearing capacity of the bearing is reduced, and even the lubricating oil is carbonized to burn tiles; the reduction of the oil temperature increases the viscosity of the oil, which increases the lubrication friction of the oil film and increases the power consumption of the bearing. In addition, the oil film is thickened, and machine vibration due to oil film vibration is generated.
For viscosity μ: the viscosity of the lubricating oil at different temperatures is different, and when the viscosity is too high, the resistance is large, so that uniform lubrication is not facilitated, and the heat dissipation effect is poor; when the viscosity is too low, the surface of the shaft cannot maintain enough oil film, and the lubrication effect is poor.
For rotational speed n: the higher the rotation speed is, the higher the temperature of the bearing and the lubricating oil is, and the viscosity of the lubricating oil is reduced under the action of centrifugal force.
Oil film pressure p y: the larger the load, the larger the oil film pressure and the smaller the oil film thickness; the smaller the load is, the smaller the oil film pressure is, the larger the oil film thickness is, the heat dissipation of the bearing is affected, and the oil film pressure mainly depends on the load.
The adopted deep learning prediction model can automatically fit the functional relation among the parameters according to the previous data.
A specific operational flow diagram is shown in fig. 3.
S3, substituting parameters of oil temperature, lubricating oil viscosity, oil film pressure and rotating speed of the shaft into the trained neural network model; the model of the neural network and the selection of parameters thereof are critical to the network performance, the invention builds an LSTM-ELM neural network prediction model, the collected parameters enter an ELM learning machine and an LSTM neural network at the same time, and on the ELM side, compared with the traditional BP neural network, the prediction accuracy of the ELM neural network on time sequence signals is far better than that of the BP neural network, a WOA optimization algorithm is adopted to find the optimal weight, a threshold value is adopted, and L2 regularization is adopted to prevent overfitting. For the network structure, two hidden layers are adopted, the first hidden layer is a full-connection layer of 10 neurons, the activation function is a full-connection layer of 5 neurons, the convergence and calculation are faster Relu, the operation efficiency is guaranteed, the response rate of an alarm can be effectively improved, the second hidden layer is a full-connection layer of 5 neurons, the activation function is softmax, the output obeys probability distribution, the output layer is three neurons, and the occurrence probabilities of three lubrication states are respectively represented. Initializing parameters by ELM, taking a training randomly generated threshold value and a weight value as an initial position vector of WOA, setting the population number N as 20 for a WOA optimization algorithm, and the maximum iteration number T max as 40, wherein the WOA optimization algorithm comprises the following specific steps:
D=|CX*(t)-X(t)| (4);
X(t+1)=X*(t)-AD (5);
A=2ar1-a (6);
C=2r2 (7);
D*=|X*(t)-X(t)| (10);
Where T is the current iteration number, X (t) represents the position of the current solution, X *(t) represents the position of the random optimal solution, D represents the distance between the two, r 1,r2 is a random number in (0, 1), a decreases linearly from 2 to 0, T max is the maximum number of iterations, a and C are coefficients, D * represents the distance from the optimal solution, b is a constant that measures the helix shape, l is a random number in (0, 1), and p refers to the probability. When the ELM diagnosis error rate is lower than 0.01%, calculating the current individual fitness value in the initialized population to obtain the optimal solution. Along with the decrease of the value a, the value A fluctuates along with the decrease, the WOA is set to iterate by adopting the formula (9) when the value A is less than or equal to 1, iterate by adopting the formula (5) when the value A is more than 1, calculate the individual fitness value, find the optimal parameter and substitute the optimal parameter into the ELM.
S4, at the LSTM side, 3 sigmoid activation functions and 2 tanh activation functions are adopted to realize information input, forgetting, updating and output, the number of stacked layers is 1, training is stopped when the training times reach 2000 times, and the dimension of the output result is 3, so that the occurrence probability of three lubrication states is respectively represented.
S5, substituting the optimal weight found by the WOA optimizing algorithm into an ELM extreme learning machine, taking the LSTM input layer and all hidden layers as ELM input, carrying out average value processing on ELM output and LSTM neural network output, and calculating to obtain probability distribution of a lubrication state; a specific neural network flow diagram is shown in fig. 4.
S6, loading a trained neural network model, substituting the trained neural network model into data acquired by the processed sensor, and calculating probability distribution of oil film lubrication states in real time, wherein the lubrication states comprise full film lubrication, boundary lubrication and mixed lubrication. When the oil film lubrication state is mainly full film lubrication, the lubrication effect is optimal, bearing abrasion can not occur basically, good protection effects can be achieved on the bearing and the journal surface, and the state attribute lamp can be green. When the oil film lubrication state is mainly mixed lubrication, partial boundary lubrication occurs at this time, but the state property lamp can be yellow within an acceptable range. When the oil film lubrication state is mainly boundary lubrication, the bearing can generate more severe friction and wear, so that the state attribute lamp is red, and an alarm is sent to the main control room of the turbine.
S7, when the occurrence probability of boundary lubrication is larger than a preset probability threshold value, sending out a warning signal to a ship main control room; when the probability of oil film boundary lubrication is larger than 0.9 at a certain moment, the oil film of the bearing is basically in boundary lubrication, the oil film temperature is kept in a higher zone, and serious abrasion occurs, which is undesirable, and at the moment, the machine crew can be enabled to add lubricating oil, antiwear agent or properly cool through starting an alarm, so that the occurrence of boundary lubrication can be effectively avoided, and the service life of the bearing is prolonged. Meanwhile, the probability threshold value can be properly adjusted according to actual conditions, and is generally preferably 0.8-0.9, so that the alarm is not too frequent, and meanwhile, the boundary lubrication can be effectively prevented.
The foregoing is a preferred embodiment of the present invention, and it will be apparent to those skilled in the art from this disclosure that changes, modifications, substitutions and alterations can be made without departing from the principles and spirit of the invention.
Claims (5)
1. The ship bearing lubrication state on-line monitoring system based on deep learning is characterized in that the monitoring system comprises the following steps:
s1, acquiring various data during bearing operation through various sensors arranged near a bearing; the data acquisition equipment comprises an ultrasonic pressure measuring device, an oil film temperature sensor, a lubricating oil viscosity sensor, a journal rotating speed sensor, an analog-to-digital converter and an upper computer;
s2, building an LSTM-ELM neural network prediction model, searching an ELM optimal weight by adopting a WOA optimization function, processing acquired signals in real time by a threshold value, and outputting the occurrence probability of boundary lubrication;
S3, substituting parameters of oil temperature, lubricating oil viscosity, oil film pressure and rotating speed of a shaft into the trained neural network model; an LSTM-ELM neural network prediction model is built, collected parameters enter an ELM learning machine and an LSTM neural network at the same time, on one side of the ELM, compared with a traditional BP neural network, a WOA optimization algorithm is adopted, an optimal weight is found, a threshold value is adopted, L2 regularization is adopted to prevent overfitting, for a network structure, two hidden layers are adopted, the first hidden layer is a fully-connected layer of 10 neurons, an activation function is converged and calculated faster Relu times, the operation efficiency is guaranteed, the response rate of an alarm is effectively improved, the second hidden layer is a fully-connected layer of 5 neurons, the activation function is softmax, probability distribution is output, the output layer is three neurons, the occurrence probability of three lubrication states is respectively represented, the parameters are initialized by the ELM, the threshold value randomly generated by training is taken as an initial position vector of the WOA;
S4, at the LSTM side, 3 sigmoid activation functions and 2 tanh activation functions are adopted to realize the input, forgetting, updating and output of information, the number of stacked layers is 1, training is stopped when the training times reach 2000 times, and the dimension of the output result is3, so that the occurrence probability of three lubrication states is respectively represented;
s5, substituting the optimal weight found by the WOA optimizing algorithm into an ELM extreme learning machine, taking the LSTM input layer and all hidden layers as ELM input, carrying out average value processing on ELM output and LSTM neural network output, and calculating to obtain probability distribution of a lubrication state;
S6, substituting the data acquired by the processed sensor into the trained neural network model, and calculating probability distribution of an oil film lubrication state in real time, wherein the lubrication state comprises full film lubrication, boundary lubrication and mixed lubrication, when the oil film lubrication state is mainly full film lubrication, the lubrication effect is optimal, bearing abrasion basically cannot occur, the protection effect is achieved on the surfaces of the bearings and the journals, so that the state attribute lamps are green, when the oil film lubrication state is mainly mixed lubrication, although partial boundary lubrication occurs, the state attribute lamps are in an acceptable range, the state attribute lamps can be yellow, when the oil film lubrication state is mainly boundary lubrication, the bearings can generate more severe friction abrasion, the state attribute lamps can be red, and an alarm is sent to a turbine main control room;
s7, when the occurrence probability of boundary lubrication is larger than a preset probability threshold value, sending out a warning signal to a ship main control room; when the probability of oil film boundary lubrication is larger than 0.9 at a certain moment, the oil film of the bearing is basically in boundary lubrication, the oil film temperature is kept in a higher zone, serious abrasion occurs, and at the moment, by starting an alarm, a crew member adds lubricating oil, antiwear agent or properly cools, boundary lubrication is effectively avoided, the service life of the bearing is prolonged, meanwhile, a probability threshold value is properly adjusted according to actual conditions, and the probability threshold value is 0.8-0.9, so that the alarm is not too frequent, and boundary lubrication is effectively prevented.
2. The online monitoring system for the lubrication state of the ship bearing based on deep learning as claimed in claim 1, wherein in the step S1, the journal rotation speed sensor is a non-contact photoelectric sensor, and is installed on one side of a bearing housing, the rotation speed sensor (1) is installed on the upper side of a rotating shaft, and the electro-optic sensing sheet (2) is installed on the rotating shaft; the oil film temperature sensor (3) and the lubricating oil viscosity sensor (4) are arranged on the inner wall of the oil outlet (5); the ultrasonic pressure measuring device comprises an ultrasonic transmitting module (6), an ultrasonic receiving module (7) and a data processing module (8), wherein the ultrasonic transmitting module (6) and the ultrasonic receiving module (7) are symmetrically arranged on the upper side and the lower side of the bearing, and the data processing module comprises a DSP and a control circuit.
3. The on-line monitoring system for the lubrication state of the ship bearings based on deep learning according to claim 2, wherein the step S1 is to measure the thicknesses d1 and d3 of the bearings at the upper and lower sides, the height d between the bearings at the upper and lower sides, and the diameter d2 of the rotating shaft in the installation direction of the ultrasonic transmitting module (6) and the receiving module (7) before the ultrasonic pressure measurement is performed; the speed of ultrasonic wave in the bearing material and the rotating shaft material can be obtained by table lookup and is set as v 1,v2,
The hydraulic pressure and the sound velocity have approximate linear relation, the linear relation is stable, and the relation formula is satisfied:
Wherein, C is the ultrasonic sound velocity in the lubricating oil, and the unit is m/s; c 0 is the ultrasonic sound velocity in lubricating oil at normal temperature and one atmosphere pressure, and the unit is m/s; p is the pressure of the lubricating oil, and the unit is Pa; k is a proportionality coefficient, wherein C 0 and K are constants, and the proportionality coefficient is obtained through experiments or table lookup; as can be seen from the formula (1), the pressure value of the oil film can be obtained by measuring the sound velocity of ultrasonic waves in the bearing lubricating oil film;
The data processing module (8) and the ultrasonic transmitting module (6) transmit ultrasonic waves, the ultrasonic receiving module (7) receives the ultrasonic waves, the total ultrasonic running time t 0 is measured, and the expression of C is obtained:
the lubricating oil pressure P expression is obtained by simplifying the combined formulas (1) and (2):
As can be seen from the formula (3), the oil pressure of the lubricating oil between the bearings, namely the oil film pressure, can be obtained only by obtaining the total running time t 0 of the ultrasonic wave between the bearings, the data processing module 8 performs the calculation and converts the pressure signal into a digital signal to be sent to an upper computer for subsequent processing;
The electric signals measured by the sensor are converted into digital signals through an analog-to-digital converter, the digital signals are transmitted to an upper computer for data processing, and the analog-to-digital converter is connected to the oil film temperature sensor, the lubricating oil viscosity sensor and the journal rotation speed sensor and is connected to the upper computer by adopting RS485 communication.
4. The online monitoring system for lubrication state of ship bearings based on deep learning as set forth in claim 1, wherein the occurrence probability of boundary lubrication in the step S2 is described using the following equation:
P=f(T)+g(μ)+h(n)+j(py);
Wherein P represents the occurrence probability of boundary lubrication, T represents the temperature of the lubricating oil during the operation of the shaft, mu represents the viscosity of the lubricating oil, n represents the rotating speed of the shaft, and P y represents the oil film pressure;
For lubricating oil temperature T: the increase of the oil temperature can lead to the decrease of the viscosity of the lubricating oil, so that the local oil film is easy to damage, the lubrication fails, the bearing capacity of the bearing is reduced, and even the lubricating oil is carbonized to burn tiles; the oil temperature is reduced to increase the viscosity of the oil, thereby increasing the lubrication friction of the oil film, increasing the power consumption of the bearing, thickening the oil film, and generating machine vibration caused by the vibration of the oil film;
For viscosity μ: the viscosity of the lubricating oil at different temperatures is different, and when the viscosity is too high, the resistance is large, so that uniform lubrication is not facilitated, and the heat dissipation effect is poor; when the viscosity is too low, the axial surface cannot keep enough oil film, and the lubrication effect is poor;
for rotational speed n: the higher the rotating speed is, the higher the temperature of the bearing and the lubricating oil is, and the viscosity of the lubricating oil is reduced under the action of centrifugal force;
Oil film pressure p y: the larger the load, the larger the oil film pressure and the smaller the oil film thickness; the smaller the load is, the smaller the oil film pressure is, the larger the oil film thickness is, the heat dissipation of the bearing is affected, and the oil film pressure mainly depends on the load;
the adopted deep learning prediction model automatically fits the functional relation among the parameters according to the previous data.
5. The online monitoring system for the lubrication state of the ship bearing based on deep learning as set forth in claim 1, wherein the step S3 sets the population number N to 20, the maximum iteration number T max to 40, and the WOA optimization algorithm specifically includes the steps of:
D=|CX*(t)-X(t)| (4);
X(t+1)=X*(t)-AD (5);
A=2ar1-a (6);
C=2r2 (7);
D*=|X*(t)-X(t)| (10);
Wherein T is the current iteration number, X (t) is the current solution position, X *(t) is the random optimal solution position, D is the distance between the two, r 1,r2 is a random number in (0, 1), a linearly decreases from 2 to 0, T max is the maximum iteration number, A and C are coefficients, D * is the distance from the optimal solution, b is a constant for measuring the spiral shape, l is a random number in (0, 1), p is the probability, and ELM diagnosis error rate is lower than 0.01%, and the current individual fitness value is calculated in the initialized population to obtain the optimal solution; along with the decrease of the value a, the value A fluctuates along with the decrease, the WOA is set to iterate by adopting the formula (9) when the value A is less than or equal to 1, iterate by adopting the formula (5) when the value A is more than 1, calculate the individual fitness value, find the optimal parameter and substitute the optimal parameter into the ELM.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210538087.2A CN114878162B (en) | 2022-05-18 | 2022-05-18 | Ship bearing lubrication state on-line monitoring system based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210538087.2A CN114878162B (en) | 2022-05-18 | 2022-05-18 | Ship bearing lubrication state on-line monitoring system based on deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114878162A CN114878162A (en) | 2022-08-09 |
CN114878162B true CN114878162B (en) | 2024-06-18 |
Family
ID=82676537
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210538087.2A Active CN114878162B (en) | 2022-05-18 | 2022-05-18 | Ship bearing lubrication state on-line monitoring system based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114878162B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117537951B (en) * | 2024-01-10 | 2024-03-26 | 西南交通大学 | Method and device for detecting internal temperature rise of superconducting suspension based on deep learning |
CN118032327A (en) * | 2024-04-15 | 2024-05-14 | 山东能源数智云科技有限公司 | Equipment intelligent lubrication monitoring method and device based on artificial intelligence |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2018025450A (en) * | 2016-08-09 | 2018-02-15 | オークマ株式会社 | Bearing diagnostic device |
DE102017223418B4 (en) * | 2017-12-20 | 2023-05-25 | Zf Friedrichshafen Ag | Model-based method and system for condition monitoring of a plain bearing, especially for wind turbines |
CN111272424B (en) * | 2020-01-20 | 2021-10-22 | 陕西理工大学 | Water-lubricated bearing on-line monitoring test platform and bearing characteristic test analysis method |
CN111881544A (en) * | 2020-06-11 | 2020-11-03 | 重庆大学 | Harmonic reducer wave generator mixed lubrication analysis method |
JP7349966B2 (en) * | 2020-07-30 | 2023-09-25 | アイダエンジニアリング株式会社 | Machine lubricating oil supply system monitoring method and device |
CN112858642A (en) * | 2021-01-13 | 2021-05-28 | 中国人民解放军92557部队 | Ship lubricating oil detection system based on intelligent Internet of things |
-
2022
- 2022-05-18 CN CN202210538087.2A patent/CN114878162B/en active Active
Non-Patent Citations (2)
Title |
---|
滑动轴承非线性油膜力的神经网络模型;秦平, 孟志强, 朱均;摩擦学学报;20020630(03);全文 * |
船舶艉轴承动态工况润滑特性分析;张天勇;朱汉华;范世东;;船舶工程;20100215(01);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN114878162A (en) | 2022-08-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN114878162B (en) | Ship bearing lubrication state on-line monitoring system based on deep learning | |
US8924162B2 (en) | Turbine-to-turbine prognostics technique for wind farms | |
CN111046491B (en) | Method and device for estimating fuel consumption of diesel engine of large ship | |
KR102457548B1 (en) | Marine Vessel Performance Monitoring | |
Qian et al. | Integrated data‐driven<? show [AQ ID= Q1]?> model‐based approach to condition monitoring of the wind turbine gearbox | |
CN105205569B (en) | State of fan gear box online evaluation method for establishing model and online evaluation method | |
CN112307688B (en) | System and method for monitoring, managing and optimizing energy efficiency of electric propulsion ship | |
CN113525655B (en) | Ship energy-electric power control management system based on machine learning | |
Coraddu et al. | A novelty detection approach to diagnosing hull and propeller fouling | |
CN102635513A (en) | Method for lubricating at least one blade pitch bearing of a wind turbine | |
CN112733283A (en) | Wind turbine generator component fault prediction method | |
CN103573307A (en) | Methods and systems for monitoring engine oil temperature of an operating engine | |
WO2011143531A2 (en) | Turbine-to-turbine prognostics technique for wind farms | |
CN113847216B (en) | Fan blade state prediction method, device, equipment and storage medium | |
TW201901031A (en) | System and method for monitoring grease of wind power generator | |
CN106981873A (en) | A kind of isolated island formula power system based on dynamic behavior is hidden failure prediction method | |
Baltazar et al. | A review on neurocomputing based wind turbines fault diagnosis and prognosis | |
Dong et al. | An online health monitoring framework for traction motors in high-speed trains using temperature signals | |
CN116383615A (en) | Multisource sensing information fusion and expansion method based on system state characterization | |
CN116629656A (en) | Ship energy consumption assessment method | |
CN116089787A (en) | Ship subsystem running state analysis method and system based on analytic hierarchy process | |
CN112598057A (en) | Coal-fired power plant boiler fault diagnosis method based on FEKNN strategy | |
Choudhary et al. | IoT Enabled Condition Monitoring of Low Voltage Motors using Fuzzy Inference System | |
Yang et al. | Fault detection of high-speed train axle bearings based on a hybridized physical and data-driven temperature model | |
Fang et al. | Failure Prediction of Wind Turbines Using Improved Gray Relation Analysis Based Support Vector Machine Method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant |