A method for evaluating corrosion and residual life of high-temperature heating surface of ultra-supercritical unit

Abstract

The application discloses a kind of ultra-supercritical unit high temperature heating surface corrosion and residual life evaluation method, comprising the following steps: one, obtains the material parameters and working condition parameters of ultra-supercritical unit high temperature heating surface;Two, the material parameters and working condition parameters of ultra-supercritical unit high temperature heating surface are brought into the corrosion amount prediction neural network model constructed in advance, and corrosion amount prediction value is obtained;Three, corrosion rate is calculated according to corrosion amount prediction value;Four, residual life prediction value is calculated according to corrosion rate.The application predicts corrosion amount using BP neural network, can conveniently and accurately predict the corrosion amount of metal material, and calculates the residual life of high temperature heating surface corrosion by predicted corrosion amount, can understand the entire life cycle of high temperature heating surface, provides important reference and technical guidance significance for the daily operation, overhaul and maintenance and safety life evaluation of large thermal power generating unit, and has engineering practical application value.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of corrosion amount and corrosion life prediction, and particularly relates to a method for evaluating corrosion and residual life of high-temperature heating surface of an ultra-supercritical unit. BACKGROUND

[0002] With the demand for energy saving and emission reduction in recent years, the efficiency of power generation boilers is gradually improved, and the thermal power generating unit has developed to an ultra-supercritical unit. The latest generation of ultra-supercritical thermal power generating unit can greatly improve the boiler power generation efficiency, reduce the primary energy consumption, and reduce the environmental pollution level, and has become a key link for upgrading the thermal power generating unit and an important content of the national energy strategy. The power generation technology of the ultra-supercritical unit increases the steam pressure and temperature of water to above the critical parameters, and the water directly changes from liquid state to vapor state in a single phase, so that the superheater pipe of the unit is in a very harsh high-temperature and high-pressure environment for a long time. The improvement of the unit parameters puts forward higher requirements for the superheater pipe, and once the failure of the high-temperature pressure-bearing component occurs, it will affect the safe operation of the unit and cause serious personal and property losses. Therefore, the ultra-supercritical unit puts forward strict requirements on the safe operation performance of the high-temperature heating surface.

[0003] In the high-temperature heating surface, the metal material is directly in contact with the supercritical working fluid with special properties, and is affected by the complex mechanisms such as oxidation of the supercritical fluid and synergistic corrosion of impurities and the working fluid, which may cause the material to be thinned due to corrosion, the mechanical stress and thermal stress to be weakened, and the leakage and burst pipe to occur at the supercritical temperature and pressure, which endangers the safety of the system, and brings severe challenges to the stable, economic and safe operation of the system. Therefore, the corrosion behavior of the alloy under the supercritical working condition becomes an important problem that cannot be ignored. Moreover, the corrosion amount prediction and residual life evaluation of the high-temperature heating surface can understand the entire life cycle of the high-temperature heating surface, and provide important reference and technical guidance for the daily operation, maintenance and safety life evaluation of the large thermal power generating unit, and has engineering practical application value.

[0004] In the prior art, the multi-linear regression analysis method and the statistical prediction method are the most common methods for predicting the corrosion amount of the material. However, the multi-linear regression analysis method ignores the influence of the interaction between various factors on the corrosion amount, and the expression is not accurate. The statistical prediction method needs to find the influence law of various factors on the corrosion amount, and the data volume is large, so it is difficult to accurately predict the corrosion amount of the metal material. Moreover, the calculation of the corrosion and residual life evaluation of the high-temperature heating surface of the ultra-supercritical unit in the prior art is large, the efficiency is low, and the accuracy is poor, and the problems still need to be solved. SUMMARY

[0005] The technical problems to be solved by the present application are to provide a method for evaluating corrosion and residual life of high-temperature heating surface of an ultra-supercritical unit, which uses a BP neural network method to predict the corrosion amount, can conveniently and accurately predict the corrosion amount of metal materials, and calculates the residual life of high-temperature heating surface corrosion through the predicted corrosion amount, so as to understand the entire life cycle of the high-temperature heating surface, and provide important reference and technical guidance for the daily operation, maintenance and safety life evaluation of large thermal power generating units, and has engineering practical application value.

[0006] To solve the above technical problems, the technical scheme adopted by the present application is as follows: a method for evaluating corrosion and residual life of high-temperature heating surface of an ultra-supercritical unit, which comprises the following steps:

[0007] Step one, obtaining material parameters and working condition parameters of the high-temperature heating surface of the ultra-supercritical unit;

[0008] Step two, inputting the material parameters and working condition parameters of the high-temperature heating surface of the ultra-supercritical unit into a pre-constructed corrosion amount prediction neural network model to obtain a corrosion amount prediction value;

[0009] Step three, calculating a corrosion rate according to the corrosion amount prediction value;

[0010] Step four, calculating a residual life prediction value according to the corrosion rate.

[0011] The method for evaluating corrosion and residual life of high-temperature heating surface of an ultra-supercritical unit, wherein the material parameters and working condition parameters include operating time t, temperature T, pressure P, fluid flow rate τ and material-related parameters M, and the material-related parameters M include the content of corrosion-resistant elements in the metal material.

[0012] The method for evaluating corrosion and residual life of high-temperature heating surface of an ultra-supercritical unit, wherein the construction process of the corrosion amount prediction neural network model in step two is as follows:

[0013] Step 201, constructing a material corrosion amount database of the high-temperature heating surface of the ultra-supercritical unit;

[0014] Step 202, constructing a corrosion amount prediction neural network model by using a BP algorithm;

[0015] Step 203, training the corrosion amount prediction neural network model by using the material corrosion amount database constructed in step 201;

[0016] Step 204, inputting a verification set to perfect the corrosion amount prediction neural network model, and obtaining a trained corrosion amount prediction neural network model.

[0017] The material corrosion amount database includes corrosion weight gain data obtained through experiments and corrosion weight gain data obtained from literature, and is expressed as: (x 1i ,..., x ni , y i ), wherein x 1i is the first parameter of the i-th sample; x ni is the n-th parameter of the i-th sample; n is the number of material parameters and working condition parameters of the sample; and y i is the corrosion amount of the i-th sample.

[0018] The corrosion amount prediction neural network model includes an input layer, a hidden layer and an output layer, the input layer is provided with n nodes corresponding to the number of material parameters and working condition parameters; the hidden layer is provided with m neurons; and the output layer is provided with one neuron corresponding to the material corrosion amount AW.

[0019] The activation function of the hidden layer is a tanh function, the activation function of the output layer is a purelin function, and the training algorithm of the corrosion amount prediction neural network model is a Polak-Ribiers conjugate gradient algorithm.

[0020] In step three, the corrosion rate is calculated according to the corrosion amount prediction value, and the calculation formula is: wherein C R is the corrosion rate, AW is the material corrosion amount, k is the material corrosion amount coefficient, p m is the density of the material, A is the area of the material exposed to the corrosion environment, and t is the corrosion time.

[0021] In the above method for evaluating the corrosion of the high-temperature heating surface of an ultra-supercritical unit and the residual life, the value of k is 8.76x10 4 .

[0022] In step four, the residual life prediction value is calculated according to the corrosion rate, and the calculation formula is: wherein t r is the residual life, d initial is the initial wall thickness, d required is the design wall thickness, C R is the corrosion rate, and t is the corrosion time.

[0023] The corrosion of the high-temperature heating surface of the ultra-supercritical unit and the residual life evaluation method has the design wall thickness δ required The calculation formula is Wherein, P is the design pressure, D O is the outer diameter of the high-temperature heating pipeline of the ultra-supercritical unit, E j is the welding joint coefficient, [σ] t is the allowable stress of the material at the design temperature, and Y is the design pressure coefficient.

[0024] Compared with the prior art, the present application has the following advantages:

[0025] 1. The corrosion amount prediction neural network model is constructed by using the BP algorithm, and the corrosion amount prediction value is obtained by using the corrosion amount prediction neural network model. Compared with the existing technology which uses a multiple linear regression analysis method to express the corrosion amount of the material, the corrosion amount of the metal material is simultaneously affected by multiple factors such as temperature, pressure, flow rate, material-related parameters, etc. in the metal material corrosion model under a supercritical environment, and more accurate prediction values can be obtained. Compared with the existing technology which uses a statistical prediction method to predict the corrosion amount of the material, the influence of various factors on the corrosion amount does not need to be found, and the corrosion amount of the metal material can be accurately predicted. The neural network model has good nonlinear characteristics, flexible and effective learning methods, and a completely distributed storage structure. A single neuron in the neural network has a self-organizing complex mode and reflects nonlinear characteristics, so that the neural network can reconstruct any nonlinear continuous function. Through learning, the network can obtain the internal law of the sequence, so that the change of the sequence can be predicted. The use of neural networks can avoid the conventional modeling process, and at the same time, it also shows good self-adaptation and self-learning ability, strong anti-interference ability. The neural network model can contain the nonlinear relationship between the corrosion influencing factors and the corrosion results in the neural network topology structure, avoiding the difficulty of finding the influence of various factors on the corrosion amount, and can conveniently and accurately predict the corrosion amount of the metal material.

[0026] 2. The present application overcomes the problem of large amount of calculation and ignores the influence of the interaction between various factors such as high-temperature heating surface working conditions and materials on the corrosion amount in the prior art.

[0027] 3. The present application uses the BP neural network to predict the corrosion amount, calculates the corrosion residual life of the high-temperature heating surface by the predicted corrosion amount, can understand the entire life cycle of the high-temperature heating surface, provides important reference and technical guidance for the daily operation, maintenance and safety life evaluation of large-scale thermal power generating units, and has engineering practical application value.

[0028] The technical solutions of the present application will be further described in detail below with reference to the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 A flow chart of the method of the present application;

[0030] Figure 2 A flow chart of the process of constructing the corrosion amount prediction neural network model of the present application;

[0031] Figure 3 A structural schematic diagram of the corrosion amount prediction neural network model constructed by the present application;

[0032] Figure 4 A comparison data graph of the predicted value and the experimental value in the embodiment of the present application. DETAILED DESCRIPTION

[0033] As Figure 1 shown, the ultra-supercritical unit high-temperature heating surface corrosion and residual life assessment method of the present application comprises the following steps:

[0034] Step one, obtaining the material parameters and working condition parameters of the ultra-supercritical unit high-temperature heating surface;

[0035] In this embodiment, the material parameters and working condition parameters include operating time t, temperature T, pressure P, fluid flow rate τ and material related parameters M, and the material related parameters M include the content of corrosion-resistant elements in the metal material. The above parameters are all main factors affecting corrosion.

[0036] Step two, bringing the material parameters and working condition parameters of the ultra-supercritical unit high-temperature heating surface into the pre-constructed corrosion amount prediction neural network model to obtain the corrosion amount prediction value;

[0037] In this embodiment, as Figure 2 shown, the construction process of the corrosion amount prediction neural network model in step two is as follows:

[0038] Step 201, constructing a material corrosion amount database of the ultra-supercritical unit high-temperature heating surface;

[0039] In this embodiment, the material corrosion amount database includes corrosion weight gain data obtained by experiments and corrosion weight gain data obtained from literature, and the material corrosion amount database is represented as: (x 1i ,..., x ni , y i ), wherein x 1i is the first parameter of the i th sample; x ni is the n th parameter of the i th sample; n is the number of material parameters and working condition parameters of the sample; and y i is the corrosion amount of the i th sample.

[0040] Step 202, constructing a corrosion amount prediction neural network model by using a BP algorithm;

[0041] In this embodiment, the corrosion prediction neural network model includes an input layer, a hidden layer, and an output layer. The input layer has n nodes, corresponding to the number of material parameters and operating condition parameters; the hidden layer has m neurons; and the output layer has 1 neuron, corresponding to the material corrosion amount ΔW.

[0042] In this embodiment, the activation function of the hidden layer is the tanh function, the activation function of the output layer is the purelin function, and the training algorithm of the erosion amount prediction neural network model is the Polak-Ribiers conjugate gradient algorithm.

[0043] Step 203: Train a neural network model for predicting corrosion using the material corrosion database constructed in Step 201;

[0044] Step 204: Substitute the validation set into the model to improve the corrosion amount prediction neural network model and obtain the trained corrosion amount prediction neural network model.

[0045] Step 3: Calculate the corrosion rate based on the predicted corrosion amount;

[0046] In this embodiment, the calculation formula used in step three to calculate the corrosion rate based on the predicted corrosion amount is as follows: Among them, C R Let ρ be the corrosion rate, ΔW be the material corrosion amount, k be the material corrosion coefficient, and ρ be the corrosion rate. m Let A be the density of the material, A be the area of ​​the material exposed to the corrosive environment, and t be the corrosion time.

[0047] In this embodiment, the value of k is 8.76 × 10 4 .

[0048] Step 4: Calculate the predicted remaining life based on the corrosion rate.

[0049] In this embodiment, the calculation formula used in step four to calculate the predicted remaining lifetime based on the corrosion rate is as follows: Among them, t r For the remaining lifetime, δ initial δ is the initial wall thickness. required To design the wall thickness, C R Let t be the corrosion rate and t be the corrosion time.

[0050] In this embodiment, the designed wall thickness δ req The calculation formula is Where P is the design pressure, and D is the design pressure. O E represents the outer diameter of the high-temperature heating pipes in ultra-supercritical units. j [σ] is the weld joint coefficient. tY is the design pressure coefficient.

[0051] In the implementation, for seamless steel pipes, the welding joint coefficient E j is 1; for ferrite, austenite or other ductile metals, the design pressure coefficient Y is constant 0.4 when the temperature is lower than 482℃.

[0052] In order to verify the technical effects generated by the present application, the method of the present application is experimentally verified.

[0053] In the experiment, 33 groups of corrosion weight gain data obtained by the experiment and 76 groups of corrosion weight gain data obtained through the literature are divided into 80% training data set and 20% test data set, and the experimental data summary table is shown in Table 1, and the literature data summary table is shown in Table 2.

[0054] The training data set is used for network training, and the test data set is used for verifying the prediction performance of the network while training, and the control condition of the verification error is used to determine when to end the training.

[0055] Table 1 Experimental data summary table

[0056]

[0057]

[0058] Table 2 Literature data summary table

[0059]

[0060]

[0061]

[0062] The time t, temperature T, pressure P, fluid flow rate τ, Cr content, Ni content and Mo content are used as network input variables, and the corrosion amount of the metal material is used as network output, and the structure diagram of the corrosion amount prediction neural network model is shown in Figure 3 .

[0063] The tanh function is selected as the activation function of the hidden layer, the purelin function is selected as the activation function of the output layer, the Polak-Ribiers conjugate gradient algorithm is selected as the training algorithm, p is the input vector, and t is the target vector. The main program is as follows:

[0064] net = newff(minmax(p), [7, 1], {‘tanh’, ‘purelin’}, ‘traincgp’);

[0065] net.trainParam.goal = 0.0001 % training goal

[0066] net.trainParam.epochs = 1000; % maximum number of epochs

[0067] net.trainParam.show = 20; % display frequency

[0068] net.trainParam.lr = 0.001; % learning rate

[0069] net.trainParam.min_fail = 1; % maximum number of failures to be tolerated

[0070] [net,tr] = train(net,p,t)

[0071] The main factors causing corrosion of metal materials are seven (time t, temperature T, pressure P, fluid flow rate τ, Cr content, Ni content, and Mo content), i.e., the input layer node is 7, and the output value is the material corrosion amount, i.e., the output node is 1. The number of hidden layer nodes is where n0 is the number of output layer nodes, n i is the number of input layer nodes, and a is a constant between 1 and 10. Then the number of hidden layer nodes is a constant between 4 and 14.

[0072] After multiple attempts at calculation, when the number of hidden layer nodes is 7, the relative error of the prediction result is the smallest, and the comparison data between the predicted value and the experimental value are as Figure 4 shown. The fitting curve obtained is y = 0.02 + 9.95x. At the same time, Table 3 gives the correlation test results of the predicted value and the experimental value, and it can be seen that the Pearson correlation coefficient is 0.997, and the correlation is significant.

[0073] Table 3 Correlation test table of predicted value and experimental value

[0074]

[0075] It can be seen that taking time t, temperature T, pressure P, fluid flow rate τ, Cr content, Ni content, and Mo content as input variables and taking the corrosion amount of metal materials as output variables can well describe the corrosion amount of metal materials under different working conditions.

[0076] The calculation result shows that the corrosion amount prediction neural network model constructed by the application can preferably reflect the nonlinear relationship between the corrosion amount and the influence factors, the deviation between the prediction value and the experimental value is small; the experimental data can be effectively learned and modeled and predicted, and especially for the complex process of metal material corrosion involving multiple factors, the neural network method shows obvious superiority.

[0077] The above is only the preferred embodiment of the application, and does not limit the application, and any simple modification, change and equivalent structure change of the above embodiment according to the technical essence of the application are still within the protection scope of the technical solution of the application.

Claims

1. An ultra-supercritical unit high-temperature heating surface corrosion and residual life evaluation method, characterized in that, The method comprises the following steps: Step one, obtaining material parameters and working condition parameters of the high-temperature heating surface of the ultra-supercritical unit; Step two, inputting the material parameters and working condition parameters of the high-temperature heating surface of the ultra-supercritical unit into a pre-constructed corrosion amount prediction neural network model to obtain a corrosion amount prediction value; Step three, calculating a corrosion rate according to the corrosion amount prediction value; Step four, calculating a residual life prediction value according to the corrosion rate; The material parameters and working condition parameters comprise an operation time t, a temperature T, a pressure P, a fluid flow rate τ and material-related parameters M, and the material-related parameters M comprise the content of corrosion-resistant elements in the metal material; The construction process of the corrosion amount prediction neural network model in step two is as follows: Step 201, constructing a material corrosion amount database of the high-temperature heating surface of the ultra-supercritical unit; Step 202, constructing a corrosion amount prediction neural network model by using a BP algorithm; Step 203, training the corrosion amount prediction neural network model by using the material corrosion amount database constructed in step 201; Step 204, inputting a verification set to perfect the corrosion amount prediction neural network model, and obtaining a trained corrosion amount prediction neural network model; Wherein, the corrosion rate is calculated according to the corrosion amount prediction value in step three, and the calculation formula is , wherein, C R is the corrosion rate, AW is the material corrosion amount, k is the material corrosion amount coefficient, p m is the density of the material, A is the area of the material exposed to the corrosion environment, and t is the corrosion time.

2. The method for evaluating the corrosion and residual life of high-temperature heating surfaces of an ultra-supercritical unit according to claim 1, characterized in that: The material corrosion amount database includes the corrosion weight gain data obtained by experiments and the corrosion weight gain data obtained from literatures, and is expressed as: (x li ,..., x ni , y i ), wherein x li is the lth parameter of the ith sample; x ni is the nth parameter of the ith sample; n is the number of material parameters and working condition parameters of the sample; and y i is the corrosion amount of the ith sample.

3. The method for evaluating the corrosion and residual life of high-temperature heating surfaces of an ultra-supercritical unit according to claim 1, characterized in that: The corrosion amount prediction neural network model comprises an input layer, a hidden layer and an output layer, n nodes are arranged in the input layer, corresponding to the number of material parameters and working condition parameters; m neurons are arranged in the hidden layer; and one neuron is arranged in the output layer, corresponding to the material corrosion amount ΔW.

4. The method for evaluating the corrosion and residual life of high-temperature heating surfaces of an ultra-supercritical unit according to claim 3, characterized in that: The activation function of the hidden layer adopts a tanh function, the activation function of the output layer adopts a purelin function, and the training algorithm of the corrosion amount prediction neural network model adopts a Polak-Ribiers conjugate gradient algorithm.

5. The method for evaluating the corrosion and residual life of high-temperature heating surfaces of an ultra-supercritical unit according to claim 1, characterized in that: said k has a value of 8.76 x 10 4 .

6. The method for evaluating the corrosion and residual life of high-temperature heating surfaces of an ultra-supercritical unit according to claim 1, characterized in that: The calculation formula used in step four for calculating the remaining life prediction value according to the corrosion rate is , wherein t r is the remaining life, δ initial is the initial wall thickness, δ required is the design wall thickness, C R is the corrosion rate, and t is the corrosion time.

7. The method for evaluating the corrosion and residual life of high-temperature heating surfaces of an ultra-supercritical unit according to claim 6, characterized in that: The design wall thickness δ required The calculation formula is Wherein, P is the design pressure, D O is the outer diameter of the high-temperature heating pipe of the ultra-supercritical unit, E j is the welding joint coefficient, [σ] t is the allowable stress of the material at the design temperature, and Y is the design pressure coefficient.

Images