Processing quality prediction method and system of steam turbine blade, medium and product

Abstract

The invention provides a turbine blade machining quality prediction method and system, a medium and a product, and the prediction method comprises the steps: obtaining machine tool machining parameter error flow data and quality deviation error flow data based on a blade machining time sequence, actual machine tool machining parameter data and actual quality deviation data of a plurality of machined blades of the same type; and based on the machining mechanism information base, the machine tool machining parameter error stream data and the quality deviation error stream data, establishing an association relationship between the machine tool machining parameter error stream data and the quality deviation error stream data, and based on the association relationship, obtaining a predicted quality deviation of the to-be-machined blade. According to the method, the machine tool machining parameter error flow data and the quality deviation error flow data are combined with the machining mechanism information base of the turbine blade machining process, and the incidence relation between the actual machining machine tool parameters and the actual quality deviation is excavated through the blade machining mechanism; and high-precision prediction of the quality deviation of the to-be-processed blade is realized through the incidence relation.

Description

Technical Field

[0001] This disclosure relates to the field of steam turbine technology, and in particular to a method, system, medium, and product for predicting the processing quality of steam turbine blades. Background Technology

[0002] Turbine blades are a core component of steam turbines, and their machining quality has a significant impact on the normal operation of the unit. Problems during machining, such as random vibrations of the machine tool itself, tool wear, and unreasonable machining parameter settings, can lead to deviations in blade machining quality, resulting in substandard products. Therefore, predicting blade machining quality before production and strictly controlling the production process parameters to ensure that blade machining quality meets production requirements are key aspects that need to be focused on during the machining process.

[0003] The manufacturing process of steam turbine blades primarily generates time-series data. By utilizing historical production quality data, time-series prediction can be used to predict the quality of future blades. Currently, the main time-series prediction methods employed include recurrent neural networks (RNNs), long short-term memory (LSTM) networks, and gated unit networks (GNUs), as well as improved algorithms that fuse convolutional neural network encoding. The accuracy of the prediction model depends on the data scale. In recent years, to address the problem of production process quality prediction with relatively small data sets, researchers have also employed time-series prediction methods based on attention mechanisms, meta-learning, and fusion transfer learning to achieve quality prediction for production processes with small sample data.

[0004] Current time series forecasting methods primarily rely on data mining. For small-sample data, forecasting methods mainly improve accuracy by deeply mining data information through model reuse and transfer learning. However, for the production of multi-variety, small-batch steam turbine blades, the discontinuity and small data scale of time series data pose challenges to improving the accuracy of time series forecasting models. Given the multi-variety, small-batch characteristics of steam turbine blade processing, it is necessary to integrate time series forecasting methods based on small-sample data to improve the accuracy of predicting steam turbine blade processing quality. Furthermore, due to the multi-variety nature of steam turbine blade processing, production time series data within a time window often includes production time series data for different types of steam turbine blades. This leads to time discontinuity in the extraction of processing data for the same type of blade, and the wear and load parameters of the cutting tools change after processing other types of blades.

[0005] Therefore, how to effectively predict the processing quality of turbine blades is a technical problem that urgently needs to be solved. Summary of the Invention

[0006] The technical problem to be solved by this disclosure is to overcome the shortcomings of the prior art in that it is impossible to effectively predict the processing quality of steam turbine blades, and to provide a method, system, medium and product for predicting the processing quality of steam turbine blades.

[0007] This disclosure solves the above-mentioned technical problems through the following technical solution:

[0008] Firstly, a method for predicting the machining quality of steam turbine blades is provided, the method comprising:

[0009] Obtain actual quality deviation data of several processed blades of the same type, and actual machine tool processing parameter data of the processed blades within a preset sampling period;

[0010] Among them, the actual machine tool processing parameter data corresponds to multiple machine tool processing parameters, and the actual processing quality deviation data corresponds to multiple quality characteristics;

[0011] Based on the blade processing sequence, the actual machine tool processing parameter data, and the actual quality deviation data, obtain machine tool processing parameter error stream data and quality deviation error stream data;

[0012] Obtain a database of processing mechanism information for the turbine blade manufacturing process;

[0013] The processing mechanism information database represents the computational relationship between the various quality characteristics of the turbine blade and the preset processing program.

[0014] Based on the machining mechanism information database, the machine tool machining parameter error flow data, and the quality deviation error flow data, an association relationship is established between the machine tool machining parameter error flow data and the quality deviation error flow data. Based on the association relationship, the predicted quality deviation of the blade to be processed is obtained.

[0015] Optionally, the step of obtaining machine tool machining parameter error stream data and quality deviation error stream data based on the blade machining sequence, the actual machine tool machining parameter data, and the actual quality deviation data includes:

[0016] Based on the blade processing sequence, the actual machine tool processing parameter data of the currently processed blade is successively subtracted from the actual machine tool processing parameter data of the previous processed blade to obtain the machine tool processing parameter error stream data.

[0017] Based on the blade processing sequence, the actual quality deviation data of the currently processed blade is successively subtracted from the actual quality deviation data of the previous processed blade to obtain the quality deviation error stream data;

[0018] And / or, the processing mechanism information database includes program processing feature comparison mechanisms and processing experience mechanisms.

[0019] Optionally, the step of obtaining the predicted quality deviation of the blade to be processed based on the correlation between the machining mechanism information database, the machine tool machining parameter error stream data, and the quality deviation error stream data, and based on the correlation, includes:

[0020] Based on the machining mechanism information database and the machine tool machining parameter error flow data, a machining parameter prediction model corresponding to each quality feature is constructed.

[0021] Based on the blade processing timing, the initial machine tool processing parameter error flow data corresponding to the previous processed blade adjacent to the blade to be processed is obtained, and the initial machine tool processing parameter error flow data is input into the processing parameter prediction model to predict the predicted machine tool processing parameter error flow data corresponding to the blade to be processed.

[0022] Based on the machining mechanism information database, the machine tool machining parameter error flow data, and the quality deviation error flow data, an error correlation model corresponding to each quality feature is constructed.

[0023] The error correlation model is used to characterize the correlation between the machine tool machining parameter error stream data and the quantity deviation error stream data.

[0024] Based on the predicted machine tool machining parameter error flow data and the error correlation model, the predicted quality deviation corresponding to the blade to be processed is obtained.

[0025] Optionally, the step of constructing a machining parameter prediction model corresponding to each quality feature based on the machining mechanism information database and the machine tool machining parameter error stream data includes:

[0026] Based on the machining mechanism information database and the machine tool machining parameter error flow data, the machine tool parameter error flow sequence dataset for each turbine blade under the preset machining program corresponding to each quality feature is extracted to obtain the machine tool machining parameter error flow dataset for each quality feature.

[0027] The machine tool machining parameter error stream dataset is divided according to a preset data ratio to obtain a first machine tool machining parameter error stream dataset and a second machine tool machining parameter error stream dataset.

[0028] Using the first machine tool machining parameter error stream dataset corresponding to the same quality feature as input and the second machine tool machining parameter error stream dataset corresponding to the same quality feature as output, a machining parameter prediction model corresponding to each quality feature is constructed.

[0029] And / or, the step of constructing an error correlation model corresponding to each quality feature based on the machining mechanism information database, the machine tool machining parameter error stream data, and the machining quality deviation error stream data includes:

[0030] Based on the machining mechanism information database and the machine tool machining parameter error flow data, the machine tool parameter error flow sequence dataset for each turbine blade under the preset machining program corresponding to each quality feature is extracted to obtain the machine tool machining parameter error flow dataset for each quality feature.

[0031] The quality deviation error stream data is classified according to the error range to obtain quality deviation error category labels;

[0032] The quality deviation error category label includes a quality deviation error category and the error range corresponding to the quality deviation error category;

[0033] Based on the processing mechanism information database and the quality deviation error category labels, the error category label sequence data of each turbine blade under the preset processing program corresponding to each quality feature is extracted to obtain the quality deviation error category label set corresponding to each quality feature;

[0034] Using the machine tool machining parameter error flow dataset corresponding to the same quality feature as input and the quality deviation error category label set corresponding to the same quality feature as output, the error correlation model corresponding to each quality feature is constructed.

[0035] Optionally, the step of obtaining the predicted quality deviation corresponding to the blade to be processed based on the predicted machine tool machining parameter error data and the error correlation model includes:

[0036] The predicted machine tool processing parameter error data corresponding to the target quality characteristics of the blade to be processed is input into the error association model corresponding to the target quality characteristics, and the predicted quality deviation error category corresponding to the target quality characteristics of the blade to be processed is output.

[0037] Based on the blade processing sequence, obtain the target quality deviation data of the previous processed blade adjacent to the blade to be processed;

[0038] The predicted quality deviation of the blade to be processed is obtained by summing the error range corresponding to the predicted quality deviation error category with the target quality deviation data.

[0039] Optionally, the processing parameter prediction model includes a first long short-term memory network layer, a second long short-term memory network layer, and a multi-head attention mechanism network layer cascaded in sequence.

[0040] And / or, the error correlation model includes a convolutional neural network layer, a third long short-term memory network layer, a fourth long short-term memory network layer, an activation function layer, and a fully connected network layer, which are cascaded in sequence.

[0041] Secondly, a machining quality prediction system for steam turbine blades is provided, the machining quality prediction system comprising:

[0042] The actual data acquisition module is used to acquire actual quality deviation data of several processed blades of the same type, and actual machine tool processing parameter data of the processed blades within a preset sampling period.

[0043] Among them, the actual machine tool processing parameter data corresponds to multiple machine tool processing parameters, and the actual processing quality deviation data corresponds to multiple quality characteristics;

[0044] The error stream data acquisition module is used to acquire machine tool processing parameter error stream data and quality deviation error stream data based on the blade processing sequence, the actual machine tool processing parameter data, and the actual quality deviation data.

[0045] The processing mechanism information acquisition module is used to acquire the processing mechanism information database of the turbine blade processing process;

[0046] The processing mechanism information database represents the computational relationship between the various quality characteristics of the turbine blade and the preset processing program.

[0047] The quality deviation prediction module is used to establish a correlation between the machine tool machining parameter error flow data and the quality deviation error flow data based on the machining mechanism information database, the machine tool machining parameter error flow data and the quality deviation error flow data, and to obtain the predicted quality deviation of the blade to be processed based on the correlation.

[0048] Thirdly, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and used to run on the processor, wherein the processor executes the computer program to implement the above-described method for predicting the machining quality of turbine blades.

[0049] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the above-described method for predicting the machining quality of turbine blades.

[0050] Fifthly, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the aforementioned method for predicting the machining quality of turbine blades.

[0051] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of this disclosure.

[0052] The positive and progressive effects of this disclosure are as follows:

[0053] The disclosed method, system, medium, and product for predicting the machining quality of turbine blades obtain machine tool machining parameter error flow data and quality deviation error flow data by processing the actual machining quality deviation data and actual machine tool machining parameter data of the machined blades. The machine tool machining parameter error flow data and quality deviation error flow data are combined with the machining mechanism information database of the turbine blade machining process. By mining the blade machining mechanism, the correlation between the actual machining machine tool parameters and the actual quality deviation is obtained. Through the correlation, high-precision prediction of the quality deviation of the blade to be machined is achieved. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the first process of the turbine blade machining quality prediction method provided in Embodiment 1 of this disclosure;

[0055] Figure 2 This is a schematic diagram of the second process of the turbine blade machining quality prediction method provided in Embodiment 1 of this disclosure;

[0056] Figure 3 This is a schematic diagram of the third process of the turbine blade machining quality prediction method provided in Embodiment 1 of this disclosure;

[0057] Figure 4 This is a schematic diagram of the fourth process of the turbine blade machining quality prediction method provided in Embodiment 1 of this disclosure;

[0058] Figure 5 This is a schematic diagram illustrating the construction process of the machining parameter prediction model in the machining quality prediction method for turbine blades provided in Embodiment 1 of this disclosure;

[0059] Figure 6 This is a schematic diagram of the fifth process of the turbine blade machining quality prediction method provided in Embodiment 1 of this disclosure;

[0060] Figure 7 This is a schematic diagram illustrating the construction process of the error correlation model in the turbine blade machining quality prediction method provided in Embodiment 1 of this disclosure;

[0061] Figure 8 This is a schematic diagram of the sixth process of the turbine blade machining quality prediction method provided in Embodiment 1 of this disclosure;

[0062] Figure 9 This is a schematic diagram of the turbine blade machining quality prediction system provided in Embodiment 2 of this disclosure;

[0063] Figure 10 This is a schematic diagram of the structure of the electronic device provided in Embodiment 3 of this disclosure. Detailed Implementation

[0064] The present disclosure is further illustrated below by way of embodiments, but the present disclosure is not limited to the scope of the embodiments described herein.

[0065] The prefixes such as "first" and "second" used in this disclosure are merely for distinguishing different descriptive objects and do not limit the position, order, priority, quantity, or content of the described objects. The use of ordinal numbers and other prefixes used to distinguish descriptive objects in this disclosure does not constitute a limitation on the described objects. The description of the described objects is given in the claims or the context of the embodiments, and should not be construed as an unnecessary limitation. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.

[0066] Example 1

[0067] This embodiment provides a method for predicting the machining quality of steam turbine blades, such as... Figure 1 As shown, the processing quality prediction method includes:

[0068] S1. Obtain actual quality deviation data of several processed blades of the same type, and actual machine tool processing parameter data of processed blades within a preset sampling period.

[0069] Among them, the actual machine tool machining parameter data corresponds to multiple machine tool machining parameters, and the actual machining quality deviation data corresponds to multiple quality characteristics.

[0070] S2. Based on the blade processing sequence, actual machine tool processing parameter data, and actual quality deviation data, obtain machine tool processing parameter error flow data and quality deviation error flow data.

[0071] S3. Obtain a database of processing mechanism information for turbine blade manufacturing processes.

[0072] Among them, the processing mechanism information database represents the operational relationship between various quality characteristics of turbine blades and preset processing programs.

[0073] S4. Based on the machining mechanism information database, machine tool machining parameter error flow data and quality deviation error flow data, establish the correlation between the machine tool machining parameter error flow data and the quality deviation error flow data, and obtain the predicted quality deviation of the blade to be processed based on the correlation.

[0074] For example, data corresponding to nine machine tool machining parameters are obtained from at least 100 identical turbine blades during the blade machining process within a preset sampling period (T), including machine tool shaft 1 load, shaft 2 load, shaft 3 load, shaft 4 load, shaft 5 load, spindle load, spindle speed, feed rate, and feed amount. This yields the actual machine tool machining parameter data for the machined blades. After machining, the turbine blade quality parameter data is measured, and the quality deviation is calculated based on the upper and lower tolerances. The turbine blade quality parameter data includes data corresponding to 62 machining quality characteristics related to the blade root, inlet edge, and outlet edge, including length, width, inner arc, flatness, cone angle, and pitch.

[0075] The blade processing sequence characterizes the order in which each blade is processed. By combining the blade processing sequence, actual machine tool processing parameter data, and actual quality deviation data, machine tool processing parameter error flow data and quality deviation error flow data are obtained.

[0076] A processing mechanism information database for the turbine blade manufacturing process is established in advance. This database represents the computational relationship between various quality characteristics of the turbine blade and the preset processing program.

[0077] The turbine blade machining quality prediction method in this embodiment processes the actual machining quality deviation data and actual machine tool machining parameter data of the machined blades to obtain machine tool machining parameter error flow data and quality deviation error flow data. The machine tool machining parameter error flow data and quality deviation error flow data are combined with the machining mechanism information database of the turbine blade machining process. By mining the blade machining mechanism, the correlation between the actual machining machine tool parameters and the actual quality deviation is obtained. Through the correlation, high-precision prediction of the quality deviation of the blade to be machined is achieved.

[0078] In an alternative implementation, such as Figure 2 As shown, step S2 above includes:

[0079] S21. Based on the blade processing sequence, the actual machine tool processing parameter data of the currently processed blade is successively subtracted from the actual machine tool processing parameter data of the previous processed blade to obtain the machine tool processing parameter error stream data.

[0080] S22. Based on the blade processing sequence, the actual quality deviation data of the currently processed blade is successively subtracted from the actual quality deviation data of the previous processed blade to obtain the quality deviation error stream data.

[0081] Acquire the machine tool shaft 1 load, shaft 2 load, shaft 3 load, shaft 4 load, shaft 5 load, spindle load, spindle speed, feed rate, and feed rate measurement data during the turbine blade machining process within a preset sampling period (fixed sampling time). The actual machine tool machining parameter data corresponding to the above 9 parameters can be represented as a vector of length T. That is, actual machine tool processing parameter data Where i = 1, 2, ..., 9 represent the machine tool processing parameter numbers, s = 1, 2, ..., n represent the selected turbine blade sequence numbers of the same category, T is the number of sampling points of the machine tool processing parameters of a blade, and n is the number of selected turbine blades, and n > 100.

[0082] After machining, the actual quality parameters of the turbine blades are measured, and the actual quality deviation is calculated based on the upper and lower tolerances. Where j=1, 2, ..., 62 represent the numbers of the 62 quality characteristics, and the actual quality deviation data. Data can be in percentage format.

[0083] Based on the blade processing sequence, the actual machine tool processing parameter data of the currently processed blades are used. Compare the actual machine tool machining parameter data with the data of the previously processed blade. The difference is used to construct the machine tool machining parameter error stream data. ;

[0084] , ,in Given a zero-based vector of length T, a machine tool machining parameter error flow dataset can be constructed. ;

[0085] .

[0086] Based on the blade processing sequence, the actual quality deviation data of the currently processed blades are analyzed. The actual quality deviation data of the previous processed blade. The difference is calculated to obtain the quality deviation error stream data. ,but , .

[0087] To address the discontinuity in machine tool parameter data within the same category of data caused by the intervening processing of other blade categories during the machining of the same category of blades, a machine tool machining parameter error flow data for turbine blades is established by sequentially subtracting the machine tool machining parameter data and quality deviation data of the machined blades from the corresponding data of the preceding blade. and quality deviation error stream data This avoids the impact of discontinuous time-series data sequences on quality deviation prediction and improves prediction accuracy.

[0088] In an optional implementation, the processing mechanism information database includes program processing feature comparison mechanisms and processing experience mechanisms.

[0089] The mechanism information database is represented in the following dictionary format. The processing program number is obtained from the internal definition of the program. Assuming that the 12 preset processing programs involved are defined as n1~n12, the preset processing programs can be represented as sequence vectors. The following table This indicates a label for a preset machining program sequence and is not data. The mechanism information database is represented in the format (quality feature, [corresponding preset machining program set]}). The machining mechanism information database includes program machining feature comparison mechanisms and machining experience mechanisms.

[0090] For example, the mechanism for comparing machining features includes: [(Feature 1, [machining program n6, machining program n7]), [(Feature 2, [machining program n2]), (Feature 3, [machining program n7]), (Feature 4, [machining program n5, machining program n7]), (Feature 6, machining program n5), (Feature 7, machining program n8), (Feature 8, machining program n8), (Feature 9, machining program n8), (Feature 10, machining program n9), (Feature 11, machining program n9), (Feature 12, machining program n8), (Feature 13, [machining program n8, machining program n9]), (Feature 14, machining program n9), (Feature 15, machining program n8), (Feature 16, [machining program n8, machining program n9]), (Feature 17, [machining program n8, machining program n9]). Program n9]), (Feature 18, [Machining program n8, Machining program n9]), (Feature 19, Machining program n9), (Feature 20, Machining program n9), (Feature 21, [Machining program n8, Machining program n9]), (Feature 22, [Machining program n8, Machining program n9]), (Feature 23, [Machining program n8, Machining program n9]), (Feature 24, [Machining program n11, Machining program n12]), (Feature 25, Machining program n11), (Feature 26, Machining program n12), (Feature 27, [Machining program n5, Machining program n11]), (Feature 28, [Machining program n7, Machining program n12]), (Feature 29, [ [Machining program n8, machining program n12]), (Feature 30, machining program n12), (Feature 31, machining program n11), (Feature 32, machining program n12), (Feature 33, [Machining program n7, machining program n10]), (Feature 34, [Machining program n1, machining program n2]), (Feature 35, machining program n1), (Feature 36, machining program n2), (Feature 37, machining program n3), (Feature 38, machining program n3), (Feature 39, machining program n3), (Feature 40, machining program n4), (Feature 41, machining program n4), (Feature 42, machining program n4), (Feature 43, [Machining program n3, machining program n12] ...30, machining program n12), (Feature 31, machining program n11), (Feature 32, machining program n12), (Feature 43, [Machining program n7, machining program n10]), (Feature 34, machining program n1, machining program n12]), (Feature 45, machining program n1), (Feature 36, machining program n2), (Feature 47, machining program n3), (Feature 48, machining program n3), (Feature 49, machining program n3), (Feature 40, machining program n4), (Feature 41, machining program n4), (Feature 42, machining program n4), (Feature 43, machining program n4), (Feature 44, machining program n4), (Feature 45, machining program n12), (Feature 46, machining program n12), (Feature 47, machining program n12 Program n4]), (Feature 44, machining program n4), (Feature 45, machining program n3), (Feature 46, [machining program n3, machining program n4]), (Feature 47, [machining program n3, machining program n4]), (Feature 48, [machining program n3, machining program n4]), (Feature 49, [machining program n5, machining program n11]), (Feature 50, [machining program n6, machining program n11]), (Feature 51, [machining program n7, machining program n12]), (Feature 52, [machining program n7, machining program n12]), (Feature 53, [machining program n5, machining program n11]), (Feature 54, [machining program n7,[Machining program n12], (Feature 55, machining program n11), (Feature 56, machining program n11), (Feature 57, machining program n11), (Feature 58, machining program n11), (Feature 59, machining program n12), (Feature 60, machining program n12), (Feature 61, machining program n12), (Feature 62, machining program n12)]. The mechanism of comparison between machining program features characterizes the comparison relationship between features and machining programs.

[0091] For example, the processing experience mechanism includes: [(key feature, feature 4), (four planes of the leaf root, [processing program n3, processing program n4, processing program n5, processing program n6]), (feature measurement value and relationship 1, [feature 1, feature 2, feature 3]), (feature measurement value and relationship 2, [feature 4, feature 5, feature 6]), (feature measurement value and relationship 3, [feature 13, feature 8, feature 11]), (feature measurement value and relationship 4, [feature 16, feature 14, feature 15]), (feature measurement value and relationship 5, [feature 21, feature 19, feature 20]), (feature measurement value and relationship 6, [feature 43, feature 38, feature 41]), (feature measurement value and relationship 7, [feature 46, feature 44, feature 45])]. Wherein, (key feature, feature 4) indicates that feature 4 is a key feature, and (feature measurement value and relationship 1, [feature 1, feature 2, feature 3]) indicates that the measurement value of feature 1 is equal to the sum of features 2 and 3.

[0092] In an alternative implementation, such as Figure 3 As shown, step S4 above includes:

[0093] S41. Based on the machining mechanism information database and machine tool machining parameter error flow data, construct a machining parameter prediction model corresponding to each quality feature.

[0094] S42. Based on the blade processing time sequence, obtain the initial machine tool processing parameter error flow data corresponding to the previous processed blade adjacent to the blade to be processed, input the initial machine tool processing parameter error flow data into the processing parameter prediction model, and predict the predicted machine tool processing parameter error flow data corresponding to the blade to be processed.

[0095] S43. Based on the machining mechanism information database, machine tool machining parameter error flow data, and quality deviation error flow data, construct the error correlation model corresponding to each quality feature.

[0096] Among them, the error correlation model is used to characterize the correlation between machine tool machining parameter error flow data and quality deviation error flow data;

[0097] S44. Based on the error flow data of the predicted machine tool processing parameters and the error correlation model, the predicted quality deviation of the blade to be processed is obtained.

[0098] In this embodiment, a machining parameter prediction model is constructed to predict the machine tool machining parameter error flow data corresponding to the blade to be machined; an error correlation model is constructed to characterize the correlation between the machine tool machining parameter error flow data and the quality deviation error flow data, and the quality deviation corresponding to the blade to be machined is predicted by combining the machining parameter error flow data predicted by the machining parameter prediction model.

[0099] In an alternative implementation, such as Figure 4 As shown, step S41 above includes:

[0100] S411. Based on the machining mechanism information database and machine tool machining parameter error flow data, extract the machine tool parameter error flow sequence dataset for each turbine blade under the preset machining program corresponding to each quality feature, and obtain the machine tool machining parameter error flow dataset corresponding to each quality feature.

[0101] S412. Divide the machine tool machining parameter error stream dataset according to the preset data ratio to obtain the first machine tool machining parameter error stream dataset and the second machine tool machining parameter error stream dataset.

[0102] S413. Using the first machine tool machining parameter error stream dataset corresponding to the same quality feature as input and the second machine tool machining parameter error stream dataset corresponding to the same quality feature as output, construct the machining parameter prediction model corresponding to each quality feature.

[0103] like Figure 5 As shown, feature extraction is first performed. Specifically, based on the mechanism information database, the machine tool parameter error flow sequence dataset for each turbine blade under the preset machining program corresponding to each quality feature j is extracted. This yields the machine tool machining parameter error stream dataset corresponding to each serialized quality feature. The machine tool machining parameter error stream dataset is divided according to a preset data ratio (e.g., 9:1) to obtain a first machine tool machining parameter error stream dataset and a second machine tool machining parameter error stream dataset. Using the first machine tool machining parameter error stream dataset corresponding to the same quality feature as input and the second machine tool machining parameter error stream dataset corresponding to the same quality feature as output, a machining parameter prediction model corresponding to each quality feature is constructed to predict machining parameter error data for future time series. The prediction is based on the processing parameters of each quality feature.

[0104] Specifically, such as Figure 5 As shown, the processing parameter prediction model includes a length temporal memory encoding network layer (a first long short temporal memory network layer and a second long short temporal memory network layer cascaded in sequence) and a multi-head attention mechanism network layer.

[0105] Since the machine tool processing parameter error stream data is a multi-parameter interval sequence data, the data corresponding to one feature of each blade is 9×N dimensional data, where N represents the data length obtained by sampling the processing process of the preset processing program corresponding to the quality feature, and 9 represents the 9 machine tool processing parameters. Based on the input two-dimensional data dimension, a multi-input port length temporal memory unit is established. To avoid overfitting caused by the small data scale, a two-layer length temporal memory encoding network (i.e., the first long short temporal memory network layer and the second long short temporal memory network layer) is used as the encoding network, and a data dimension reconstruction layer is added to the encoded data.

[0106] The structure of the two-layer long and short sequence memory coding network is as follows: it uses the sequence data of machine tool parameter error stream. The input data for the Long Short-Term Memory (LSTM) network is as follows, where the structure of the first LSM network layer is: the input gate output is represented as... The output of the forget gate is The output result of the output gate is represented as The memory cells of the memory unit output as The output of the memory update function is: The output of the hidden state is represented as The structure of the second long short-term memory network layer is as follows: the input gate output is represented as... The output of the forget gate is The output result of the output gate is represented as The memory cells of the memory unit output as The output of the memory update function is: The output of the hidden state is represented as .

[0107] Based on the output of the second long short-term memory network layer, the matrices corresponding to the nine machine tool machining parameters output in the time sequence are fused and concatenated into a one-dimensional vector using multi-channel feature data. A multi-head attention mechanism network layer is constructed to decode the fused vector using this mechanism. The query, key, and value projection transformations for the v-th head of the multi-head network can be expressed as follows: , and ,in , , The process involves constructing a scaling dot product attention mechanism, which consists of a linear matrix of query, key, and value. The head output is ,in To scale the dot product output, The output of the dot product scaling attention mechanism is for the v-th head. For data Dimensions, further multi-head output splicing into Where v is the head length, For multi-head concatenation matrices, key feature extraction is performed. Where LayerNorm represents the data normalization operation, Represents the normalized bias matrix. Feature extraction for normalization; temporal association mining ,in and It is a linear incidence matrix. and For linear correlation bias, Based on linear correlation features, the output provides time-series prediction data with high accuracy. ,in This represents the j-th quality feature corresponding to the j-th quality feature. The processing parameter error data predicted by the time series (i.e., the predicted processing parameter error data of the blade to be processed). To predict the output matrix, This is used to predict the output bias.

[0108] Based on the mean square error between the predicted results and the actual sequence results As the loss function, where The function represents the average value of a vector, k represents the number of the machine tool machining parameter, and the Adam optimizer is used as the gradient update learning tool for the feedback network. A maximum number of learning iterations is set to obtain the machining parameter prediction model corresponding to the quality feature j, ultimately yielding the predicted machining parameter error data for future time series. Similarly, the processing parameter prediction model for the next quality feature is constructed sequentially until the processing parameter prediction models for all quality features are obtained.

[0109] In an alternative implementation, such as Figure 6 As shown, step S43 above includes:

[0110] S431. Based on the machining mechanism information database and machine tool machining parameter error flow data, extract the machine tool parameter error flow sequence dataset for each turbine blade under the preset machining program corresponding to each quality feature, and obtain the machine tool machining parameter error flow dataset corresponding to each quality feature.

[0111] S432. Classify the quality deviation error stream data according to the error range to obtain quality deviation error category labels.

[0112] The quality deviation error category label includes the quality deviation error category and the corresponding error range.

[0113] S433. Based on the processing mechanism information database and quality deviation error category labels, extract the error category label sequence data of each turbine blade under the preset processing program corresponding to each quality feature, and obtain the quality deviation error category label set corresponding to each quality feature.

[0114] S434. Using the machine tool processing parameter error flow dataset corresponding to the same quality feature as input and the quality deviation error category label set corresponding to the same quality feature as output, construct the error correlation model corresponding to each quality feature.

[0115] like Figure 7 As shown, feature extraction is first performed. Specifically, based on the mechanism information database, the machine tool parameter error flow sequence dataset for each turbine blade under the preset machining program corresponding to each quality feature j is extracted. This yields the machine tool machining parameter error flow dataset corresponding to each quality feature. .

[0116] Quality deviation error stream data Based on the error range, five categories are defined, resulting in quality deviation error category labels. Quality deviation error category label This includes the quality deviation error category and the corresponding error range for each quality deviation error category. {[Error Category 1: Less than -50%], [Error Category 2: -50% to -10%], [Error Category 3: -10% to 10%], [Error Category 4: 10% to 50%], [Error Category 5: Greater than 50%]}.

[0117] like Figure 7 As shown, feature extraction is first performed based on the processing mechanism information database and quality deviation error category labels. Extract the error category label sequence data of each turbine blade under the preset machining program corresponding to each quality feature j. , This yields the quality deviation error category label set corresponding to each quality feature. The machine tool machining parameter error flow dataset corresponding to the same quality feature j. As input, the quality deviation error category label set corresponding to the same quality feature j. As the output, an error correlation model corresponding to each quality feature is constructed to classify the quality deviation error of the blade to be processed. The predictions are based on the corresponding error correlation model for each quality feature.

[0118] Specifically, such as Figure 7As shown, the error correlation model includes a series of convolutional neural network layers, long short-term memory network layers (third long short-term memory network layer and fourth long short-term memory network layer), activation function layer and fully connected network layer.

[0119] The convolutional neural network layer consists of a first convolutional network layer, a second convolutional network layer, and a pooling layer. The first convolutional network layer is represented as follows: ,in This represents the sequence data extracted from the machining program corresponding to the s-th blade under the j-th quality feature, i.e., the aforementioned machine tool machining parameter error stream dataset. This is used as the input to the error correlation model. Wherein, This is the first-layer two-dimensional convolution kernel; * indicates the convolution operation. For the bias term of a one-dimensional convolutional layer, For the first layer output, the activation function is the ReLU function, which satisfies... ,in, This is the output of the activation function.

[0120] The second convolutional network layer is represented as follows: ,in This is the output of the second convolutional network. This is the second convolutional kernel. For the bias term of the second convolutional network, the activation function is also set to... .

[0121] Pooling layers are represented as follows: ,in The output of the pooling layer sequence is represented by MaxPool, which is the pooling operation function. MaxPool extracts the output of the second convolutional layer within a selected two-dimensional window. The maximum value is used as the output of the pooling layer to improve the robustness of the network.

[0122] Based on the above pooling layer, the output is... Perform a flattening operation, that is, flatten the data into one-dimensional vector data, represented as: The Flatten function represents the one-dimensional flattening operation function for two-dimensional data.

[0123] For the third long short-term memory network layer: input gate unit calculation ,in The output is the input gate calculation unit. The input gate is a weight matrix of the input data sequence. The hidden state is input as a sequence weight matrix. The input sequence for the hidden state is obtained from the output of the memory unit. For the input gate bias term, For activation function; forget gate unit calculation ,in Output of the forget gate calculation unit. The input data sequence weight matrix is ​​used as the input to the forget gate. Input sequence for hidden state The weight matrix, For the forget gate bias term; output gate unit calculation ,in Output to the output gate calculation unit. The weight matrix is ​​the input data sequence to the output gate. Input sequence for hidden state The weight matrix, For the output gate bias term; memory cell calculation includes: candidate memory cell formula is ,in Output for memory unit cells, Input the weight matrix into the memory cell. The weight matrix is ​​the input sequence of the hidden states of the memory unit. For memory cell bias terms, the memory update function is: The hidden state unit output is .

[0124] The fourth long short-term memory (LSTM) network layer has the same structure as the third LSM network layer, and corresponds to the third LSM network layer. The input gate output result is represented as follows: The output of the forget gate is The output result of the output gate is represented as The memory cells of the memory unit output as The output of the memory update function is: The output of the hidden state is represented as .

[0125] The activation function layer uses the Sigmoid function as the activation function, and its output can be expressed as follows: Finally, a fully connected network layer outputs the predicted quality deviation error category of the blade to be processed. , , Represented as the weights of the fully connected network layers. This represents the bias term for the fully connected network layer. {[Error Category 1: Less than -50%], [Error Category 2: -50% to -10%], [Error Category 3: -10% to 10%], [Error Category 4: 10% to 50%], [Error Category 5: Greater than 50%]} represent the error categories of the quality deviation of the blade to be processed predicted by the error correlation model.

[0126] Quality deviation error category predicted by the model , with quality deviation error category label Using the cross-entropy as the loss function, the model can be expressed as: Where b represents the quality deviation error category predicted by the model, and further using the loss function, a feedback learning network based on the Adam optimizer is established to update the parameters, forming the machine tool machining parameter error stream data under the j-th quality feature. Error stream data of mass deviation between blades The error correlation model is constructed and saved. Based on the same principle, the error correlation model for the next quality feature is constructed sequentially until the error correlation models for all quality features are obtained.

[0127] In an alternative implementation, such as Figure 8 As shown, step S44 above includes:

[0128] S441. Input the predicted machine tool processing parameter error data corresponding to the target quality characteristics of the blade to be processed into the error association model with the target quality characteristics, and output the predicted quality deviation error category corresponding to the target quality characteristics of the blade to be processed.

[0129] S442. Obtain the target quality deviation data of the previous processed blade adjacent to the blade to be processed based on the blade processing time sequence.

[0130] S443. Calculate the sum of the error range corresponding to the predicted quality deviation error category and the target quality deviation data to obtain the predicted quality deviation of the target quality characteristics of the blade to be processed.

[0131] Using the machining parameter prediction model corresponding to the target quality feature (e.g., the j-th quality feature), the predicted machine tool machining parameter error data corresponding to the target quality feature of the blade to be processed (the (s+1)-th blade) is predicted. The predicted machine tool machining parameter error data is input into the error correlation model with the target quality features, and the predicted quality deviation error category corresponding to the target quality features of the blade to be processed is output. .

[0132] Based on the blade processing time sequence, obtain the target quality deviation data of the previous processed blade (s-th blade) adjacent to the blade to be processed (s+1-th blade). Calculate the predicted quality deviation error category The sum of the corresponding error range and the target quality deviation data yields the predicted quality deviation of the target quality characteristics of the blade to be processed. This allows for the prediction of quality deviations in the blades to be processed.

[0133] Example 2

[0134] Corresponding to the aforementioned embodiment of the method for predicting the machining quality of turbine blades, this disclosure also provides an embodiment of a system for predicting the machining quality of turbine blades, such as... Figure 9 As shown, the turbine blade machining quality prediction system includes:

[0135] Actual data acquisition module 1 is used to acquire actual quality deviation data of several processed blades of the same type, and actual machine tool processing parameter data of processed blades within a preset sampling period.

[0136] Among them, the actual machine tool processing parameter data corresponds to multiple machine tool processing parameters, and the actual processing quality deviation data corresponds to multiple quality characteristics;

[0137] Error flow data acquisition module 2 is used to acquire machine tool processing parameter error flow data and quality deviation error flow data based on blade processing timing, actual machine tool processing parameter data and actual quality deviation data;

[0138] The processing mechanism information acquisition module 3 is used to acquire the processing mechanism information database of the turbine blade processing process;

[0139] Among them, the processing mechanism information database represents the operational relationship between various quality characteristics of turbine blades and preset processing programs;

[0140] The quality deviation prediction module 4 is used to establish the correlation between the machine tool machining parameter error flow data and the quality deviation error flow data based on the machining mechanism information database, machine tool machining parameter error flow data and quality deviation error flow data, and obtain the predicted quality deviation of the blade to be processed based on the correlation.

[0141] In an optional implementation, the error stream data acquisition module 2 includes:

[0142] The machine tool error flow acquisition unit 21 is used to obtain machine tool processing parameter error flow data by subtracting the actual machine tool processing parameter data of the currently processed blade from the actual machine tool processing parameter data of the previous processed blade based on the blade processing sequence.

[0143] The quality error stream acquisition unit 22 is used to obtain quality deviation error stream data by subtracting the actual quality deviation data of the currently processed blade from the actual quality deviation data of the previous processed blade based on the blade processing sequence.

[0144] In an optional implementation, the processing mechanism information database includes program processing feature comparison mechanisms and processing experience mechanisms.

[0145] In an optional implementation, the quality deviation prediction module 4 includes:

[0146] The first construction unit 41 is used to construct a machining parameter prediction model corresponding to each quality feature based on the machining mechanism information database and machine tool machining parameter error flow data.

[0147] The first prediction unit 42 is used to obtain the initial machine tool processing parameter error flow data corresponding to the previous processed blade adjacent to the blade to be processed based on the blade processing time sequence, input the initial machine tool processing parameter error flow data into the processing parameter prediction model, and predict the predicted machine tool processing parameter error flow data corresponding to the blade to be processed.

[0148] The second building unit 43 is used to build error correlation models corresponding to each quality feature based on the machining mechanism information database, machine tool machining parameter error flow data and quality deviation error flow data.

[0149] Among them, the error correlation model is used to characterize the correlation between machine tool machining parameter error flow data and quantity deviation error flow data;

[0150] The second prediction unit 44 is used to obtain the predicted quality deviation of the blade to be processed based on the error flow data of the machine tool processing parameters and the error correlation model.

[0151] In an optional embodiment, the first construction unit 41 is used to extract the machine tool parameter error flow sequence dataset for each turbine blade under a preset machining program corresponding to each quality feature based on the machining mechanism information database and machine tool machining parameter error flow data, to obtain the machine tool machining parameter error flow dataset corresponding to each quality feature; divide the machine tool machining parameter error flow dataset according to a preset data ratio to obtain the first machine tool machining parameter error flow dataset and the second machine tool machining parameter error flow dataset; and construct the machining parameter prediction model corresponding to each quality feature by using the first machine tool machining parameter error flow dataset corresponding to the same quality feature as input and the second machine tool machining parameter error flow dataset corresponding to the same quality feature as output.

[0152] In an optional embodiment, the second construction unit 43 is used to extract the machine tool parameter error flow sequence dataset for each turbine blade under each preset machining program corresponding to each quality feature, based on the machining mechanism information database and machine tool machining parameter error flow data, to obtain the machine tool machining parameter error flow dataset corresponding to each quality feature; classify the quality deviation error flow data according to the error range to obtain quality deviation error category labels; wherein, the quality deviation error category label includes the quality deviation error category and the error range corresponding to the quality deviation error category; extract the error category label sequence dataset for each turbine blade under each preset machining program corresponding to each quality feature based on the machining mechanism information database and the quality deviation error category label, to obtain the quality deviation error category label set corresponding to each quality feature; and construct the error association model corresponding to each quality feature by using the machine tool machining parameter error flow dataset corresponding to the same quality feature as input and the quality deviation error category label set corresponding to the same quality feature as output.

[0153] In an optional embodiment, the second prediction unit 44 is used to input the predicted machine tool processing parameter error data corresponding to the target quality feature of the blade to be processed into the error association model with the target quality feature and the corresponding error, and output the predicted quality deviation error category corresponding to the target quality feature of the blade to be processed; obtain the target quality deviation data of the previous processed blade adjacent to the blade to be processed based on the blade processing time sequence; calculate the sum of the error range corresponding to the predicted quality deviation error category and the target quality deviation data to obtain the predicted quality deviation of the target quality feature of the blade to be processed.

[0154] In an optional implementation, the processing parameter prediction model includes a first long short-term memory network layer, a second long short-term memory network layer, and a multi-head attention mechanism network layer cascaded in sequence.

[0155] In one alternative implementation, the error correlation model includes a convolutional neural network layer, a third long short-term memory network layer, a fourth long short-term memory network layer, an activation function layer, and a fully connected network layer, which are cascaded in sequence.

[0156] The turbine blade machining quality prediction system of this embodiment processes the actual machining quality deviation data and actual machine tool machining parameter data of the machined blades to obtain machine tool machining parameter error flow data and quality deviation error flow data. The machine tool machining parameter error flow data and quality deviation error flow data are combined with the machining mechanism information database of the turbine blade machining process. By mining the blade machining mechanism, the correlation between the actual machining machine tool parameters and the actual quality deviation is obtained. Through the correlation, high-precision prediction of the quality deviation of the blade to be machined is achieved.

[0157] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The system embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs.

[0158] Example 3

[0159] Figure 10 This disclosure presents a schematic diagram of the structure of an electronic device in an example embodiment. The electronic device includes a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, it implements the turbine blade processing quality prediction method provided in Embodiment 1 above. Figure 10 The electronic device 90 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments disclosed herein.

[0160] like Figure 10 As shown, the electronic device 90 can be manifested as a general-purpose computing device, such as a server device. The components of the electronic device 90 may include, but are not limited to: at least one processor 91, at least one memory 92, and a bus 93 connecting different system components (including memory 92 and processor 91).

[0161] Bus 93 includes a data bus, an address bus, and a control bus.

[0162] The memory 92 may include volatile memory, such as random access memory (RAM) 921 and / or cache memory 922, and may further include read-only memory (ROM) 923.

[0163] The memory 92 may also include a program tool 925 (or utility) having a set (at least one) program module 924, such program module 924 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0164] The processor 91 executes various functional applications and data processing by running computer programs stored in the memory 92, such as the turbine blade machining quality prediction method provided in Embodiment 1 above.

[0165] Electronic device 90 can also communicate with one or more external devices 94 (e.g., keyboard, pointing device, etc.). This communication can be performed through input / output (I / O) interface 95. Furthermore, electronic device 90 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) via network adapter 96. As shown, network adapter 96 communicates with other modules of electronic device 90 via bus 93. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with electronic device 90, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.

[0166] It should be noted that although several units / modules or sub-units / modules of the electronic device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0167] Example 4

[0168] This disclosure also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the turbine blade machining quality prediction method provided in Embodiment 1 above.

[0169] The readable storage medium may be more specifically adopted, including but not limited to: portable disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0170] Example 5

[0171] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the turbine blade machining quality prediction method provided in Embodiment 1 above.

[0172] The program code for executing the computer program product disclosed herein can be written in any combination of one or more programming languages. The program code can be executed entirely on a user device, partially on a user device, as a stand-alone software package, partially on a user device and partially on a remote device, or entirely on a remote device.

[0173] While specific embodiments of this disclosure have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of this disclosure is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of this disclosure, but all such changes and modifications fall within the scope of protection of this disclosure.

Claims

1. A method for predicting the machining quality of steam turbine blades, characterized in that, The processing quality prediction method includes: Obtain actual quality deviation data of several processed blades of the same type, and actual machine tool processing parameter data of the processed blades within a preset sampling period; Among them, the actual machine tool processing parameter data corresponds to multiple machine tool processing parameters, and the actual processing quality deviation data corresponds to multiple quality characteristics; Based on the blade processing sequence, the actual machine tool processing parameter data, and the actual quality deviation data, obtain machine tool processing parameter error stream data and quality deviation error stream data; Obtain a database of processing mechanism information for the turbine blade manufacturing process; The processing mechanism information database represents the computational relationship between the various quality characteristics of the turbine blade and the preset processing program. Based on the machining mechanism information database, the machine tool machining parameter error flow data, and the quality deviation error flow data, an association relationship is established between the machine tool machining parameter error flow data and the quality deviation error flow data. Based on the association relationship, the predicted quality deviation of the blade to be processed is obtained.

2. The processing quality prediction method according to claim 1, characterized in that, The steps of obtaining machine tool machining parameter error stream data and quality deviation error stream data based on blade machining timing, actual machine tool machining parameter data, and actual quality deviation data include: Based on the blade processing sequence, the actual machine tool processing parameter data of the currently processed blade is subtracted from the actual machine tool processing parameter data of the previous processed blade to obtain the machine tool processing parameter error stream data. Based on the blade processing sequence, the actual quality deviation data of the currently processed blade is successively subtracted from the actual quality deviation data of the previous processed blade to obtain the quality deviation error stream data; And / or, the processing mechanism information database includes program processing feature comparison mechanisms and processing experience mechanisms.

3. The processing quality prediction method according to claim 1, characterized in that, The step of obtaining the predicted quality deviation of the blade to be processed based on the machining mechanism information database, the machine tool machining parameter error flow data, and the quality deviation error flow data, and the correlation between the machine tool machining parameter error flow data and the quality deviation error flow data, includes: Based on the machining mechanism information database and the machine tool machining parameter error flow data, a machining parameter prediction model corresponding to each quality feature is constructed. Based on the blade processing timing, the initial machine tool processing parameter error flow data corresponding to the previous processed blade adjacent to the blade to be processed is obtained, and the initial machine tool processing parameter error flow data is input into the processing parameter prediction model to predict the predicted machine tool processing parameter error flow data corresponding to the blade to be processed. Based on the machining mechanism information database, the machine tool machining parameter error flow data, and the quality deviation error flow data, an error correlation model corresponding to each quality feature is constructed. The error correlation model is used to characterize the correlation between the machine tool machining parameter error stream data and the quantity deviation error stream data. Based on the predicted machine tool machining parameter error flow data and the error correlation model, the predicted quality deviation corresponding to the blade to be processed is obtained.

4. The processing quality prediction method according to claim 3, characterized in that, The steps of constructing a machining parameter prediction model corresponding to each quality feature based on the machining mechanism information database and the machine tool machining parameter error flow data include: Based on the machining mechanism information database and the machine tool machining parameter error flow data, the machine tool parameter error flow sequence dataset for each turbine blade under the preset machining program corresponding to each quality feature is extracted to obtain the machine tool machining parameter error flow dataset for each quality feature. The machine tool machining parameter error stream dataset is divided according to a preset data ratio to obtain a first machine tool machining parameter error stream dataset and a second machine tool machining parameter error stream dataset. Using the first machine tool machining parameter error stream dataset corresponding to the same quality feature as input and the second machine tool machining parameter error stream dataset corresponding to the same quality feature as output, a machining parameter prediction model corresponding to each quality feature is constructed. And / or, the step of constructing an error correlation model corresponding to each quality feature based on the machining mechanism information database, the machine tool machining parameter error stream data, and the machining quality deviation error stream data includes: Based on the machining mechanism information database and the machine tool machining parameter error flow data, the machine tool parameter error flow sequence dataset for each turbine blade under the preset machining program corresponding to each quality feature is extracted to obtain the machine tool machining parameter error flow dataset for each quality feature. The quality deviation error stream data is classified according to the error range to obtain quality deviation error category labels; The quality deviation error category label includes a quality deviation error category and the error range corresponding to the quality deviation error category; Based on the processing mechanism information database and the quality deviation error category label, the error category label sequence dataset of each turbine blade under the preset processing program corresponding to each quality feature is extracted to obtain the quality deviation error category label set corresponding to each quality feature; Using the machine tool machining parameter error flow dataset corresponding to the same quality feature as input and the quality deviation error category label set corresponding to the same quality feature as output, the error correlation model corresponding to each quality feature is constructed.

5. The processing quality prediction method according to claim 3, characterized in that, The step of obtaining the predicted quality deviation corresponding to the blade to be processed based on the predicted machine tool processing parameter error data and the error correlation model includes: The predicted machine tool processing parameter error data corresponding to the target quality characteristics of the blade to be processed is input into the error association model corresponding to the target quality characteristics, and the predicted quality deviation error category corresponding to the target quality characteristics of the blade to be processed is output. Based on the blade processing sequence, obtain the target quality deviation data of the previous processed blade adjacent to the blade to be processed; The predicted quality deviation of the blade to be processed is obtained by summing the error range corresponding to the predicted quality deviation error category with the target quality deviation data.

6. The processing quality prediction method according to claim 3, characterized in that, The processing parameter prediction model includes a first long short-term memory network layer, a second long short-term memory network layer, and a multi-head attention mechanism network layer cascaded in sequence. And / or, the error correlation model includes a convolutional neural network layer, a third long short-term memory network layer, a fourth long short-term memory network layer, an activation function layer, and a fully connected network layer, which are cascaded in sequence.

7. A machining quality prediction system for steam turbine blades, characterized in that, The processing quality prediction system includes: The actual data acquisition module is used to acquire actual quality deviation data of several processed blades of the same type, and actual machine tool processing parameter data of the processed blades within a preset sampling period. Among them, the actual machine tool processing parameter data corresponds to multiple machine tool processing parameters, and the actual processing quality deviation data corresponds to multiple quality characteristics; The error stream data acquisition module is used to acquire machine tool processing parameter error stream data and quality deviation error stream data based on the blade processing sequence, the actual machine tool processing parameter data, and the actual quality deviation data. The processing mechanism information acquisition module is used to acquire the processing mechanism information database of the turbine blade processing process; The processing mechanism information database represents the computational relationship between the various quality characteristics of the turbine blade and the preset processing program. The quality deviation prediction module is used to establish a correlation between the machine tool machining parameter error flow data and the quality deviation error flow data based on the machining mechanism information database, the machine tool machining parameter error flow data and the quality deviation error flow data, and to obtain the predicted quality deviation of the blade to be processed based on the correlation.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, characterized in that, When the processor executes the computer program, it implements the method for predicting the machining quality of turbine blades as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for predicting the machining quality of turbine blades as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for predicting the machining quality of turbine blades as described in any one of claims 1 to 6.

Images