CN111324989A - GA-BP neural network-based gear contact fatigue life prediction method - Google Patents
GA-BP neural network-based gear contact fatigue life prediction method Download PDFInfo
- Publication number
- CN111324989A CN111324989A CN202010197216.7A CN202010197216A CN111324989A CN 111324989 A CN111324989 A CN 111324989A CN 202010197216 A CN202010197216 A CN 202010197216A CN 111324989 A CN111324989 A CN 111324989A
- Authority
- CN
- China
- Prior art keywords
- layer
- output
- neural network
- hidden layer
- weight
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 38
- 238000012360 testing method Methods 0.000 claims abstract description 36
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 14
- 230000002068 genetic effect Effects 0.000 claims abstract description 14
- 238000011156 evaluation Methods 0.000 claims abstract description 8
- 238000003062 neural network model Methods 0.000 claims abstract description 7
- 238000009661 fatigue test Methods 0.000 claims abstract description 4
- 238000009795 derivation Methods 0.000 claims abstract description 3
- 230000006870 function Effects 0.000 claims description 37
- 230000004913 activation Effects 0.000 claims description 21
- 239000000463 material Substances 0.000 claims description 17
- 238000010606 normalization Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 8
- 230000035772 mutation Effects 0.000 claims description 7
- 238000012546 transfer Methods 0.000 claims description 6
- 108090000623 proteins and genes Proteins 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- 101150084044 P gene Proteins 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 239000002131 composite material Substances 0.000 claims description 2
- 150000001875 compounds Chemical class 0.000 claims description 2
- 238000011478 gradient descent method Methods 0.000 claims description 2
- 238000012886 linear function Methods 0.000 claims description 2
- 238000004519 manufacturing process Methods 0.000 abstract description 4
- 230000008901 benefit Effects 0.000 abstract description 3
- 230000007246 mechanism Effects 0.000 abstract description 3
- 238000013461 design Methods 0.000 abstract 1
- 238000005461 lubrication Methods 0.000 description 12
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 238000012897 Levenberg–Marquardt algorithm Methods 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000001050 lubricating effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Physiology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a gear contact fatigue life prediction method based on a GA-BP neural network, which comprises the following steps: 1. collecting gear contact fatigue test data and normalizing the data to be used as sample data of a BP neural network model; 2. constructing a BP neural network structure; 3. optimizing the weight and the threshold of the BP neural network by using a genetic algorithm, and training the BP neural network; 4. calculating prediction accuracy evaluation parameter determination coefficientAcquiring an optimized weight and a threshold when the set value is met; 5. and testing the established GA-BP neural network by using a test sample set. Compared with the existing gear contact fatigue life prediction method based on the physical model, the method has the advantages of low cost and accurate predictionThe gear fatigue life prediction method has the advantages of being high in degree, free of derivation according to a failure mechanism, capable of achieving gear fatigue life prediction, improving prediction accuracy, simple to use and capable of providing a new technical means for gear design and manufacture.
Description
Technical Field
The invention belongs to the technical field of intelligent manufacturing, and particularly relates to a method for predicting contact fatigue life of a gear.
Background
The gear is an important mechanical basic part, is widely applied to various industrial fields for motion and power transmission, and the fatigue life of the gear directly determines the service performance of the whole machine. With the development of heavy equipment such as wind power, high-speed rail and aviation towards high reliability, long service life and intellectualization, higher requirements are put forward on the service life and reliability of the gear. If the gear is subjected to fatigue failure damage in the service stage, the equipment is shut down to influence the production, and the personal and property safety is endangered. The fatigue life of the gear is accurately predicted, sudden vicious accidents of equipment can be avoided, and long-term safe and reliable operation of mechanical equipment is ensured.
The conventional gear fatigue life prediction method mainly comprises a fatigue accumulated damage theory, a mathematical model and the like, wherein the fatigue accumulated damage theory considers that the maximum life of a part is N times when the part is under the action of a constant stress amplitude, namely the part is in operation for N times, and a dangerous point of the part is in fatigue failure; the mathematical model predicts the fatigue life of the gear by establishing the relationship between the gear parameters and the fatigue life; although the method can predict the fatigue life and achieve certain accuracy, various defects exist, such as imperfect accumulation theory of fatigue damage, no consideration of the influence of loading load sequence on the fatigue damage and the like, the methods are time-consuming and high in cost, and the result often has larger deviation from the test value.
The data driving method has the advantage that the internal mechanism of the material is not required to be comprehensively clarified, and a new technical means is provided for predicting the contact fatigue life of the gear. The BP neural network has strong nonlinear fitting capability and self-adaptive capability, so that the BP neural network is widely applied to the prediction of the service life of parts, but the initial weight and the threshold are randomly selected, so that the problem of falling into a local minimum value is easily caused, and therefore, the initial weight and the threshold of the BP neural network are optimized by adopting a genetic algorithm, and the accurate prediction of the contact fatigue life of the gear is realized.
The terms: the genetic algorithm optimized BP neural network is called GA-BP neural network for short.
Disclosure of Invention
Aiming at the problems in the prior art, the technical problem to be solved by the invention is to provide a gear contact fatigue life prediction method based on a GA-BP neural network, which can predict the gear fatigue life, can improve the prediction accuracy, and is simple to use and low in cost.
The technical problem to be solved by the invention is realized by the technical scheme, which comprises the following steps:
s1, collecting test data of the gear contact fatigue test, taking the structural characteristic parameters and the material parameters of the gear as input parameters and the gear contact fatigue life as output parameters, and carrying out normalization processing on each group of test data to take the test data as a training sample and a test sample;
s2, respectively determining the number of nodes of an input layer and an output layer according to the input parameters and the output parameters in the S1, determining the number of hidden layers and the number of nodes of the hidden layers, selecting the type of an activation function between each layer of the neural network, and constructing a BP neural network structure;
s3, optimizing the initial weight and threshold of the BP neural network by using a genetic algorithm, setting initial training parameters, and training the BP neural network structure by adopting the training samples in S1;
s4, calculating prediction accuracy evaluation parameter determination coefficient R2If the weight value meets the set value, extracting the trained weight value and threshold value to obtain a GA-BP neural network model; if not, returning to S3 to train again until determining the coefficient R2The set value is met;
and S5, testing the established GA-BP neural network by using the test sample set, and evaluating the network performance and the prediction accuracy.
Therefore, a GA-BP neural network model is obtained to predict the contact fatigue life of the gear.
Compared with the prior art, the invention has the technical effects that:
compared with the existing gear contact fatigue life prediction method based on a physical model, the gear contact fatigue life prediction method is low in cost, high in prediction precision, free of derivation according to failure mechanisms, capable of achieving gear fatigue life prediction, capable of improving prediction accuracy, simple to use, capable of providing a new evaluation means for manufacturing and using of gears, saving a large amount of time and cost, and has popularization and application values.
Drawings
The drawings of the invention are illustrated as follows:
FIG. 1 is a flow chart of one embodiment of the present method;
FIG. 2 is a block diagram of one embodiment of the present method;
FIG. 3 is a diagram of prediction values and decision coefficients according to an embodiment of the present invention;
FIG. 4 is a comparison graph of predicted values and test values according to one embodiment of the present invention.
Detailed Description
The invention is further illustrated by the following examples in conjunction with the accompanying drawings:
as shown in fig. 1, the present embodiment includes the following steps:
s1, collecting test data of the gear contact fatigue test, analyzing the influence of gear contact fatigue life by using gear parameters, carrying out normalization processing on each group of test data by using the structural characteristic parameters and the material parameters of the gear as input parameters and the gear contact fatigue life as output parameters, and using the normalized test data as training samples and test samples.
143 sets of gear contact fatigue life test data were collected from a number of colleges and research institutes. The selected input parameters include but are not limited to tooth number, roughness, contact stress, tooth surface hardness and material parameters, the gear contact fatigue life is taken as an output parameter, and the unit of test data is shown in table 1:
TABLE 1 units of test data
Parameter(s) | Unit of | Parameter(s) | Unit of |
Lubrication parameter | Hardness of tooth surface | HV | |
Roughness of | Ra | Parameters of the material | |
Stress of contact | MPa | Contact fatigue life of gear | Cycle |
Mapping the collected data to a (0, 1) interval through a normalization formula, wherein the normalization formula of the data is as follows:
in the formula, xmIs normalized number, x is number to be normalized or number output by training setminOutputting the minimum value, x, for the minimum value or training set to be normalizedmaxThe maximum value of the input to be normalized or the maximum value of the output of a training set is obtained, (wherein the output predicted value of the testing set is subjected to inverse normalization to obtain the contact fatigue life of the gearAnd (5) predicting the value. )
The data after the normalization process are divided into 108 training sample sets and 35 testing sample sets.
The partial training sample set before normalization is shown in table 2.
TABLE 2 partial training sample set data before normalization
Note: the material type and lubrication mode are numerically distinguished, wherein material 1 represents 20MnCr5, material 2 represents 18Cr2Ni4WA, material 5 represents AISI 4118, material 10 represents 16MnCr5, lubrication 1 represents no lubrication, and lubrication 2 represents oil lubrication.
And S2, determining the number of nodes of an input layer and the number of nodes of an output layer according to the input parameters and the output parameters in the S1, determining the number of hidden layers and the number of nodes of the hidden layers, selecting the type of an activation function between the layers of the neural network, and constructing the BP neural network structure.
In this embodiment, the number of nodes of the input layer of the BP neural network is 5, the number of nodes of the hidden layer is 2, the number of nodes of the output layer is 1, and the range of the number of nodes of the hidden layer is determined according to an empirical formula, where the empirical formula is:
in the formula, n is the number of hidden layer nodes, m is the number of input layer nodes, p is the number of output layer nodes, and a ranges from 1 to 10. And determining the number of the optimal hidden layer nodes to be 8 by a trial and error method.
The BP neural network structure is '5-8-8-1', an activation function between an input layer and a first hidden layer selects a hyperbolic tangent function Tanh, an activation function between the first hidden layer and a second hidden layer selects an S-shaped function Sigmoid, an activation function between the second hidden layer and an output layer selects a linear function Purelin, and the activation functions between the layers are as follows:
purelin activation function: f. of3(x)=x
x is the total input of any node of the hidden layer or the output layer, and f (x) is the total output of any node of the hidden layer or the output layer.
Fig. 2 shows a structure of a BP neural network, where in this embodiment, a transfer formula from an input layer to a first hidden layer is:
in the formula, H1jJ is the output of the jth node of the first hidden layer, 1,2, …,8, f1As a function of Tanh activation, ωijIs the weight value, x, between the ith node of the input layer and the jth node of the first hidden layeriIs an input value of the ith node of the input layer, ajIs the threshold of the jth node of the first hidden layer.
The transfer formula from the first hidden layer to the second hidden layer is:
in the formula, H2kFor the output of the kth node of the second hidden layer, k is 1,2, …,8, f2For Sigmoid activation functions, ωjkIs the weight between the jth node of the first hidden layer and the kth node of the second hidden layer, bkIs the threshold of the kth node of the second hidden layer.
The second hidden layer to output layer transfer formula is:
wherein y is the output value of the output layer, f3Activating functions for Purelin,ωkIs the weight between the kth node of the second hidden layer and the output layer, and c is the output layer threshold.
S3, optimizing the initial weight and threshold of the BP neural network by using a genetic algorithm, setting initial training parameters including iteration times, minimum error of a training target, learning rate and maximum failure times, and training the BP neural network by using training samples.
The initial parameters of the genetic algorithm are: 70 population scale, 100 evolution times, 0.8 cross probability and 0.2 mutation probability.
The specific process of genetic algorithm optimization is as follows:
step 1) population initialization, wherein according to a network structure of a BP neural network 5-8-8-1 shown in fig. 2, the total number of 5 × 8+8 × 8+8 × 1 is 112 weight values, and the total number of 8+8+1 is 17 threshold values, so that the individual coding length of a genetic algorithm is 112+ 17-129, the individual number population scale is 70, an individual coding method is real number coding, each individual is a real number string and consists of the BP neural network weight value and the threshold value.
Step 2) selecting a fitness function
The fitness function is the inverse of the sum of the square of the output layer prediction output and the expected output error:
in the formula, FlAnd the fitness function value of the ith individual, y is the predicted output of the BP neural network, and d is the expected output.
Step 3), selecting operation: genetic algorithm selection roulette method, selection strategy based on fitness proportion, probability p of each individual l being selectedlComprises the following steps:
wherein k is a coefficient, FlIs the first oneFitness function value of the volume.
Step 4), cross operation: individuals employ real number encoding and cross-manipulation methods employ real number cross-manipulation, see the literature "a genetic algorithm project," Whitley, darrell. 65-85 ("genetic Algorithm guide", Whitley, Darrell. statistics and computing4.2 (1994): 65-85.): the r individual hrAnd the s-th individual hsThe method of interleaving at t bits is as follows:
hrt′=hrt(1-e)+hste
hst′=hst(1-e)+hrte
in which e is [0,1]A random number in between; h isrt' is the r-th individual after the crossover operation, hst' is the s-th individual after the crossover operation.
Step 5), mutation operation: selecting the p gene h of the o individualopCarrying out mutation by the following operation method:
in the formula, hmaxIs gene hopUpper bound of, hminIs gene hopLower boundary of (f), (g) r2(1-g/Gmax)2,r2Is a random number, G is the current iteration number, GmaxFor maximum evolutionary number, r is [0,1 ]]Random number of cells, hop' is the p-th gene of the o-th individual after mutation operation.
And 6), repeating the step 3), the step 4) and the step 5) until the evolution times reach 100, and decoding the optimal individual to be used as the weight and the threshold value after the BP neural network is optimized.
The weight and the threshold after the BP neural network optimization are as follows:
TABLE 3 weight ω between input layer and first hidden layerij
0.4218 | 0.2447 | 0.3578 | 0.4907 | 0.3844 | -0.2171 | 0.2578 | 0.4580 |
-0.2863 | 0.9478 | 0.1458 | 0.7965 | 0.1166 | -0.8059 | -0.2495 | -0.8747 |
-0.7241 | -0.1998 | 0.7721 | 0.6914 | -0.5112 | -0.0388 | 0.4085 | 0.9574 |
0.7924 | 0.2281 | 0.8561 | 0.3451 | 0.3491 | 0.3151 | 0.6033 | 0.5980 |
-0.8740 | 0.6742 | 0.8846 | 0.2431 | -0.5790 | -0.0948 | 0.4490 | 0.9830 |
Note: the ith row and the jth column of values represent the weight omega between the ith node of the input layer and the jth node of the first hidden layerij
TABLE 4 weight ω between first and second hidden layersjk
0.7752 | 0.9144 | -0.9630 | 0.5424 | -0.1378 | 0.4338 | -0.0500 | 0.6697 |
0.2636 | 0.0761 | -0.9036 | 0.2976 | -0.9921 | 0.3810 | -0.5291 | 0.9403 |
0.2336 | 0.3169 | -0.3379 | -0.3860 | 0.6275 | 0.0070 | 0.4118 | -0.6723 |
0.8509 | 0.5511 | -0.8442 | -0.7990 | -0.2139 | 0.4650 | 0.3608 | 0.4789 |
-0.3446 | -0.0218 | 0.6606 | -0.2344 | -0.0162 | 0.6991 | 0.0150 | 0.5359 |
-0.5811 | 0.3932 | 0.5897 | 0.9315 | -0.4478 | 0.5806 | -0.4985 | 0.9826 |
-0.2206 | 0.7731 | -0.7684 | -0.0700 | -0.3670 | -0.5834 | 0.7251 | -7277 |
-0.5653 | -0.4367 | -0.1178 | -0.6378 | -0.2122 | -0.1526 | -0.7138 | 0.9720 |
-0.8740 | 0.6742 | 0.8846 | 0.2431 | -0.5790 | -0.0948 | 0.4490 | 0.9830 |
Note: the jth row and kth column values represent the weight ω between the jth node of the first hidden layer and the kth node of the second hidden layerjk
TABLE 5 weight ω between the second hidden layer and the output layerk
0.9144 | 0.0761 | 0.3169 | 0.5511 | -0.0218 | 0.3932 | -0.7731 | -0.4367 |
Note: the kth column value represents the weight ω between the kth node of the second hidden layer and the output layerk
TABLE 6 threshold a of the first hidden layerj
-0.0736 | -0.3354 | 0.1699 | 0.5334 | -0.4973 | -0.6228 | -0.1609 | 0.5524 |
Note: the jth column value represents the threshold value a of the jth node of the first hidden layerj
TABLE 7 threshold b for the second hidden layerk
0.6615 | 0.1516 | -0.8594 | -0.5865 | -0.3272 | 0.1648 | 0.5183 | -0.3050 |
Note: the kth column value represents the threshold b of the kth node of the second hidden layerk
Output node threshold: and c is 0.2342.
In the step, a Levenberg-Marquardt algorithm is adopted for training to train the BP neural network, and initial training parameters are set as follows: iteration number 1000, training target minimum error 1e-4Learning rate 0.001, maximum number of failures 15, concrete training process includes:
step 1), carrying out forward propagation on the training sample, and calculating the square sum of the predicted output and the expected output error of the output layer.
The forward propagation specifically includes: training samples are input from an input layer, processed by an activation function of a first hidden layer, output by a node of the first hidden layer, processed by an activation function of a second hidden layer and output by a node of the second hidden layer; and after the activation function processing of the output layer, the actual output value is obtained through the output of the output layer node, and the error sum of the square of the predicted output value and the expected output value of the output node is calculated.
Randomly selecting the u-th input sample and the corresponding output test data, wherein the sum of the square of the error between the predicted output and the expected output is represented by the formula:
E(u)=(d(u)-y(u))2
where e (u) is the sum of the square of the predicted output and the expected output error of the u-th input sample, d (u) is the expected output of the u-th input sample (i.e., the output test data), and y (u) is the predicted output of the u-th input sample.
Step 2), judging whether the sum of the square of the error between the predicted output and the expected output is less than the minimum error of the training target (1 e)-4) The requirements of (1); if not, entering the step 3) to carry out a back propagation process, and if so, finishing the training of the BP neural network modelAnd (4) bundling.
And 3) performing a back propagation process by adopting a gradient descent method (gradient is a vector formed by the error function on the weight and the partial derivative of the threshold), obtaining a weight updating formula from the second hidden layer to the output layer, a weight updating formula between the input layer and the first hidden layer and a weight updating formula between the first hidden layer and the second hidden layer by utilizing the square sum function E (u) of the predicted output and the expected output error to obtain a weight composite derivative, and respectively updating the weight of the output layer, the weight between the input layer and the first hidden layer and the weight between the first hidden layer and the second hidden layer according to the updating formulas.
Wherein, the weight ω between the input layer and the first hidden layerijThe update formula is:
weight ω between first hidden layer and second hidden layerjkThe update formula is:
weight ω from the second hidden layer to the output layerkThe update formula is:
in the formula, η represents the learning rate (0.001).
Step 4), inputting the updated output layer weight obtained in the step 3), the weight between the input layer and the first hidden layer, the weight between the first hidden layer and the second hidden layer into the input layer and the hidden layer in the forward propagation in the step 1) to continue the forward propagation, and returning to the step 1).
S4, calculating prediction accuracy evaluation parameter determination coefficient R2If the weight value meets a set value (generally greater than 0.99), extracting the weight value and the threshold value after training to obtain a GA-BP neural network model; if not, go back to S3 to train again until determiningCoefficient R2The set value is met;
in this step, a prediction accuracy evaluation parameter determination coefficient R2The calculation formula is as follows:
in the formula (I), the compound is shown in the specification,in order to predict the value of the model,is the sample average value, yoIs the sample data value, and v is the number of training samples.
S5, testing the established GA-BP neural network by using the test sample set, performing inverse normalization on the prediction result to obtain a predicted gear contact fatigue life value, and calculating a prediction precision evaluation parameter decision coefficient R2And evaluating the network performance and the prediction accuracy, and testing partial data of the sample set as shown in a table 8:
table 8 partial test sample set
Note: the material type and the lubrication mode are numerically distinguished, wherein material 1 represents 20MnCr5, material 2 represents 18Cr2Ni4WA, material 5 represents AISI 4118, lubrication 1 represents no lubrication, and lubrication 2 represents oil lubrication.
And calculating through a GA-BP neural network model, performing inverse normalization on the prediction result to obtain a predicted gear contact fatigue life value, and comparing the predicted gear contact fatigue life value with a test value, as shown in fig. 3 and 4. As can be seen from fig. 3: test set decision coefficient R20.9587; the gear contact fatigue life prediction precision of the test sample set is high and stable. In FIG. 4, the abscissa of the data point is measuredThe actual value of the contact fatigue life of the gear is tested, the ordinate is the predicted value of the contact fatigue life of the gear, the linear straight line indicates that the predicted value is equal to the actual value, the closer the data point is to the straight line, the higher the prediction precision is, and the higher the prediction precision can be seen.
And introducing a group of data (roughness Ra0.6, contact stress 1997.09MPa, surface hardness HV695, material 18CrNiMo, lubricating mode oil lubrication and real life 1.4E +07Cycle) which are not concentrated in the test sample into the model to obtain a predicted value 1.51E +07Cycle, so as to meet the precision requirement.
The method improves the prediction accuracy of the contact fatigue life of the gear, particularly overcomes the defects of complex prediction and lower prediction accuracy of the existing gear fatigue life, and is convenient, effective and low in cost.
Claims (4)
1. A gear contact fatigue life prediction method based on a GA-BP neural network is characterized by comprising the following steps:
s1, collecting test data of the gear contact fatigue test, taking the structural characteristic parameters and the material parameters of the gear as input parameters and the gear contact fatigue life as output parameters, and carrying out normalization processing on each group of test data to take the test data as a training sample and a test sample;
s2, respectively determining the number of nodes of an input layer and an output layer according to the input parameters and the output parameters in the S1, determining the number of hidden layers and the number of nodes of the hidden layers, selecting the type of an activation function between each layer of the neural network, and constructing a BP neural network structure;
establishing a prediction model by using a BP neural network containing 2 hidden layers, wherein the number of nodes of the hidden layers meets the following requirements:
in the formula, n is the number of hidden layer nodes, m is the number of input layer nodes, p is the number of output layer nodes, and a ranges from 1 to 10;
the activation function between the input layer and the first hidden layer selects a hyperbolic tangent function Tanh, the activation function between the first hidden layer and the second hidden layer selects an S-shaped function Sigmoid, and the activation function between the second hidden layer and the output layer selects a linear function Purelin;
the transfer formula from the input layer of the BP neural network to the first hidden layer is as follows:
in the formula, H1jJ is the output of the jth node of the first hidden layer, 1,2, …, n, f1As a function of Tanh activation, ωijIs the weight value, x, between the ith node of the input layer and the jth node of the first hidden layeriIs an input value of the ith node of the input layer, ajA threshold value of the jth node of the first hidden layer;
the transfer formula from the first hidden layer to the second hidden layer of the BP neural network is as follows:
in the formula, H2kFor the output of the kth node of the second hidden layer, k is 1,2, …, n, f2For Sigmoid activation functions, ωjkIs the weight between the jth node of the first hidden layer and the kth node of the second hidden layer, bkA threshold value of a kth node of the second hidden layer;
the transfer formula from the second hidden layer to the output layer of the BP neural network is as follows:
wherein y is the output value of the output layer, f3As Purelin activation function, ωkC is the weight between the kth node of the second hidden layer and the output layer, and c is the threshold value of the output layer;
s3, optimizing the initial weight and threshold of the BP neural network by using a genetic algorithm, setting initial training parameters, and training the BP neural network structure by adopting the training samples in S1;
the genetic algorithm optimization process comprises the following steps:
step 1), population initialization: setting a population number q, wherein individual codes of the population are real number codes, and each individual consists of a BP neural network weight and a threshold;
step 2) selecting a fitness function
The fitness function is the inverse of the sum of the square of the output layer prediction output and the expected output error:
in the formula, FlThe fitness function value of the ith individual is obtained, y is the prediction output of the BP neural network, and d is the expected output;
step 3), selecting operation: genetic algorithm selection roulette method, probability p of each individual l being selectedlComprises the following steps:
wherein k is a coefficient, FlThe fitness function value of the ith individual;
step 4), cross operation: the individuals adopt real number coding, the cross operation method adopts a real number cross method, and the r-th individual hrAnd the s-th individual hsThe method of interleaving at t bits is as follows:
hrt′=hrt(1-e)+hste
hst′=hst(1-e)+hrte
in which e is [0,1]A random number in between; h isrt' is the r-th individual after the crossover operation, hst' is the s-th individual after the crossover operation;
step 5), mutation operation: selecting the p gene h of the o individualopPerforming mutation by the following method:
In the formula, hmaxIs gene hopUpper bound of, hminIs gene hopLower boundary of (f), (g) r2(1-g/Gmax)2,r2Is a random number, G is the current iteration number, GmaxFor maximum evolutionary number, r is [0,1 ]]Random number of cells, hop' is the p gene of the o individual after mutation operation;
step 6), repeating the step 3), the step 4) and the step 5) until the evolution times are reached, and decoding the optimal individual to be used as the weight and the threshold value after the BP neural network is optimized;
training the BP neural network by using the training set in S1;
s4, calculating prediction accuracy evaluation parameter determination coefficient R2If the weight value meets the set value, extracting the trained weight value and threshold value to obtain a GA-BP neural network model; if not, returning to S3 to train again until determining the coefficient R2The set value is met;
prediction accuracy evaluation parameter determination coefficient R2The calculation formula is as follows:
in the formula (I), the compound is shown in the specification,in order to predict the value of the model,is the sample average value, yoIs the sample data value, v is the number of training samples;
and S5, testing the established GA-BP neural network by using the test sample set, and evaluating the network performance and the prediction accuracy.
2. A GA-BP neural network-based gear contact fatigue life prediction method according to claim 1, wherein in S1, the input parameters are tooth number, roughness, contact stress, tooth surface hardness and material parameters.
3. The GA-BP neural network-based gear contact fatigue life prediction method of claim 2, wherein in S2, the BP neural network is constructed such that the number of nodes of the input layer is 5, the number of hidden layers is 2, the number of nodes of the output layer is 1, and the number of nodes of hidden layers is 8.
4. A method for predicting gear contact fatigue life based on GA-BP neural network as claimed in claim 1,2 or 3, wherein in S4, the step of training BP neural network with the training set in S1 is:
step 1), carrying out forward propagation on a training sample, and calculating the square sum of the predicted output of an output layer and an expected output error;
randomly selecting the u-th input sample and the corresponding output test data, wherein the sum of the square of the error between the predicted output and the expected output is represented by the formula:
E(u)=(d(u)-y(u))2
wherein E (u) is the sum of the predicted output and the desired output error squared for the u-th input sample, d (u) is the desired output for the u-th input sample, and y (u) is the predicted output for the u-th input sample;
step 2), judging whether the sum of the square of the error between the predicted output and the expected output meets the requirement of being smaller than the minimum error of the training target or not; if not, entering the step 3) to carry out a back propagation process, if so, finishing the training of the BP neural network model, and ending;
step 3), performing a back propagation process by adopting a gradient descent method, and obtaining a weight updating formula from the second hidden layer to the output layer, a weight updating formula between the input layer and the first hidden layer, and a weight updating formula between the first hidden layer and the second hidden layer by utilizing the composite derivation of the predicted output and an expected output error square sum function E (u) degree weight:
wherein, the weight ω between the input layer and the first hidden layerijThe update formula is:
weight ω between first hidden layer and second hidden layerjkThe update formula is:
weight ω from the second hidden layer to the output layerkThe update formula is:
wherein η is the learning rate;
step 4), inputting the updated output layer weight obtained in the step 3), the weight between the input layer and the first hidden layer, the weight between the first hidden layer and the second hidden layer into the input layer and the hidden layer in the forward propagation in the step 1) to continue the forward propagation, and returning to the step 1).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010197216.7A CN111324989B (en) | 2020-03-19 | 2020-03-19 | Gear contact fatigue life prediction method based on GA-BP neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010197216.7A CN111324989B (en) | 2020-03-19 | 2020-03-19 | Gear contact fatigue life prediction method based on GA-BP neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111324989A true CN111324989A (en) | 2020-06-23 |
CN111324989B CN111324989B (en) | 2024-01-30 |
Family
ID=71164044
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010197216.7A Active CN111324989B (en) | 2020-03-19 | 2020-03-19 | Gear contact fatigue life prediction method based on GA-BP neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111324989B (en) |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111709942A (en) * | 2020-06-29 | 2020-09-25 | 中南大学 | Zinc flotation dosing amount prediction control method based on texture degree optimization |
CN111814401A (en) * | 2020-07-08 | 2020-10-23 | 重庆大学 | LED life prediction method of BP neural network based on genetic algorithm |
CN111950099A (en) * | 2020-08-03 | 2020-11-17 | 中国石油大学(华东) | Method, system, medium and computer equipment for testing mechanical property of equipment material |
CN112035779A (en) * | 2020-09-03 | 2020-12-04 | 郑州机械研究所有限公司 | Method for judging residual life of gear |
CN112149822A (en) * | 2020-07-21 | 2020-12-29 | 吉林建筑大学 | Drinking water disinfection byproduct prediction method and system |
CN112149849A (en) * | 2020-07-21 | 2020-12-29 | 吉林建筑大学 | Method for predicting disinfection by-products of drinking water based on multiple linear regression method |
CN112308228A (en) * | 2020-11-12 | 2021-02-02 | 安徽江机重型数控机床股份有限公司 | Fault detection early warning method for cutting fluid pumping system of numerical control machine tool |
CN112464554A (en) * | 2020-11-03 | 2021-03-09 | 桂林理工大学 | Operating parameter optimization method of gasoline refining equipment |
CN112623027A (en) * | 2020-12-26 | 2021-04-09 | 浙江天行健智能科技有限公司 | Road feel simulation method based on BP neural network |
CN112651164A (en) * | 2021-01-20 | 2021-04-13 | 华东理工大学 | Creep fatigue life prediction method based on machine learning |
CN112651176A (en) * | 2020-12-24 | 2021-04-13 | 广西大学 | GA-BP algorithm-based shielded door motor output force prediction method |
CN112685834A (en) * | 2020-12-31 | 2021-04-20 | 合肥工业大学智能制造技术研究院 | Collision energy absorption prediction method, medium and terminal for vehicle body front end structural component |
CN112926163A (en) * | 2021-04-07 | 2021-06-08 | 唐山学院 | Method and device for constructing bolt fatigue life prediction model based on Workbench and Matlab |
CN113011463A (en) * | 2021-02-25 | 2021-06-22 | 上海工程技术大学 | Chinese herbal medicine equipment bearing service life prediction method based on improved GA-BP |
CN113063589A (en) * | 2021-03-18 | 2021-07-02 | 重庆青山工业有限责任公司 | Gear microscopic error vibration prediction method based on neural network |
CN113239599A (en) * | 2021-06-15 | 2021-08-10 | 江苏理工学院 | Intelligent tire wear life estimation method and device based on BP neural network |
CN113640712A (en) * | 2021-08-06 | 2021-11-12 | 中国人民解放军海军工程大学 | Prediction method for vertical component of vertical induced magnetic field of ship |
CN114638164A (en) * | 2022-03-21 | 2022-06-17 | 西安热工研究院有限公司 | Method for predicting high-temperature creep life of pressure pipeline of power station |
CN114997692A (en) * | 2022-06-17 | 2022-09-02 | 哈尔滨工业大学 | Rotary table rotation reference angular pendulum error evaluation method based on BP neural network |
CN115293054A (en) * | 2022-10-08 | 2022-11-04 | 西南石油大学 | Neural network-based dented pipeline failure evaluation method |
CN115879245A (en) * | 2023-02-21 | 2023-03-31 | 南通市中吕齿轮有限公司 | Method for evaluating long-life transmission fatigue reliability of medium carbon steel gear |
CN117669390A (en) * | 2024-02-01 | 2024-03-08 | 中国石油大学(华东) | Metal full-stage fatigue crack growth prediction method and system based on neural network |
Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102693450A (en) * | 2012-05-16 | 2012-09-26 | 北京理工大学 | A prediction method for crankshaft fatigue life based on genetic nerve network |
CN103077267A (en) * | 2012-12-28 | 2013-05-01 | 电子科技大学 | Parameter sound source modeling method based on improved BP (Back Propagation) neural network |
CN105913150A (en) * | 2016-04-12 | 2016-08-31 | 河海大学常州校区 | BP neural network photovoltaic power station generating capacity prediction method based on genetic algorithm |
CN106886663A (en) * | 2017-03-29 | 2017-06-23 | 北京理工大学 | Tooth bending Prediction method for fatigue life and device |
CN106951983A (en) * | 2017-02-27 | 2017-07-14 | 浙江工业大学 | Injector performance Forecasting Methodology based on the artificial neural network using many parent genetic algorithms |
CN107942290A (en) * | 2017-11-16 | 2018-04-20 | 东南大学 | Binaural sound sources localization method based on BP neural network |
CN108318244A (en) * | 2018-01-23 | 2018-07-24 | 重庆大学 | Consider the comentation hardening Gear Contact fatigue methods of risk assessment of residual stress |
CN108362510A (en) * | 2017-11-30 | 2018-08-03 | 中国航空综合技术研究所 | A kind of engineering goods method of fault pattern recognition based on evidence neural network model |
CN108460461A (en) * | 2018-02-06 | 2018-08-28 | 吉林大学 | Mars earth shear parameters prediction technique based on GA-BP neural networks |
CN108764473A (en) * | 2018-05-23 | 2018-11-06 | 河北工程大学 | A kind of BP neural network water demands forecasting method based on correlation analysis |
US20190147369A1 (en) * | 2017-11-14 | 2019-05-16 | Adobe Inc. | Rule Determination for Black-Box Machine-Learning Models |
US20190242936A1 (en) * | 2018-02-05 | 2019-08-08 | Wuhan University | Fault diagnosis method for series hybrid electric vehicle ac/dc converter |
CN110147624A (en) * | 2019-05-24 | 2019-08-20 | 重庆大学 | A kind of Gear Contact Prediction method for fatigue life based on loading spectrum |
CN110232444A (en) * | 2019-06-17 | 2019-09-13 | 武汉轻工大学 | Optimization method, device, equipment and the storage medium of geology monitoring BP neural network |
CN110363344A (en) * | 2019-07-11 | 2019-10-22 | 安徽理工大学 | Probability integral parameter prediction method based on MIV-GP algorithm optimization BP neural network |
-
2020
- 2020-03-19 CN CN202010197216.7A patent/CN111324989B/en active Active
Patent Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102693450A (en) * | 2012-05-16 | 2012-09-26 | 北京理工大学 | A prediction method for crankshaft fatigue life based on genetic nerve network |
CN103077267A (en) * | 2012-12-28 | 2013-05-01 | 电子科技大学 | Parameter sound source modeling method based on improved BP (Back Propagation) neural network |
CN105913150A (en) * | 2016-04-12 | 2016-08-31 | 河海大学常州校区 | BP neural network photovoltaic power station generating capacity prediction method based on genetic algorithm |
CN106951983A (en) * | 2017-02-27 | 2017-07-14 | 浙江工业大学 | Injector performance Forecasting Methodology based on the artificial neural network using many parent genetic algorithms |
CN106886663A (en) * | 2017-03-29 | 2017-06-23 | 北京理工大学 | Tooth bending Prediction method for fatigue life and device |
US20190147369A1 (en) * | 2017-11-14 | 2019-05-16 | Adobe Inc. | Rule Determination for Black-Box Machine-Learning Models |
CN107942290A (en) * | 2017-11-16 | 2018-04-20 | 东南大学 | Binaural sound sources localization method based on BP neural network |
CN108362510A (en) * | 2017-11-30 | 2018-08-03 | 中国航空综合技术研究所 | A kind of engineering goods method of fault pattern recognition based on evidence neural network model |
CN108318244A (en) * | 2018-01-23 | 2018-07-24 | 重庆大学 | Consider the comentation hardening Gear Contact fatigue methods of risk assessment of residual stress |
US20190242936A1 (en) * | 2018-02-05 | 2019-08-08 | Wuhan University | Fault diagnosis method for series hybrid electric vehicle ac/dc converter |
CN108460461A (en) * | 2018-02-06 | 2018-08-28 | 吉林大学 | Mars earth shear parameters prediction technique based on GA-BP neural networks |
CN108764473A (en) * | 2018-05-23 | 2018-11-06 | 河北工程大学 | A kind of BP neural network water demands forecasting method based on correlation analysis |
CN110147624A (en) * | 2019-05-24 | 2019-08-20 | 重庆大学 | A kind of Gear Contact Prediction method for fatigue life based on loading spectrum |
CN110232444A (en) * | 2019-06-17 | 2019-09-13 | 武汉轻工大学 | Optimization method, device, equipment and the storage medium of geology monitoring BP neural network |
CN110363344A (en) * | 2019-07-11 | 2019-10-22 | 安徽理工大学 | Probability integral parameter prediction method based on MIV-GP algorithm optimization BP neural network |
Non-Patent Citations (13)
Title |
---|
余发山;康洪;: "基于GA优化BP神经网络的液压钻机故障诊断", 电子测量技术, no. 02, pages 134 - 137 * |
刘春艳;凌建春;寇林元;仇丽霞;武俊青;: "GA-BP神经网络与BP神经网络性能比较", 中国卫生统计, no. 02, pages 173 - 176 * |
吴志杰;孔凡敏;李康;: "基于遗传算法的BP神经网络的LED寿命预测模型", 半导体技术, no. 05, pages 375 - 380 * |
安艳秋, 陈举华, 张洪才: "基于进化神经网络的齿轮可靠性预测", 山东大学学报(工学版), no. 03, pages 227 - 231 * |
张明月;王新云;夏巨谌;纪刚;: "基于BP神经网络和遗传算法的齿轮坯预锻件多目标优化设计", 锻压技术, no. 06, pages 22 - 26 * |
张细政;郑亮;刘志华;: "基于遗传算法优化BP神经网络的风机齿轮箱故障诊断", 湖南工程学院学报(自然科学版), no. 03, pages 1 - 6 * |
李强;杨天邦;涂公平;: "GA-BP神经网络模型应用于岩芯扫描仪测定海洋沉积物中多种组分的半定量分析", 分析仪器, no. 01, pages 75 - 79 * |
李战芬;韩意;刘彦臣;樊孝仁;: "基于神经网络遗传算法优化的曲轴疲劳寿命预测", 中北大学学报(自然科学版), no. 04, pages 401 - 406 * |
李春生;李霄野;张可佳;: "基于遗传算法改进的BP神经网络房价预测分析", 计算机技术与发展, no. 08, pages 144 - 147 * |
皮骏;马圣;贺嘉诚;孔庆国;马龙;: "遗传算法优化的SVM在航空发动机磨损故障诊断中的应用", 润滑与密封, no. 10, pages 89 - 97 * |
范晓东;邱波;刘园园;魏诗雅;段福庆;: "一种基于遗传优化的BP神经网络的测光红移估计算法", 光谱学与光谱分析, no. 08, pages 2374 - 2378 * |
赵东波;陆金桂;姚灵灵;: "液压支架顶梁疲劳寿命的改进神经网络预测", 矿业研究与开发, no. 12, pages 106 - 109 * |
马林茂;李德富;郭海湘;李伟伟;: "基于遗传算法优化BP神经网络在原油产量预测中的应用:以大庆油田BED试验区为例", 数学的实践与认识, no. 24, pages 117 - 128 * |
Cited By (30)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111709942A (en) * | 2020-06-29 | 2020-09-25 | 中南大学 | Zinc flotation dosing amount prediction control method based on texture degree optimization |
CN111814401A (en) * | 2020-07-08 | 2020-10-23 | 重庆大学 | LED life prediction method of BP neural network based on genetic algorithm |
CN111814401B (en) * | 2020-07-08 | 2023-10-27 | 重庆大学 | LED life prediction method of BP neural network based on genetic algorithm |
CN112149822A (en) * | 2020-07-21 | 2020-12-29 | 吉林建筑大学 | Drinking water disinfection byproduct prediction method and system |
CN112149849A (en) * | 2020-07-21 | 2020-12-29 | 吉林建筑大学 | Method for predicting disinfection by-products of drinking water based on multiple linear regression method |
CN112149849B (en) * | 2020-07-21 | 2024-05-14 | 吉林建筑大学 | Method for predicting disinfection byproducts of drinking water based on multiple linear regression method |
CN111950099A (en) * | 2020-08-03 | 2020-11-17 | 中国石油大学(华东) | Method, system, medium and computer equipment for testing mechanical property of equipment material |
CN112035779B (en) * | 2020-09-03 | 2021-06-22 | 郑州机械研究所有限公司 | Method for judging residual life of gear |
CN112035779A (en) * | 2020-09-03 | 2020-12-04 | 郑州机械研究所有限公司 | Method for judging residual life of gear |
CN112464554A (en) * | 2020-11-03 | 2021-03-09 | 桂林理工大学 | Operating parameter optimization method of gasoline refining equipment |
CN112308228A (en) * | 2020-11-12 | 2021-02-02 | 安徽江机重型数控机床股份有限公司 | Fault detection early warning method for cutting fluid pumping system of numerical control machine tool |
CN112651176A (en) * | 2020-12-24 | 2021-04-13 | 广西大学 | GA-BP algorithm-based shielded door motor output force prediction method |
CN112623027A (en) * | 2020-12-26 | 2021-04-09 | 浙江天行健智能科技有限公司 | Road feel simulation method based on BP neural network |
CN112685834A (en) * | 2020-12-31 | 2021-04-20 | 合肥工业大学智能制造技术研究院 | Collision energy absorption prediction method, medium and terminal for vehicle body front end structural component |
CN112651164A (en) * | 2021-01-20 | 2021-04-13 | 华东理工大学 | Creep fatigue life prediction method based on machine learning |
CN113011463B (en) * | 2021-02-25 | 2022-09-06 | 上海工程技术大学 | Improved GA-BP-based Chinese herbal medicine equipment bearing life prediction method |
CN113011463A (en) * | 2021-02-25 | 2021-06-22 | 上海工程技术大学 | Chinese herbal medicine equipment bearing service life prediction method based on improved GA-BP |
CN113063589A (en) * | 2021-03-18 | 2021-07-02 | 重庆青山工业有限责任公司 | Gear microscopic error vibration prediction method based on neural network |
CN112926163B (en) * | 2021-04-07 | 2022-04-29 | 唐山学院 | Method and device for constructing bolt fatigue life prediction model based on Workbench and Matlab |
CN112926163A (en) * | 2021-04-07 | 2021-06-08 | 唐山学院 | Method and device for constructing bolt fatigue life prediction model based on Workbench and Matlab |
CN113239599A (en) * | 2021-06-15 | 2021-08-10 | 江苏理工学院 | Intelligent tire wear life estimation method and device based on BP neural network |
CN113640712A (en) * | 2021-08-06 | 2021-11-12 | 中国人民解放军海军工程大学 | Prediction method for vertical component of vertical induced magnetic field of ship |
CN113640712B (en) * | 2021-08-06 | 2023-11-24 | 中国人民解放军海军工程大学 | Prediction method for vertical component of vertical induction magnetic field of ship |
CN114638164A (en) * | 2022-03-21 | 2022-06-17 | 西安热工研究院有限公司 | Method for predicting high-temperature creep life of pressure pipeline of power station |
CN114638164B (en) * | 2022-03-21 | 2024-03-15 | 西安热工研究院有限公司 | Method for predicting high-temperature creep life of power station pressure pipeline |
CN114997692A (en) * | 2022-06-17 | 2022-09-02 | 哈尔滨工业大学 | Rotary table rotation reference angular pendulum error evaluation method based on BP neural network |
CN115293054A (en) * | 2022-10-08 | 2022-11-04 | 西南石油大学 | Neural network-based dented pipeline failure evaluation method |
CN115293054B (en) * | 2022-10-08 | 2022-12-13 | 西南石油大学 | Neural network-based dented pipeline failure evaluation method |
CN115879245A (en) * | 2023-02-21 | 2023-03-31 | 南通市中吕齿轮有限公司 | Method for evaluating long-life transmission fatigue reliability of medium carbon steel gear |
CN117669390A (en) * | 2024-02-01 | 2024-03-08 | 中国石油大学(华东) | Metal full-stage fatigue crack growth prediction method and system based on neural network |
Also Published As
Publication number | Publication date |
---|---|
CN111324989B (en) | 2024-01-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111324989B (en) | Gear contact fatigue life prediction method based on GA-BP neural network | |
CN108344564B (en) | A kind of state recognition of main shaft features Testbed and prediction technique based on deep learning | |
CN112116153B (en) | Park multivariate load joint prediction method coupling Copula and stacked LSTM network | |
WO2018053935A1 (en) | Failure mode occurrence probability based operating status fuzzy evaluation and prediction method for rotating device | |
CN109766745B (en) | Reinforced learning tri-state combined long-time and short-time memory neural network system and training and predicting method | |
CN110544011A (en) | Intelligent system combat effectiveness evaluation and optimization method | |
CN105868863B (en) | Electromechanical system multi-level reliability prediction method based on fuzzy judgment | |
CN109165421B (en) | Ship shafting bearing load value prediction method based on genetic algorithm optimization BP neural network | |
CN113033105B (en) | Method for predicting fatigue performance of steel material based on mechanical theory guiding transfer learning | |
CN112364560A (en) | Intelligent prediction method for working hours of mine rock drilling equipment | |
CN106096723A (en) | A kind of based on hybrid neural networks algorithm for complex industrial properties of product appraisal procedure | |
CN115221793A (en) | Tunnel surrounding rock deformation prediction method and device | |
CN114021432A (en) | Stress corrosion cracking crack propagation rate prediction method and system | |
CN109840335A (en) | Based on the radial forging pit prediction optimization method for strengthening T-S fuzzy neural network | |
CN114707692A (en) | Wetland effluent ammonia nitrogen concentration prediction method and system based on hybrid neural network | |
CN114547795A (en) | Rolling bearing residual life prediction method based on data driving | |
CN110633868A (en) | Method for predicting properties of oil and gas in exploration well oil testing layer by optimizing neural network through genetic algorithm | |
CN112767692A (en) | Short-term traffic flow prediction system based on SARIMA-GA-Elman combined model | |
CN110991616B (en) | Method for predicting BOD of effluent based on pruning feedforward small-world neural network | |
CN115983831A (en) | Operation and maintenance method for rolling bearing set of train-level traction motor | |
CN115618751A (en) | Steel plate mechanical property prediction method | |
CN112733457B (en) | Load electricity utilization characteristic modeling method for improving double-layer long-short term memory network | |
CN112365022A (en) | Engine bearing fault prediction method based on multiple stages | |
CN117522173B (en) | Natural gas hydrate depressurization exploitation productivity prediction method based on deep neural network | |
CN113779852A (en) | Method for quantitatively evaluating influence of random parameter disturbance on product fatigue life distribution |
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 |