A fault diagnosis system and method for steam turbine based on big data analysis
Through big data analysis and SVM model, the failure of steam turbine blade coating is predicted, and the problem of difficulty in preventing and diagnosing the failure of blade water corrosion protective coating in the existing technology is solved, and more accurate fault diagnosis and maintenance decisions are achieved.
Patent Information
- Application Number
- CN202411200457.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-29
AI Technical Summary
The prior art is difficult to effectively prevent and diagnose the failure of the water-eroded protective coating of the turbine blades, resulting in unstable operation of the turbine and poses safety hazards.
Using a fault diagnosis system based on big data analysis, the SVM prediction model is constructed to predict future coating failures and calculate failure averages by collecting and processing the spacing data between the blade matrix and the coating in the pit of the final blade sample of the turbine. Rate, cluster and classify, generate repair or replacement solutions.
It improves the comprehensiveness and accuracy of fault diagnosis, simplifies the diagnosis process, can promptly predict the coating failure situation, and ensures the scientific and accurate maintenance period of the turbine blades.
Smart Images

Figure CN118709060B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of fault diagnosis and adjustment systems, and in particular, relates to a fault diagnosis system and method for a steam turbine based on big data analysis. Background Art
[0002] Chinese patent CN104049539B discloses a full-operating-condition simulation system for a steam turbine heat recovery system. It uses main steam flow, main steam temperature, hot reheat steam temperature, feed water pump pressure, condensate pump pressure, circulating water temperature and circulating water flow as input parameters to simulate the full-operating-condition operation status of the steam turbine heat recovery system; it provides support for the optimization design, condition monitoring, fault diagnosis and variable-operating-condition performance calculation of the steam turbine heat recovery system.
[0003] Steam turbines are vital equipment in the production process of the energy and power industry. Their operating stability and lifespan affect the efficiency and capacity of power generation. The initiation and expansion of the gap between the water erosion protection coating of the turbine blades and the substrate of the last-stage blade until the coating fails is a huge safety hazard to the stable operation of the steam turbine. Therefore, if the failure of the water erosion protection coating of the turbine blades cannot be prevented in advance, the steam turbine will not be able to operate normally. Summary of the invention
[0004] In view of the problems in the related technology, the present invention proposes a fault diagnosis system and method for a steam turbine based on big data analysis to overcome the above-mentioned technical problems existing in the existing related technology.
[0005] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:
[0006] The present invention is a fault diagnosis method for a steam turbine based on big data analysis, comprising the following steps:
[0007] S1, collecting the spacing data between the blade substrate and the coating in the pits of multiple different types of steam turbine last-stage blade samples to obtain a spacing data matrix set; calculating the mean of the data in the spacing data matrix set to obtain an average spacing data matrix;
[0008] S2, using the average spacing data matrix to build the final SVM prediction model;
[0009] S3, collecting the spacing data between the blade substrate and the coating in the pits of a plurality of turbine last-stage blade samples with different characteristics, and obtaining an average spacing data set matrix set; using the final SVM prediction model to predict the average spacing data set matrix set, and obtaining an average spacing future data set matrix set;
[0010] S4, calculating the average failure rate of a plurality of steam turbine last-stage blade samples with different characteristics according to the average spacing future data set matrix set and the average spacing data set matrix set, and obtaining the average failure rate data set matrix;
[0011] S5, performing a clustering operation on the failure average rate data set matrix to obtain a failure average rate data set classification matrix set; performing label generation on each failure average rate data set classification matrix in the failure average rate data set classification matrix set to obtain a rate category label data set matrix; using feature data of a plurality of steam turbine last stage blade samples with different features and the rate category label data set matrix to construct a final SVM classification model;
[0012] S6. Perform fault diagnosis on the last-stage blade of the steam turbine to be diagnosed according to the final SVM classification model, and generate a maintenance or replacement plan;
[0013] The types of turbine last-stage blades include straight blades with equal cross-section, straight blades with variable cross-section, twisted blades, rear-loaded blades, front-loaded blades, impulse blades, reaction blades, regulating-stage blades, last-stage blades, intermediate-stage blades, carbon steel blades, alloy steel blades, stainless steel blades, heat-resistant steel blades, ceramic blades, cast blades, welded blades, assembled blades, low-temperature blades, medium-temperature blades, high-temperature blades, single-stage blades, double-stage blades, multi-stage blades, etc.; the substrate and coating materials of different types of turbine last-stage blades are different, so that the spacing data between the blade substrate and the coating in the pit are different; therefore, by collecting the spacing data between the blade substrate and the coating in the pit of multiple different types of turbine last-stage blade samples, the richness of the data is improved; and by calculating the mean of the spacing data, the error and randomness of the measurement data are reduced, and the accuracy of the data is improved; a good data foundation is provided for the subsequent construction of an accurate prediction model; by constructing the final SVM prediction model , which provides a model basis for the subsequent prediction of spacing data in the future; by predicting the average spacing data in the future, the future coating failure can be predicted in time, so that corresponding measures can be taken in advance to prevent it, which improves the comprehensiveness of fault diagnosis; by calculating the average failure rate of multiple turbine last-stage blade samples with different characteristics, it provides a data basis for the subsequent establishment of the corresponding relationship between the characteristic data and the average failure rate of the turbine last-stage blade samples; clustering the average failure rate data set matrix, and then generating corresponding category labels for each clustering classification, which reduces the complexity of the data, and then reduces the difficulty of subsequent construction of the classification model, and at the same time provides label data for building the final SVM classification model; by constructing the final SVM classification model, the corresponding coating failure rate can be obtained based on the characteristic data of the last-stage blade of the turbine to be diagnosed, which simplifies the diagnosis process and improves the diagnosis efficiency.
[0014] Preferably, the characteristic data includes coating thickness data, coating elastic modulus data and substrate elastic modulus data;
[0015] Preferably, the S1 comprises the following steps:
[0016] S11. Collect multiple different types of steam turbine last-stage blade samples with pits to obtain a last-stage blade sample set. , Indicates the collected i Steam turbine last stage blade samples, Indicates the total number of samples of the last stage blades of the steam turbine collected;
[0017] S12. Setting the first historical time point set , Indicates the first set of historical time points k A historical point in time, represents the total number of historical time points in the first historical time point set; and then sets the last blade sample set Several pit points of each leaf sample in the matrix are obtained. c ;as follows,
[0018] ;
[0019] in, Represents the final leaf sample set Middle i The first j A pit point, Represents the final leaf sample set Middle i The total number of pits on each leaf sample;
[0020] The final leaf sample set Each last stage blade sample is set to the running state and is set at the first historical time point Each historical time point in the pit point matrix c The distance data between the blade substrate and the coating at each pit point is measured and collected to obtain a distance data matrix set. , Represents the first historical time point set Middle k The spacing data matrix of historical time points is as follows:
[0021] ;
[0022] in, express No. iThe first j The distance data between the blade substrate and the coating at each pit point;
[0023] Calculate the spacing data matrix set The average value of each row of data in each spacing data matrix is obtained to obtain the average spacing data matrix ;as follows,
[0024] ;
[0025] in, Represents the final leaf sample set Middle i The last stage blade samples of the steam turbine are collected at the first historical time point Middle k The average distance data between the blade substrate and the coating at each historical time point;
[0026] Since the coating failure site is usually at the blade pit and the distance between the coating and the blade substrate changes continuously over time during operation, the spacing data at the blade pit corresponding to each time point is collected by setting the first historical time point set, providing data support for the subsequent construction of a prediction model for predicting future spacing data.
[0027] Preferably, S2 comprises the following steps:
[0028] S21, setting a first training data ratio; and calculating an average spacing data matrix according to the first training data ratio. Perform data division to obtain an average spacing training data matrix and an average spacing test data matrix;
[0029] S22, constructing an initial SVM prediction model and setting a first training error threshold; inputting the average spacing training data matrix into the initial SVM prediction model for training operation; during the training process, when the training error is less than the first training error threshold, stopping the training to obtain a trained SVM prediction model; otherwise, continuing the training;
[0030] S23, set a goodness of fit threshold; input the average spacing test data matrix into the trained SVM prediction model for testing; when the test is completed, calculate the goodness of fit value; when the goodness of fit value is greater than or equal to the goodness of fit threshold, use the trained SVM prediction model as the final SVM prediction model; otherwise, return to S22 to continue training the trained SVM prediction model until the goodness of fit value is greater than or equal to the goodness of fit threshold, and obtain the final SVM prediction model;
[0031] The calculation formula of the goodness of fit value is as follows: ;
[0032] In the formula, d represents the goodness-of-fit value; Indicates the total number of fitted data; and represent the actual value and the fitted value respectively;
[0033] The accuracy of the prediction model is evaluated by calculating the goodness of fit of the trained SVM prediction model, so that the final SVM prediction model constructed has better prediction effect and ability for the spacing data; the accuracy of subsequent predictions of the spacing data is improved, thereby improving the accuracy of the entire fault diagnosis.
[0034] Preferably, S3 comprises the following steps:
[0035] S31. Set coating thickness data set , coating elastic modulus dataset And the matrix elastic modulus data set ; , , Respectively represent the set j Coating thickness data, i The coating elastic modulus data and k The matrix elastic modulus data, , , Respectively represent the total number of set coating thickness data, the total number of coating elastic modulus data and the total number of substrate elastic modulus data;
[0036] S32, according to the coating thickness data set , coating elastic modulus dataset And the matrix elastic modulus data set Collect multiple samples of the last stage blades of the steam turbine with pits to obtain the last stage blade sample set matrix ;as follows,
[0037] ;
[0038] in, The coating elastic modulus data is , coating thickness data is A sample set of steam turbine last stage blades; , express The elastic modulus data of the matrix is Samples of steam turbine last stage blades;
[0039] S33: Setting a second historical time point set , Indicates the first set of the second historical time points i A historical point in time, Indicates the total number of historical time points in the set second historical time point set;
[0040] According to the second historical time point set Collect the final blade sample set matrix The average spacing data between the blade substrate and the coating at the pit on each last-stage blade sample is obtained to obtain the average spacing data set matrix set , Indicates that the second historical time point is concentrated at the The average spacing dataset matrix of historical time points is as follows:
[0041] ;
[0042] in, express The elastic modulus data of the middle coating is , coating thickness data is The average distance data set between the blade substrate and the coating corresponding to the last stage blade sample set of the steam turbine; , express The elastic modulus data of the matrix is The average distance data between the blade substrate and the coating on the last stage blade samples of the steam turbine;
[0043] S34. Setting a future time point set , Indicates that the future time point set is the first i A future time point, Indicates the total number of future time points in the set future time point set;
[0044] According to the future time point set The average spacing dataset matrix is set Input into the final SVM prediction model for prediction operation to obtain the average spacing future data set matrix set , Indicates the future time point concentration The average spacing of future time points in the future data set matrix is as follows:
[0045] ;
[0046] in, express The elastic modulus data of the middle coating is , coating thickness data is The average distance between the blade substrate and the coating corresponding to the steam turbine last stage blade sample set is a future data set;
[0047] , express The elastic modulus data of the matrix is Future data on the average distance between the blade substrate and the coating on the last stage blade samples of steam turbines;
[0048] Collecting coating thickness data set, coating elastic modulus data set and substrate elastic modulus data set has a great influence on coating failure rate. Therefore, collecting the corresponding last-stage blade samples of the combination of coating thickness data set, coating elastic modulus data set and substrate elastic modulus data set makes the last-stage blade samples richer, and thus makes the corresponding spacing data obtained subsequently richer; by using the final SVM prediction model to predict the average spacing data set matrix set, the average spacing future data set matrix set is obtained; which provides a data basis for the subsequent calculation of the coating failure rate of each last-stage blade sample.
[0049] Preferably, S4 comprises the following steps:
[0050] S41, setting a coating failure spacing threshold; and storing the average spacing future data set matrix set The average distance between the blade substrate and the coating on each turbine last stage blade sample in the future data is compared with the coating failure distance threshold; as time increases, the average distance future data set matrix set is recorded The future data of the average spacing between the blade substrate and the coating that is greater than or equal to the coating failure spacing threshold on each turbine last-stage blade sample and the corresponding future time point are obtained to obtain the failure future time point set matrix and the future dataset matrix of mean interval between failures ; They are as follows,
[0051] ; ;
[0052] in, , Respectively represent the coating elastic modulus data as , coating thickness data is The failure future time point set and the failure average spacing future data set corresponding to the steam turbine last stage blade sample set; , , , Respectively The elastic modulus data of the matrix is The future failure time of the turbine last stage blade sample and The elastic modulus data of the matrix is Future data of mean distance to failure of turbine last stage blade samples;
[0053] S42, the failure average spacing future data set matrix The matrix set with the average spacing data set The average spacing dataset matrix of the first historical time point is subtracted to obtain the average spacing difference dataset matrix ;as follows,
[0054] ;
[0055] in, The coating elastic modulus data is , coating thickness data is The average spacing difference data set corresponding to the last stage blade sample set of the steam turbine, , express The elastic modulus data of the matrix is The average spacing difference data of the last stage blade samples of the steam turbine is calculated as follows: ;
[0056] According to the average spacing difference data set matrix , failure future time point set matrix And the second historical time point set Calculate the final blade sample set matrix The average failure rate data of each last-stage blade sample in the , and the average failure rate data set matrix is obtained ;as follows,
[0057] ;
[0058] in, The coating elastic modulus data is , coating thickness data is The average failure rate data set corresponding to the last stage blade sample set of the steam turbine; , express The elastic modulus data of the matrix is The average failure rate data of the last stage blade samples of the steam turbine is calculated as follows: ;
[0059] The coating failure rate is calculated by subtracting the spacing at the time of coating failure from the spacing at the initial moment and then dividing it by the time difference between the two. The coating failure rate is important for the maintenance period of the blades and provides a basis for subsequent diagnosis of turbine blades.
[0060] Preferably, the S5 comprises the following steps:
[0061] S51, the failure average rate data set matrix Perform clustering operations to obtain the classification matrix set of the average failure rate data set , Represents the average failure rate data set matrix The clustering result i The classification matrix of the average failure rate data set, Represents the average failure rate data set matrix The total number of classification matrices of the average failure rate data set obtained by clustering;
[0062] Classify the failure average rate data set into a matrix set Each failure average rate data set classification matrix in the rate category label is set to obtain the rate category label set , express The corresponding rate category label;
[0063] According to the rate class label set And the classification matrix set of the average failure rate data set Set the failure average rate data set matrix The rate category label data of each failure average rate data in the matrix is obtained. m ;as follows,
[0064] ;
[0065] in, The coating elastic modulus data is , coating thickness data is The rate category label data set corresponding to the failure average rate data set corresponding to the last stage blade sample set of the steam turbine; , express The elastic modulus data of the matrix is The rate category label data corresponding to the average failure rate data of the last stage blade samples of the steam turbine;
[0066] S52, according to the coating thickness data set , coating elastic modulus dataset And the matrix elastic modulus data set Get the final blade sample set matrix The coating thickness data, coating elastic modulus data and substrate elastic modulus data of each last-stage blade sample are combined to obtain a comprehensive data set matrix ;as follows,
[0067] ;
[0068] in, express The corresponding comprehensive dataset; ;
[0069] S53, constructing an initial SVM classification model, wherein the kernel function of the initial SVM classification model is a radial basis function; using the comprehensive data set matrix And the rate category label dataset matrix m Train and test the initial SVM classification model to obtain the final SVM classification model;
[0070] The failure average rate data set matrix is clustered, and then corresponding category labels are generated for each clustered classification, which reduces the complexity of the data and the difficulty of subsequent classification model construction. At the same time, it provides label data for building the final SVM classification model.
[0071] Preferably, the S53 comprises the following steps:
[0072] S531, setting a second training data ratio; and calculating the comprehensive data set matrix according to the second training data ratio. And the failure average rate data set matrix Perform data partitioning to obtain a comprehensive training data set matrix, a comprehensive test data set matrix, a failure average rate training data set matrix, and a failure average rate test data set matrix;
[0073] S532, setting a second training error threshold; inputting the comprehensive training data set matrix and the failure average rate training data set matrix into the initial SVM classification model for training operation; during the training process, when the training error is less than the second training error threshold, stopping the training to obtain a trained SVM classification model; otherwise, continuing the training;
[0074] S533, setting accuracy threshold ; Input the comprehensive test data set matrix and the failure average rate test data set matrix into the trained SVM classification model for testing; After the test is completed, the first test accuracy is obtained ;
[0075] when When , the trained SVM classification model is used as the final SVM classification model; otherwise, the trained SVM classification model is optimized until So far, the final SVM classification model is obtained;
[0076] By dividing the comprehensive data set matrix and the average failure rate data set matrix into training sets and test sets, data support is provided for training and testing the initial SVM classification model.
[0077] Preferably, in S533, the trained SVM classification model is optimized until So far, the final SVM classification model includes the following steps:
[0078] S5331, build dragonfly population , represents the first i A dragonfly, represents the size of the dragonfly population; sets the maximum number of iterations of the dragonfly population to , the current number of iterations is And the search space dimension is two-dimensional;
[0079] S5332, setting the initial penalty factor of the trained SVM classification model to And the width of the radial basis function is ; According to the initial penalty factor And width The Chebyshev chaotic mapping method is used to set the initial position of each dragonfly in the dragonfly population, and the initial position set is obtained. , , Respectively represent the first i The one-dimensional component and the two-dimensional component of the initial position of the dragonfly; the iterative formula of the Chebyshev chaotic map is as follows, ;
[0080] In the formula, , Respectively represent i +1 round of chaotic sequence iteration data and i Round chaotic sequence iterative data; β represents the mapping coefficient;
[0081] S5333: According to the accuracy threshold And test accuracy Set the fitness function of the dragonfly population ;as follows, ;
[0082] In the formula, d A positive number indicates a correction parameter;
[0083] S5334, start the iteration operation; before the iteration, the current number of iterations is Set to 1; in each iteration, the fitness function of the dragonfly population is used. Calculate the fitness value of each dragonfly to obtain a fitness value set; take the dragonfly position corresponding to the maximum fitness value in the fitness value set as the global optimal position; and calculate the fitness function of the dragonfly population according to the global optimal position. The position of each dragonfly is updated; after the update is completed, the current number of iterations is Add 1 and enter the next iteration;
[0084] S5335, when When , stop the iteration and get the final global optimal position; otherwise, continue to iterate;
[0085] The final global optimal position is used as the optimal SVM parameter set; the optimal SVM parameter set is substituted into the trained SVM classification model to obtain the optimized SVM classification model; the comprehensive test data set matrix and the failure average rate test data set matrix in S443 are input into the optimized SVM classification model for testing; after the test is completed, the second test accuracy is obtained. ;
[0086] when When , the optimized SVM classification model is used as the final SVM classification model; otherwise, return to S4434 to continue the iterative operation until So far, the final SVM classification model is obtained;
[0087] The dragonfly optimization algorithm has the characteristics of being able to perform global search effectively in the search space, helping to find the global optimal solution or at least a better solution, and reducing the risk of falling into the local optimal solution; therefore, the dragonfly optimization algorithm is used to iteratively optimize the penalty factor parameters of the trained SVM classification model and the width parameters of the radial basis kernel function, and the test accuracy is used as the variable to measure the dragonfly fitness, so that the trained SVM classification model can meet the test requirements more efficiently.
[0088] Preferably, the S6 comprises the following steps:
[0089] S61, setting a last-stage blade of a steam turbine to be diagnosed, wherein the last-stage blade of the steam turbine to be diagnosed has a pit; obtaining data of the last-stage blade of the steam turbine to be diagnosed, and obtaining a blade data set to be diagnosed , , , They respectively represent the coating thickness data, coating elastic modulus data and substrate elastic modulus data of the last stage blade of the steam turbine to be diagnosed;
[0090] S62, the leaf data set to be diagnosed Input into the final SVM classification model for classification, and obtain the failure rate label data of the blade coating to be diagnosed;
[0091] S63, setting coating failure rate label data threshold and maintenance time interval ; When the coating failure rate label data of the blade to be diagnosed is less than the coating failure rate label data threshold, the last stage blade of the steam turbine to be diagnosed needs to be repaired or replaced immediately; when the coating failure rate label data of the blade to be diagnosed is greater than or equal to the coating failure rate label data threshold, it is necessary to repair or replace the last stage blade of the steam turbine to be diagnosed immediately; Repair or replace the last stage blades of the steam turbine to be diagnosed;
[0092] The rate of coating failure determines how long the current coating can last. Therefore, judging the maintenance period of turbine blades by the coating failure rate makes the judgment process more scientific and accurate.
[0093] A fault diagnosis system for a steam turbine based on big data analysis, comprising a first blade spacing data acquisition module, a spacing mean value calculation module, a prediction model construction module, a second blade spacing data acquisition module, a prediction module, a failure average rate calculation module, a clustering module, a label generation module, a classification model construction module and a fault diagnosis module;
[0094] The first blade spacing data acquisition module is used to collect the spacing data between the blade substrate and the coating in the pits of a plurality of different types of steam turbine last-stage blade samples to obtain a spacing data matrix set;
[0095] The spacing mean calculation module is used to calculate the mean of the data in the spacing data matrix to obtain the average spacing data matrix;
[0096] The prediction model building module is used to construct a final SVM prediction model using an average spacing data matrix;
[0097] The second blade spacing data acquisition module is used to collect the spacing data between the blade substrate and the coating in the pits of a plurality of turbine last stage blade samples with different characteristics, and obtain an average spacing data set matrix set;
[0098] The prediction module is used to predict the average spacing data set matrix set using the final SVM prediction model to obtain the average spacing future data set matrix set;
[0099] The failure average rate calculation module is used to calculate the failure average rate of a plurality of steam turbine last stage blade samples with different characteristics according to the average spacing future data set matrix set and the average spacing data set matrix set to obtain the failure average rate data set matrix;
[0100] The clustering module is used to perform a clustering operation on the failure average rate data set matrix to obtain a failure average rate data set classification matrix set;
[0101] The label generation module is used to generate labels for each failure average rate data set classification matrix in the failure average rate data set classification matrix set to obtain a rate category label data set matrix;
[0102] The classification model building module is used to construct a final SVM classification model using characteristic data of a plurality of steam turbine last stage blade samples with different characteristics and a rate category label data set matrix;
[0103] The fault diagnosis module is used to perform fault diagnosis on the last-stage blades of the steam turbine to be diagnosed according to the final SVM classification model, and generate a maintenance or replacement plan.
[0104] The present invention has the following beneficial effects:
[0105] 1. In the present invention, by constructing the final SVM prediction model, a model basis is provided for the subsequent prediction of the spacing data in the future; by predicting the average spacing data in the future, the future coating failure situation can be predicted in time, thereby improving the comprehensiveness of fault diagnosis; by calculating the average failure rate of a plurality of turbine last-stage blade samples with different characteristics, a data basis is provided for the subsequent establishment of the corresponding relationship between the characteristic data of the turbine last-stage blade samples and the average failure rate; the failure average rate data set matrix is clustered, and then a corresponding category label is generated for each classification obtained by clustering, thereby reducing the complexity of the data, thereby reducing the difficulty of subsequently constructing the classification model, and at the same time providing label data for constructing the final SVM classification model; by constructing the final SVM classification model, the corresponding coating failure rate can be obtained based on the characteristic data of the turbine last-stage blade to be diagnosed, thereby simplifying the diagnosis process and improving the diagnosis efficiency.
[0106] 2. In the present invention, the time that the current coating can support is determined by the coating failure rate. Therefore, the maintenance period of the turbine blade is determined by the coating failure rate, making the determination process more scientific and accurate.
[0107] 3. In the present invention, the penalty factor parameters of the trained SVM classification model and the width parameters of the radial basis kernel function are iteratively optimized by adopting the dragonfly optimization algorithm, and the test accuracy is used as a variable to measure the dragonfly fitness, so that the trained SVM classification model can meet the test requirements more efficiently.
[0108] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS
[0109] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying creative work.
[0110] Figure 1 The present invention is a flow chart of a steam turbine fault diagnosis system based on big data analysis for diagnosing a steam turbine fault. DETAILED DESCRIPTION
[0111] The following will be combined with the drawings in the embodiments of the invention to clearly and completely describe the technical solutions in the embodiments of the invention. Obviously, the described embodiments are only part of the embodiments of the invention, not all of the embodiments. Based on the embodiments in the invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the invention.
[0112] In the description of the present invention, it is necessary to understand that the terms "opening", "upper", "lower", "top", "middle", "inside" and the like indicating orientation or positional relationship are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the invention.
[0113] Embodiment 1
[0114] This embodiment is a fault diagnosis method for a steam turbine based on big data analysis, comprising the following steps:
[0115] S1, collecting the spacing data between the blade substrate and the coating in the pits of multiple different types of steam turbine last-stage blade samples to obtain a spacing data matrix set; calculating the mean of the data in the spacing data matrix set to obtain an average spacing data matrix;
[0116] The S1 comprises the following steps:
[0117] S11. Collect multiple different types of steam turbine last-stage blade samples with pits to obtain a last-stage blade sample set. , Indicates the collected i Steam turbine last stage blade samples, Indicates the total number of samples of the last stage blades of the steam turbine collected;
[0118] S12. Setting the first historical time point set , Indicates the first set of historical time points k A historical point in time, represents the total number of historical time points in the first historical time point set; and then sets the last blade sample set Several pit points of each leaf sample in the matrix are obtained. c ;as follows,
[0119] ;
[0120] in, Represents the final leaf sample set Middle i The first j A pit point, Represents the final leaf sample set Middle i The total number of pits on each leaf sample;
[0121] The final leaf sample set Each last stage blade sample is set to the running state and is set at the first historical time point Each historical time point in the pit point matrix c The distance data between the blade substrate and the coating at each pit point is measured and collected to obtain a distance data matrix set. , Represents the first historical time point set Middle k The spacing data matrix of historical time points is as follows:
[0122] ;
[0123] in, express Middle i The first j The distance data between the blade substrate and the coating at each pit point;
[0124] Calculate the spacing data matrix set The average value of each row of data in each spacing data matrix is obtained to obtain the average spacing data matrix ;as follows,
[0125] ;
[0126] in, Represents the final leaf sample set Middle i The last stage blade samples of the steam turbine are collected at the first historical time point Middle k The average distance data between the blade substrate and the coating at each historical time point;
[0127] S2, using the average spacing data matrix to build the final SVM prediction model;
[0128] The S2 comprises the following steps:
[0129] S21, setting a first training data ratio; and calculating an average spacing data matrix according to the first training data ratio. Perform data division to obtain an average spacing training data matrix and an average spacing test data matrix;
[0130] S22, constructing an initial SVM prediction model and setting a first training error threshold; inputting the average spacing training data matrix into the initial SVM prediction model for training operation; during the training process, when the training error is less than the first training error threshold, stopping the training to obtain a trained SVM prediction model; otherwise, continuing the training;
[0131] S23, set a goodness of fit threshold; input the average spacing test data matrix into the trained SVM prediction model for testing; when the test is completed, calculate the goodness of fit value; when the goodness of fit value is greater than or equal to the goodness of fit threshold, use the trained SVM prediction model as the final SVM prediction model; otherwise, return to S22 to continue training the trained SVM prediction model until the goodness of fit value is greater than or equal to the goodness of fit threshold, and obtain the final SVM prediction model;
[0132] The calculation formula of the goodness of fit value is as follows:
[0133] ;
[0134] In the formula, d represents the goodness-of-fit value; Indicates the total number of fitted data; and represent the actual value and the fitted value respectively;
[0135] S3, collecting the spacing data between the blade substrate and the coating in the pits of a plurality of turbine last-stage blade samples with different characteristics, and obtaining an average spacing data set matrix set; using the final SVM prediction model to predict the average spacing data set matrix set, and obtaining an average spacing future data set matrix set;
[0136] The S3 comprises the following steps:
[0137] S31. Set coating thickness data set , coating elastic modulus dataset And the matrix elastic modulus data set ; , , Respectively represent the set j Coating thickness data, i The coating elastic modulus data and k The matrix elastic modulus data, , , Respectively represent the total number of set coating thickness data, the total number of coating elastic modulus data and the total number of substrate elastic modulus data;
[0138] S32, according to the coating thickness data set , coating elastic modulus dataset And the matrix elastic modulus data set Collect multiple samples of the last stage blades of the steam turbine with pits to obtain the last stage blade sample set matrix ;as follows,
[0139] ;
[0140] in, The coating elastic modulus data is , coating thickness data is A sample set of steam turbine last stage blades; , express The elastic modulus data of the matrix is Samples of steam turbine last stage blades;
[0141] S33: Setting a second historical time point set , Indicates the first set of the second historical time points i A historical point in time, Indicates the total number of historical time points in the set second historical time point set;
[0142] According to the second historical time point set Collect the final blade sample set matrix The average spacing data between the blade substrate and the coating at the pit on each last-stage blade sample is obtained to obtain the average spacing data set matrix set , Indicates that the second historical time point is concentrated at the The average spacing dataset matrix of historical time points is as follows: ;
[0143] in, express The elastic modulus data of the middle coating is , coating thickness data is The average distance data set between the blade substrate and the coating corresponding to the last stage blade sample set of the steam turbine; , express The elastic modulus data of the matrix is The average distance data between the blade substrate and the coating on the last stage blade samples of the steam turbine;
[0144] S34. Setting a future time point set , Indicates that the future time point set is the first i A future time point, Indicates the total number of future time points in the set future time point set;
[0145] According to the future time point set The average spacing dataset matrix is set Input into the final SVM prediction model for prediction operation to obtain the average spacing future data set matrix set , Indicates the future time point concentration The average spacing of future time points in the future data set matrix is as follows:
[0146] ;
[0147] in, express The elastic modulus data of the middle coating is , coating thickness data is The average distance between the blade substrate and the coating corresponding to the steam turbine last stage blade sample set is a future data set; , express The elastic modulus data of the matrix is Future data on the average distance between the blade substrate and the coating on the last stage blade samples of steam turbines;
[0148] S4, calculating the average failure rate of a plurality of steam turbine last-stage blade samples with different characteristics according to the average spacing future data set matrix set and the average spacing data set matrix set, and obtaining the average failure rate data set matrix;
[0149] The S4 comprises the following steps:
[0150] S41, setting a coating failure spacing threshold; and storing the average spacing future data set matrix set The average distance between the blade substrate and the coating on each turbine last stage blade sample in the future data is compared with the coating failure distance threshold; as time increases, the average distance future data set matrix set is recorded The future data of the average spacing between the blade substrate and the coating that is greater than or equal to the coating failure spacing threshold on each turbine last-stage blade sample and the corresponding future time point are obtained to obtain the failure future time point set matrix and the future dataset matrix of mean interval between failures ; They are as follows,
[0151] ; ;
[0152] in, , Respectively represent the coating elastic modulus data as , coating thickness data is The failure future time point set and the failure average spacing future data set corresponding to the steam turbine last stage blade sample set; , , , Respectively The elastic modulus data of the matrix is The future failure time of the turbine last stage blade sample and The elastic modulus data of the matrix is Future data of mean distance to failure of turbine last stage blade samples;
[0153] S42, the failure average spacing future data set matrix The matrix set with the average spacing data set The average spacing dataset matrix of the first historical time point is subtracted to obtain the average spacing difference dataset matrix ;as follows,
[0154] ;
[0155] in, The coating elastic modulus data is , coating thickness data is The average spacing difference data set corresponding to the last stage blade sample set of the steam turbine, , express The elastic modulus data of the matrix is The average spacing difference data of the last stage blade samples of the steam turbine is calculated as follows: ;
[0156] According to the average spacing difference data set matrix , failure future time point set matrix And the second historical time point set Calculate the final blade sample set matrix The average failure rate data of each last-stage blade sample in the , and the average failure rate data set matrix is obtained ;as follows,
[0157] ;
[0158] in, The coating elastic modulus data is , coating thickness data is The average failure rate data set corresponding to the last stage blade sample set of the steam turbine; , express The elastic modulus data of the matrix is The average failure rate data of the last stage blade samples of the steam turbine is calculated as follows:
[0159] ;
[0160] S5, performing a clustering operation on the failure average rate data set matrix to obtain a failure average rate data set classification matrix set; performing label generation on each failure average rate data set classification matrix in the failure average rate data set classification matrix set to obtain a rate category label data set matrix; using feature data of a plurality of steam turbine last stage blade samples with different features and the rate category label data set matrix to construct a final SVM classification model;
[0161] The S5 comprises the following steps:
[0162] S51, the failure average rate data set matrix Perform clustering operations to obtain the classification matrix set of the average failure rate data set , Represents the average failure rate data set matrix The clustering result i The classification matrix of the average failure rate data set, Represents the average failure rate data set matrix The total number of classification matrices of the average failure rate data set obtained by clustering;
[0163] Classify the failure average rate data set into a matrix set Each failure average rate data set classification matrix in the rate category label is set to obtain the rate category label set , express The corresponding rate category label;
[0164] According to the rate class label set And the classification matrix set of the average failure rate data set Set the failure average rate data set matrix The rate category label data of each failure average rate data in the matrix is obtained. m ;as follows,
[0165] ;
[0166] in, The coating elastic modulus data is , coating thickness data is The rate category label data set corresponding to the failure average rate data set corresponding to the last stage blade sample set of the steam turbine; , express The elastic modulus data of the matrix is The rate category label data corresponding to the average failure rate data of the last stage blade samples of the steam turbine;
[0167] S52, according to the coating thickness data set , coating elastic modulus dataset And the matrix elastic modulus data set Get the final blade sample set matrix The coating thickness data, coating elastic modulus data and substrate elastic modulus data of each last-stage blade sample are combined to obtain a comprehensive data set matrix ;as follows,
[0168] ;
[0169] in, express The corresponding comprehensive dataset; ;
[0170] S53, constructing an initial SVM classification model, wherein the kernel function of the initial SVM classification model is a radial basis function; using the comprehensive data set matrix And the rate category label dataset matrix m Train and test the initial SVM classification model to obtain the final SVM classification model;
[0171] The S53 comprises the following steps:
[0172] S531, setting a second training data ratio; and calculating the comprehensive data set matrix according to the second training data ratio. And the failure average rate data set matrix Perform data partitioning to obtain a comprehensive training data set matrix, a comprehensive test data set matrix, a failure average rate training data set matrix, and a failure average rate test data set matrix;
[0173] S532, setting a second training error threshold; inputting the comprehensive training data set matrix and the failure average rate training data set matrix into the initial SVM classification model for training operation; during the training process, when the training error is less than the second training error threshold, stopping the training to obtain a trained SVM classification model; otherwise, continuing the training;
[0174] S533, setting accuracy threshold ; Input the comprehensive test data set matrix and the failure average rate test data set matrix into the trained SVM classification model for testing; After the test is completed, the first test accuracy is obtained ;
[0175] when When , the trained SVM classification model is used as the final SVM classification model; otherwise, the trained SVM classification model is optimized until So far, the final SVM classification model is obtained;
[0176] In S533, the trained SVM classification model is optimized until So far, the final SVM classification model includes the following steps:
[0177] S5331, build dragonfly population , represents the first i A dragonfly, represents the size of the dragonfly population; sets the maximum number of iterations of the dragonfly population to , the current number of iterations is And the search space dimension is two-dimensional;
[0178] S5332, setting the initial penalty factor of the trained SVM classification model to And the width of the radial basis function is ; According to the initial penalty factor And width The Chebyshev chaotic mapping method is used to set the initial position of each dragonfly in the dragonfly population, and the initial position set is obtained. , , Respectively represent the first i The one-dimensional component and the two-dimensional component of the initial position of the dragonfly; the iterative formula of the Chebyshev chaotic map is as follows,
[0179] ;
[0180] In the formula, , Respectively represent i +1 round of chaotic sequence iteration data and i Round chaotic sequence iterative data; β represents the mapping coefficient;
[0181] S5333: According to the accuracy threshold And test accuracy Set the fitness function of the dragonfly population ;as follows,
[0182] ;
[0183] In the formula, d A positive number indicates a correction parameter;
[0184] S5334, start the iteration operation; before the iteration, the current number of iterations is Set to 1; in each iteration, the fitness function of the dragonfly population is used. Calculate the fitness value of each dragonfly to obtain a fitness value set; take the dragonfly position corresponding to the maximum fitness value in the fitness value set as the global optimal position; and calculate the fitness function of the dragonfly population according to the global optimal position. The position of each dragonfly is updated; after the update is completed, the current number of iterations is Add 1 and enter the next iteration;
[0185] S5335, when When , stop the iteration and get the final global optimal position; otherwise, continue to iterate;
[0186] The final global optimal position is used as the optimal SVM parameter set; the optimal SVM parameter set is substituted into the trained SVM classification model to obtain the optimized SVM classification model; the comprehensive test data set matrix and the failure average rate test data set matrix in S443 are input into the optimized SVM classification model for testing; after the test is completed, the second test accuracy is obtained. ;
[0187] when When , the optimized SVM classification model is used as the final SVM classification model; otherwise, return to S4434 to continue the iterative operation until So far, the final SVM classification model is obtained;
[0188] S6. Perform fault diagnosis on the last-stage blade of the steam turbine to be diagnosed according to the final SVM classification model, and generate a maintenance or replacement plan;
[0189] The S6 comprises the following steps:
[0190] S61, setting a last-stage blade of a steam turbine to be diagnosed, wherein the last-stage blade of the steam turbine to be diagnosed has a pit; obtaining data of the last-stage blade of the steam turbine to be diagnosed, and obtaining a blade data set to be diagnosed , , , They respectively represent the coating thickness data, coating elastic modulus data and substrate elastic modulus data of the last stage blade of the steam turbine to be diagnosed;
[0191] S62, the leaf data set to be diagnosed Input into the final SVM classification model for classification, and obtain the failure rate label data of the blade coating to be diagnosed;
[0192] S63, setting coating failure rate label data threshold and maintenance time interval ; When the coating failure rate label data of the blade to be diagnosed is less than the coating failure rate label data threshold, the last stage blade of the steam turbine to be diagnosed needs to be repaired or replaced immediately; when the coating failure rate label data of the blade to be diagnosed is greater than or equal to the coating failure rate label data threshold, it is necessary to repair or replace the last stage blade of the steam turbine to be diagnosed immediately; Repair or replace the last stage blades of the turbine to be diagnosed.
[0193] Embodiment 2
[0194] This embodiment discloses a fault diagnosis system for a steam turbine based on big data analysis, the system can implement the method of the above embodiment, including a first blade spacing data acquisition module, a spacing mean value calculation module, a prediction model construction module, a second blade spacing data acquisition module, a prediction module, a failure average rate calculation module, a clustering module, a label generation module, a classification model construction module and a fault diagnosis module;
[0195] The first blade spacing data acquisition module is used to collect the spacing data between the blade substrate and the coating in the pits of a plurality of different types of steam turbine last-stage blade samples to obtain a spacing data matrix set;
[0196] The spacing mean calculation module is used to calculate the mean of the data in the spacing data matrix to obtain the average spacing data matrix;
[0197] The prediction model building module is used to construct a final SVM prediction model using an average spacing data matrix;
[0198] The second blade spacing data acquisition module is used to collect the spacing data between the blade substrate and the coating in the pits of a plurality of turbine last stage blade samples with different characteristics, and obtain an average spacing data set matrix set;
[0199] The prediction module is used to predict the average spacing data set matrix set using the final SVM prediction model to obtain the average spacing future data set matrix set;
[0200] The failure average rate calculation module is used to calculate the failure average rate of a plurality of steam turbine last stage blade samples with different characteristics according to the average spacing future data set matrix set and the average spacing data set matrix set to obtain the failure average rate data set matrix;
[0201] The clustering module is used to perform a clustering operation on the failure average rate data set matrix to obtain a failure average rate data set classification matrix set;
[0202] The label generation module is used to generate labels for each failure average rate data set classification matrix in the failure average rate data set classification matrix set to obtain a rate category label data set matrix;
[0203] The classification model building module is used to construct a final SVM classification model using characteristic data of a plurality of steam turbine last stage blade samples with different characteristics and a rate category label data set matrix;
[0204] The fault diagnosis module is used to perform fault diagnosis on the last-stage blades of the steam turbine to be diagnosed according to the final SVM classification model, and generate a maintenance or replacement plan.
[0205] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.
[0206] The preferred embodiments of the invention disclosed above are only used to help explain the invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the invention, so that those skilled in the art can understand and use the invention well.
Claims
1. A fault diagnosis method for a steam turbine based on big data analysis, characterized in that: The following steps are involved: S1, collecting the spacing data between the blade substrate and the coating in the pits of multiple different types of steam turbine last-stage blade samples to obtain a spacing data matrix set; Calculate the mean of the data in the spacing data matrix to obtain the average spacing data matrix; S2, using the average spacing data matrix to build the final SVM prediction model; S3, collecting the spacing data between the blade substrate and the coating in the pits of a plurality of turbine last-stage blade samples with different characteristics, and obtaining an average spacing data set matrix set; The final SVM prediction model is used to predict the average spacing data set matrix set to obtain the average spacing future data set matrix set; S4, calculating the average failure rate of a plurality of steam turbine last-stage blade samples with different characteristics according to the average spacing future data set matrix set and the average spacing data set matrix set, and obtaining the average failure rate data set matrix; S5, performing a clustering operation on the failure average rate data set matrix to obtain a failure average rate data set classification matrix set; Generate labels for each failure average rate data set classification matrix in the failure average rate data set classification matrix set to obtain a rate category label data set matrix; The final SVM classification model is constructed using the characteristic data of multiple steam turbine last stage blade samples with different characteristics and the rate category label data set matrix; S6. Perform fault diagnosis on the last-stage blade of the steam turbine to be diagnosed according to the final SVM classification model, and generate a maintenance or replacement plan; The S3 comprises the following steps: S31, setting a coating thickness data set, a coating elastic modulus data set, and a substrate elastic modulus data set; S32, collecting a plurality of samples of last-stage blades of steam turbines with pits according to the coating thickness data set, the coating elastic modulus data set and the substrate elastic modulus data set to obtain a last-stage blade sample set matrix; S33, setting a second historical time point set; collecting average spacing data between the blade substrate and the coating at the pit on each final blade sample in the final blade sample set matrix according to the second historical time point set, to obtain an average spacing data set matrix set; S34, setting a future time point set; inputting the average spacing data set matrix set into the final SVM prediction model for prediction operation according to the future time point set, to obtain the average spacing future data set matrix set; The S4 comprises the following steps: S41, setting a coating failure spacing threshold; comparing the future data of the average spacing between the blade substrate and the coating on each steam turbine last-stage blade sample in the average spacing future data set matrix set with the coating failure spacing threshold; as time increases, recording the first future data of the average spacing between the blade substrate and the coating on each steam turbine last-stage blade sample in the average spacing future data set matrix set that is greater than or equal to the coating failure spacing threshold and the corresponding future time point, to obtain a failure future time point set matrix and a failure average spacing future data set matrix; S42, subtracting the failure average spacing future data set matrix from the average spacing data set matrix at the first historical time point in the average spacing data set matrix set to obtain an average spacing difference data set matrix; calculating the failure average rate data of each last-stage blade sample in the last-stage blade sample set matrix according to the average spacing difference data set matrix, the failure future time point set matrix and the second historical time point set to obtain a failure average rate data set matrix; The S5 comprises the following steps: S51, performing a clustering operation on the failure average rate data set matrix to obtain a failure average rate data set classification matrix set; setting a rate category label for each failure average rate data set classification matrix in the failure average rate data set classification matrix set to obtain a rate category label set; setting rate category label data for each failure average rate data in the failure average rate data set matrix according to the rate category label set and the failure average rate data set classification matrix set to obtain a rate category label data set matrix; S52, obtaining the coating thickness data, coating elastic modulus data and substrate elastic modulus data of each last-stage blade sample in the last-stage blade sample set matrix according to the coating thickness data set, coating elastic modulus data set and substrate elastic modulus data set, and combining them to obtain a comprehensive data set matrix; S53, constructing an initial SVM classification model; using the comprehensive data set matrix and the rate category label data set matrix to train and test the initial SVM classification model to obtain a final SVM classification model.
2. A method for fault diagnosis of a steam turbine based on big data analysis according to claim 1, characterized in that: The S1 comprises the following steps: S11, collecting a plurality of different types of steam turbine last-stage blade samples with pits to obtain a last-stage blade sample set; S12. Set each final-stage blade sample in the final-stage blade sample set to an operating state and measure and collect the spacing data between the blade substrate and the coating in each pit point at each historical time point to obtain a spacing data matrix set; calculate the average value of each row of data in each spacing data matrix in the spacing data matrix set to obtain an average spacing data matrix.
3. The method for fault diagnosis of a steam turbine based on big data analysis according to claim 1, characterized in that: The S53 comprises the following steps: S531, setting a second training data ratio; dividing the comprehensive data set matrix and the failure average rate data set matrix according to the second training data ratio to obtain a comprehensive training data set matrix, a comprehensive test data set matrix, a failure average rate training data set matrix, and a failure average rate test data set matrix; S532, setting a second training error threshold; inputting the comprehensive training data set matrix and the failure average rate training data set matrix into the initial SVM classification model for training operation; during the training process, when the training error is less than the second training error threshold, stopping the training to obtain a trained SVM classification model; otherwise, continuing the training; S533, setting accuracy threshold ; Input the comprehensive test data set matrix and the failure average rate test data set matrix into the trained SVM classification model for testing; After the test is completed, the first test accuracy is obtained ;when When , the trained SVM classification model is used as the final SVM classification model; otherwise, the trained SVM classification model is optimized until So far, the final SVM classification model is obtained.
4. A method for fault diagnosis of a steam turbine based on big data analysis according to claim 3, characterized in that: In S533, the trained SVM classification model is optimized until So far, the final SVM classification model includes the following steps: S5331, construct a dragonfly population; set the maximum number of iterations of the dragonfly population to , the current number of iterations is ; S5332, setting the initial penalty factor of the trained SVM classification model and the width of the radial basis function; setting the initial position of each dragonfly in the dragonfly population by Chebyshev chaotic mapping according to the initial penalty factor and the width, to obtain an initial position set; S5333: According to the accuracy threshold And test accuracy Set the fitness function of the dragonfly population; S5334, start iterative operation; in each round of iteration, calculate the fitness value of each dragonfly according to the fitness function of the dragonfly population and use the dragonfly position corresponding to the maximum fitness value as the global optimal position; update the position of each dragonfly according to the global optimal position and the fitness function of the dragonfly population; S5335, when When , stop the iteration and get the final global optimal position; otherwise, continue the iteration; take the final global optimal position as the optimal SVM parameter set; substitute the optimal SVM parameter set into the trained SVM classification model to get the optimized SVM classification model; input the comprehensive test data set matrix and the failure average rate test data set matrix into the optimized SVM classification model for testing; after the test is completed, get the second test accuracy ;when When , the optimized SVM classification model is used as the final SVM classification model; otherwise, return to S5334 to continue the iterative operation until So far, the final SVM classification model is obtained.
5. A method for fault diagnosis of a steam turbine based on big data analysis according to claim 4, characterized in that: The S6 comprises the following steps: S61, setting a last-stage blade of a steam turbine to be diagnosed, wherein the last-stage blade of the steam turbine to be diagnosed has a pit; acquiring data of the last-stage blade of the steam turbine to be diagnosed, and obtaining a blade data set to be diagnosed; S62, inputting the blade data set to be diagnosed into the final SVM classification model for classification, and obtaining the blade coating failure rate label data to be diagnosed; S63. Set the coating failure rate label data threshold and maintenance time interval; when the coating failure rate label data of the blade to be diagnosed is less than the coating failure rate label data threshold, the last-stage blade of the steam turbine to be diagnosed needs to be repaired or replaced immediately; when the coating failure rate label data of the blade to be diagnosed is greater than or equal to the coating failure rate label data threshold, the last-stage blade of the steam turbine to be diagnosed needs to be repaired or replaced within the maintenance time interval.
6. A system for implementing the fault diagnosis method for a steam turbine based on big data analysis as described in any one of claims 1 to 5.
