Position prediction method for aero-engine blade milling based on geometric features

Abstract

The present invention discloses a method for predicting the position accuracy of aeroengine blade milling based on geometric features, comprising the following steps: determining key machining parameters and collecting current data during the machining process; importing a blade model into CAD software, parameterizing the blade's curved surface, establishing a UV coordinate system, evenly arranging detection points, detecting the position accuracy at each detection point, and calculating the curvature of each detection point along a V-direction streamline using the CAD software; calculating the residual height h at each detection point, and forming a data set from the corresponding current data, curvature along a V-direction streamline at the detection point, residual height at the detection point, and position accuracy under different key machining parameters; constructing a neural network model; and using the data set to train the neural network model for predicting the position accuracy of aeroengine blade milling. The present invention not only learns complex nonlinear relationships from machining data but also improves the model's generalization and prediction accuracy through geometric features.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent manufacturing and relates to a method for predicting the position of aero-engine blade milling based on geometric features. Background Art

[0002] As a key component of aircraft engines, the machining accuracy of aircraft engine blades is crucial to overall engine performance. Due to their complex structure, diverse shapes, and stringent dimensional requirements, positional deviations during machining directly impact the engine's aerodynamic performance, combustion efficiency, and service life. Traditional methods for predicting blade machining positional deviations rely on extensive experiments or empirical rules. These methods struggle to accurately account for various factors involved in machining complex surfaces, such as workpiece shape, tool wear, and the dynamic characteristics of the machine tool.

[0003] With the development of intelligent manufacturing technology, research based on data-driven methods has become an effective means to solve the problem of blade machining position prediction. By utilizing a large amount of historical machining data, the neural network prediction model can automatically learn and capture the hidden laws in the machining process, and achieve accurate prediction of position deviation. However, purely data-driven models often require a large amount of historical data, and the prediction accuracy is limited when faced with small sample data; and the use of only machining data ignores the influence of the bending and torsional characteristics of the workpiece on the position deviation, further limiting the prediction accuracy of the model. To this end, this patent proposes a method for predicting the position of aero-engine blade milling based on geometric features.

[0004] By incorporating geometric features into intelligent algorithms, not only is the size of the data set expanded, but the network model can also learn the bending and torsional characteristics of the blades, thereby improving the overall efficiency and quality of aero-engine manufacturing. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention provides a method for predicting the position accuracy of aeroengine blade milling processes based on geometric features. This method fully considers the bending and torsional characteristics of aeroengine blades, calculates the curvature at each point, and then calculates the residual height based on this curvature and machining parameters. This is then incorporated into a neural network as domain knowledge, expanding the dataset size and improving the network's prediction accuracy.

[0006] To achieve the above object, the technical solution of the present invention is:

[0007] A method for predicting the position accuracy of aero-engine blade milling based on geometric features, the method comprising the following contents:

[0008] Determine four key machining parameters: spindle speed, feed rate, cutting depth, and path spacing. Each key machining parameter is varied at different levels to form an orthogonal experimental table. Machining is performed according to the orthogonal experimental table, and current data during the machining process is collected.

[0009] Import the blade model into the CAD software, parameterize the blade surface, and establish a UV coordinate system, where the V direction is developed along the long axis of the blade and the U direction is developed along the chord direction of the blade. Evenly arrange the inspection points and inspect the position of each inspection point. Generate the V-direction streamline at each inspection point using the surface feature function in the CAD software, and calculate the curvature of each inspection point along the V-direction streamline using the CAD software.

[0010] The residual height h at each detection point is calculated according to the following formula:

[0011]

[0012] Where h is the residual height at the detection point, ρ is the curvature of the V-direction streamline at the detection point, R is the radius of the ball-end milling cutter, and I is the path spacing;

[0013] The corresponding current data under different key processing parameters, the curvature of the V-direction streamline at the detection point, the residual height of the detection point and the position of the corresponding detection point are combined into a data set;

[0014] Build a neural network model:

[0015] The neural network model includes a CNN-LSTM network and a DNN network. The current data is used as the input of the CNN-LSTM network, and the residual height is used as the input of the DNN network. The output of the CNN-LSTM network is spliced ​​and fused with the key processing parameters and the curvature along the V-direction streamline. After that, the first feature result is obtained by processing and predicting through the first fully connected layer. The residual height is used as domain knowledge and the second feature result is obtained after dimensionality upgrade through the DNN network. The first feature result and the second feature result are spliced ​​and then processed and predicted through the second fully connected layer to obtain the final position.

[0016] The dataset is used to train the neural network model to obtain a trained neural network model for position prediction in aero-engine blade milling.

[0017] Furthermore, the CNN-LSTM network is composed of a CNN and an LSTM in series. The current data is input into the CNN to obtain a low-dimensional local feature representation, and then the time-dependent feature n1 in the sequence is extracted through the LSTM in the time dimension.

[0018] Then, a new feature n2 is obtained by combining n1 with key processing parameters and curvature along the V-direction streamline using a stitching layer.

[0019] Pass n2 through a fully connected layer with 10 neurons for preliminary prediction to obtain the first feature result n3;

[0020] After the residual height is subjected to a nonlinear transformation operation through a DNN network with three hidden layers, the second feature result n4 is obtained; the first feature result n3 and the second feature result n4 are concatenated through a concatinate layer to obtain n5; n5 is calculated through a fully connected layer with 10 neurons to obtain the final result.

[0021] Furthermore, the numbers of neurons in the three hidden layers are 100, 100, and 10 respectively.

[0022] Furthermore, the data set is divided into training sets and the root mean square error RMSE is used as the loss function when training the neural network model.

[0023]

[0024] Among them, RMSE is the root mean square error, y predi is the position value of the i-th detection point predicted by the neural network model, y truei is the actual position value of the i-th detection point, and n is the number of detection points;

[0025] Draw the loss curve during training and observe the changing trend of the loss value to determine whether the model has converged and determine the appropriate number of training rounds.

[0026] The present invention also protects a computer-readable storage medium having a computer program stored thereon, which can implement the steps of the method when the program is executed by a processor.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] By combining geometric features with neural networks, the network model of this invention not only learns complex nonlinear relationships from processing data but also enhances the model's generalization and prediction accuracy through geometric features. The residual height is processed through a deep learning network (DNN) for dimensionality upscaling and effectively integrated with the relevant features of the extracted current data to generate a position estimation model with high predictive accuracy. This model demonstrates strong adaptability and robustness under various processing conditions.

[0029] The present invention fully considers the differences between different points, calculates the curvature along the V-direction streamline at each point, and calculates the residual height based on the curvature and key processing parameters as prior knowledge and integrates it into the neural network, expanding the scale of the data set and improving the prediction accuracy of the network.

[0030] The method of the present invention not only improves the predictive reliability of blade processing accuracy, but also enhances the precision control capability under complex working conditions to adapt to complex working conditions or small sample scenarios, and can provide strong support and guarantee for the high-precision manufacturing of aircraft engine blades. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a structural schematic diagram of the neural network model in the method for predicting the position accuracy of aero-engine blade milling based on geometric features of the present invention.

[0032] Figure 2 This is a diagram showing the distribution of detection points and the establishment of the coordinate system in the present invention.

[0033] Figure 3 This is a diagram of the training loss convergence process in the embodiment.

[0034] Figure 4 This is a comparison chart of the prediction results accuracy.

[0035] Figure 5 To evaluate the overall effect of the prediction results. DETAILED DESCRIPTION

[0036] In order to make the technical solution of the present invention clearer, the present invention is further described below with reference to the accompanying drawings.

[0037] The present invention provides a method for predicting the position accuracy of aero-engine blade milling based on geometric features, which is implemented in the following steps:

[0038] Step 1: Experiment Design. Four key machining parameters were determined: spindle speed, feed rate, depth of cut, and path spacing. Each key machining parameter was varied at different levels to form an orthogonal experimental table. This orthogonal experimental method was used to evaluate the impact of each key machining parameter on the accuracy of five-axis milling of aircraft engine blades.

[0039] a. Determine the four key machining parameters: spindle speed, feed rate, cutting depth, and path spacing.

[0040] b. Set different levels for each selected key processing parameter to fully cover the possible working range. According to the principle of orthogonal experimental design, construct an orthogonal experimental matrix and ensure that all key processing parameter combinations are fully and reasonably investigated.

[0041] Step 2: Data Collection. Perform machining according to the orthogonal experimental table in Step 1; collect current data during machining; arrange inspection points and check the position accuracy at each point; calculate the curvature along the V-direction streamline at each inspection point and calculate the residual height at the inspection point.

[0042] (1) Carry out processing one by one according to the designed experimental combination to ensure strict control of each processing parameter.

[0043] (2) Collect current data.

[0044] a. Determine the sampling frequency of current data to ensure the timeliness and accuracy of the data. The sampling frequency should be no less than five times per second. The current data of each blade during the machining process should be collected based on the CNC system of the machine tool.

[0045] b. Clearly define the collection time periods during each blade processing, including the start, middle, and end of processing.

[0046] (3) Collecting position data. Use 8 U-direction streamlines and 16 V-direction streamlines to evenly divide the blade surface. The U-direction streamlines are the tool feed direction. The intersection of the U-direction streamlines and the V-direction streamlines is the position detection point. For each blade, at least 64 detection points are evenly selected on the blade base (concave surface) and the blade back (convex surface) to obtain the actual processing results of the blade. In the embodiment, 64 detection points are distributed on the blade base and blade back of each group of blades, and a total of 128 points are detected.

[0047] (4) Import the blade model. Import the blade model into the CAD software. Common formats include STEP, IGES, STL, etc. to ensure that the model can be used in the CAD software.

[0048] (5) Determine the UV coordinate system. The surface of the blade is usually a complex three-dimensional surface. In order to calculate the streamlines and curvature of the surface, the surface is parameterized and a UV coordinate system is established. The V direction is developed along the long axis of the blade, and the U direction is developed along the chord direction of the blade. The established UV coordinate system is as follows: Figure 2 As shown in the figure, the intersection of U and V directions is the detection point of position accuracy.

[0049] The actual machining results of the blade are compared with its corresponding 3D CAD model to obtain the position accuracy of these inspection points under the same key machining parameters.

[0050] (6) Generate V-direction streamlines. Generate V-direction streamlines along each inspection point using the surface feature function in the CAD software.

[0051] (7) The curvature of each detection point along the V-direction streamline is calculated using CAD software. In this embodiment, nine blades with a total of 1152 detection points are constructed. The curvature matrix of the 1152 detection points is expressed as [ρ1, ρ2, ..., ρ1152 ].

[0052] (8) Calculation of residual height.

[0053] The residual height h at each detection point is calculated according to the following formula: the residual height matrix [h1,h2,…h 1152 ].

[0054]

[0055] Where h is the residual height at the detection point, ρ is the curvature of the V-direction streamline at the detection point, R is the radius of the ball-end milling cutter, and I is the path spacing;

[0056] The current data corresponding to different key processing parameters, the curvature along the V-direction streamline at the detection point, the residual height of the detection point, and the position of the corresponding detection point are combined into a data set; the key processing parameters, current data, curvature along the V-direction streamline, residual height, and position corresponding to each detection point constitute a sample. The key processing parameters and current data of the same blade are the same.

[0057] The third step is to build a neural network model, including a CNN-LSTM network and a DNN network. The current data is used as the input of the CNN-LSTM network, and the residual height calculated according to the curvature along the V-direction streamline is used as the input of the DNN network. The output of the CNN-LSTM network is spliced ​​and fused with the key processing parameters and the curvature along the V-direction streamline, and then processed and predicted by the first fully connected layer (at least one layer) to obtain the first feature result; the residual height is used as domain knowledge and the second feature result is obtained after dimensionality upgrade through the DNN network. The first feature result and the second feature result are spliced ​​and then processed and predicted by the second fully connected layer to obtain the final position.

[0058] The number of hidden layers of the DNN network in the present invention is at least two, preferably 3-5.

[0059] Taking into account the problem of small amount of data in aero-engine blade milling, the present invention calculates the residual height at each detection point as domain knowledge and embeds it into the neural network, thereby enhancing the interpretability of the neural network and improving the prediction accuracy of the neural network.

[0060] The specific data processing process of the neural network model is:

[0061] (1) Feature extraction module: CNN-LSTM is used to extract features from current data.

[0062] a. The input current data first passes through the convolutional neural network (CNN) to obtain a low-dimensional local feature representation.

[0063] b. Use the output of CNN as the input of LSTM, and extract the time-dependent feature n1 in the sequence through LSTM in the time dimension.

[0064] (2) Feature fusion. The key processing parameters and the blade morphology have an important influence on the processing quality. Therefore, the features extracted from the current data are fused with the key processing parameters and the curvature along the V-direction streamline using a splicing layer, and a preliminary prediction is made.

[0065] a. The key machining parameters (spindle speed, feed rate, cutting depth, path spacing) and the curvature along the V-direction streamline are spliced ​​into n1 to obtain a new feature n2.

[0066] b. The new feature n2 obtained by splicing is preliminarily predicted through a fully connected layer with 10 neurons to obtain the preliminary prediction result n3

[0067] (3) Embedding domain knowledge. A deep neural network is used to implement nonlinear transformation of the residual height. The nonlinearly transformed features are combined with the initial prediction results to make them linearly differentiable in the high-dimensional output space, greatly improving the ability to handle complex nonlinear problems. The residual height after nonlinear transformation is spliced ​​with the preliminary prediction results to prepare for the final prediction.

[0068] a. Apply a nonlinear transformation to the residual height from the first step through a DNN with three hidden layers to obtain feature n4. The number of neurons in the three hidden layers is 100, 100, and 10, respectively.

[0069] b. Concatenate the initial prediction result n3 and the residual height result n4 after nonlinear transformation through the Concatnate layer to obtain n5

[0070] (4) Calculate the final result. Pass n5 through a fully connected layer with 10 neurons to obtain the final result.

[0071] The fourth step is to divide the data set into training set and test set according to a certain ratio, and train the neural network model in the third step.

[0072] a. Define the loss function: For the position prediction task, the root mean square error (RMSE) is selected as the loss function, which is the square root of the difference between the model prediction value and the actual value.

[0073]

[0074] The loss function is used to measure the difference between the model's predictions and the actual results.

[0075] RMSE is the root mean square error, y predi is the position value of the i-th detection point predicted by the neural network model, ytruei is the actual position value of the i-th detection point, n is the number of detection points, that is, the number of samples;

[0076] b. Select an optimization algorithm: Choose an appropriate optimization algorithm to update the model parameters. Common optimization algorithms include stochastic gradient descent (SGD), Adam, and Adagrad. Select an appropriate optimization algorithm based on the actual situation and data characteristics. This example uses the Adam optimization algorithm to update the model parameters.

[0077] c. Divide the dataset: Divide the dataset into training, validation, and test sets. The training set is used to update model parameters, the validation set is used to adjust hyperparameters, and the test set is used to evaluate model performance.

[0078] d. Adjust hyperparameters: Tune the model's hyperparameters, including learning rate, batch size, number of hidden layer nodes, etc. You can use methods such as grid search, random search, or Bayesian optimization to search for the optimal hyperparameter combination.

[0079] e. Train the model: Use the training set to train the model and monitor the change in loss on the validation set. Adjust the hyperparameters based on the loss on the validation set until the optimal hyperparameter combination is found.

[0080] f. Evaluate the model: Evaluate the model's performance on the test set, calculating at least one of the mean square error (MSE) and mean average error (MAE) values ​​for position predictions as evaluation criteria. Use the evaluation results to determine the model's accuracy and make necessary adjustments and improvements.

[0081] g. Visualize the loss curve: Draw the loss curve of the model during training and observe the changing trend of the loss value to determine whether the model has converged and determine the appropriate number of training rounds.

[0082] Example:

[0083] The steps of the method for predicting the position accuracy of aero-engine blade milling based on geometric features in this embodiment are as follows:

[0084] Step 1: Experiment Design. Four key machining parameters were determined: spindle speed, feed rate, depth of cut, and path spacing. Each key machining parameter was varied at different levels to form an orthogonal experimental table. This orthogonal experimental method was used to evaluate the impact of each key machining parameter on the accuracy of five-axis milling of aircraft engine blades.

[0085] a. Determine the four key machining parameters: spindle speed, feed rate, cutting depth and path spacing.

[0086] b. Set different levels for each selected key machining parameter to fully cover the possible operating range. The spindle speed range is [13,000-15,000] rpm, the feed rate range is [1,000-2,000] mm / min, the depth of cut range is [0.12-0.14] mm, and the path spacing range is [0.07-0.09]. A four-factor, three-level orthogonal experiment was designed. The orthogonal experiment table is shown in Table 1.

[0087]

[0088] Step 2: Data collection.

[0089] (1) Process each experimental combination according to the orthogonal experimental table to ensure strict control of each processing parameter.

[0090] (2) Collect current data.

[0091] a. Determine the sampling frequency of current data to ensure the timeliness and accuracy of the data. Sampling is done five times per second. The current data of each blade during the machining process is collected based on JDMon_V1.5.

[0092] b. Clearly define the collection time periods during each blade processing, including the start, middle, and end of processing.

[0093] (3) Collecting position data. The blade surface is evenly divided using 8 U-direction streamlines and 16 V-direction streamlines. The U-direction streamlines are the tool feed direction. The intersection of the U-direction streamlines and the V-direction streamlines is the position detection point. 64 detection points are distributed on the blade base and back of each blade group, for a total of 128 detection points.

[0094] (4) Calculate the curvature of each inspection point along the V-direction streamline using CAD software; calculate the residual height at 128 inspection points using the following formula:

[0095] ;

[0096] The third step is to build a neural network model. First, the features extracted by the CNN-LSTM are integrated with key processing parameters and the curvature along the V-direction streamline. After several layers of fully connected layers, a feature result is predicted. Then, the residual height is used as domain knowledge and dimensionally upgraded through a deep neural network (DNN) to obtain a second feature result. These two feature results are concatenated to predict the final position.

[0097] (1) Feature extraction module: CNN-LSTM is used to extract features from current data.

[0098] a. The input current data is first passed through a convolutional neural network to obtain a low-dimensional local feature representation. A CNN is constructed using TensorFlow. Experimental results show that the best performance is achieved when filters = 1, kernel_size = 4, and pool_size = 3.

[0099] b. Use the CNN output as the input to the LSTM network, extracting the time-dependent feature n1 from the sequence in the time dimension. Using TensorFlow to construct the LSTM, the optimal setting of lstm_units=4 is achieved.

[0100] (2) Feature fusion. Considering only the current data ignores the impact of key processing parameters and blade morphology on processing quality. Therefore, the features extracted from the current data are fused with the key processing parameters and the curvature along the V-direction streamline using a splicing layer, and a preliminary prediction is made.

[0101] a. Concatenate the vector P containing key machining parameters (spindle speed, feed rate, depth of cut, path spacing) and the curvature along the V-axis streamline to n1 to create a new feature n2. Use the concatenate function in the Keras functional API to combine P and n1, setting axis=1 to concatenate the two vectors along the first dimension.

[0102]

[0103] b. The new feature n2 obtained by splicing is preliminarily predicted through a fully connected layer with 10 neurons to obtain the preliminary prediction result n3

[0104] (3) Embedding domain knowledge. A deep neural network is used to implement nonlinear transformation of the residual height. The nonlinearly transformed features are combined with the initial prediction results to make them linearly differentiable in the high-dimensional output space, greatly improving the ability to handle complex nonlinear problems. The residual height after nonlinear transformation is spliced ​​with the preliminary prediction results to prepare for the final prediction.

[0105] a. The residual height in the first step is transformed nonlinearly through a deep neural network (DNN) with three hidden layers. The three hidden layers contain 100, 100, and 10 neurons respectively. Feature n4 is obtained through the DNN.

[0106] b. Concatenate n3 and n4 through the Concatnate layer to get n5

[0107]

[0108] (4) Calculate the final result. Pass n5 through a fully connected layer with 10 neurons to obtain the final result.

[0109] The fourth step is position prediction.

[0110] The neural network model was trained by splitting the 1,152 samples into training and test sets with a 9:1 ratio. The following two evaluation metrics were used to assess the predictive effectiveness of the model: MSE (mean squared error) and MAE (mean absolute error).

[0111]

[0112]

[0113] from Figure 4 As can be seen, the predicted value and the true value are very close. In most cases, the two curves almost overlap, which shows that the prediction results of the model are very consistent with the actual situation, indicating that the model can accurately capture the characteristics and laws of the data; and for most data points, the error between the predicted value and the true value is very small, indicating that the model has high stability and reliability. Figure 3 The different lines in the figure represent the changes in training loss on the training set and validation set. It can be seen that the training loss continues to decrease with the increase in batches and is basically stable after 60 times, indicating that the neural network model of the present invention can converge quickly, verifying the effectiveness of the model design. Figure 5 It can be seen that the MSE of the model of the present invention is 0.0016 and the MAE is 0.025. In comparison, the MSE of the CNN-LSTM network that only considers current data and processing parameters is 0.003 and the MAE is 0.0038, which proves that the model of the present invention has high accuracy and can accurately estimate the accuracy of different inspection points, which is conducive to timely adjustment of areas with large local deviations in the later stage.

[0114] Any matters not described in the present invention are applicable to the prior art.

Claims

1. A method for predicting the position accuracy of aero-engine blade milling based on geometric features, characterized in that: The prediction method Includes the following: Determine four key machining parameters: spindle speed, feed rate, cutting depth, and path spacing. Each key machining parameter is varied at different levels to form an orthogonal experimental table. Machining is performed according to the orthogonal experimental table, and current data during the machining process is collected. Import the blade model into the CAD software, parameterize the blade surface, and establish a UV coordinate system, where the V direction is developed along the long axis of the blade and the U direction is developed along the chord direction of the blade. Evenly arrange the inspection points and inspect the position of each inspection point. Generate the V-direction streamline at each inspection point using the surface feature function in the CAD software, and calculate the curvature of each inspection point along the V-direction streamline using the CAD software. The residual height h at each detection point is calculated according to the following formula: Where h is the residual height at the detection point, ρ is the curvature of the V-direction streamline at the detection point, R is the radius of the ball-end milling cutter, and I is the path spacing; The corresponding current data under different key processing parameters, the curvature of the V-direction streamline at the detection point, the residual height of the detection point and the position of the corresponding detection point are combined into a data set; Build a neural network model: The neural network model includes a CNN-LSTM network and a DNN network. The current data is used as the input of the CNN-LSTM network, and the residual height is used as the input of the DNN network. The output of the CNN-LSTM network is spliced ​​and fused with the key processing parameters and the curvature along the V-direction streamline. After that, the first feature result is obtained by processing and predicting through the first fully connected layer. The residual height is used as domain knowledge and the second feature result is obtained after dimensionality upgrade through the DNN network. The first feature result and the second feature result are spliced ​​and then processed and predicted through the second fully connected layer to obtain the final position. The dataset is used to train the neural network model to obtain a trained neural network model for position prediction in aero-engine blade milling.

2. The method according to claim 1, characterized in that The CNN-LSTM network is composed of CNN and LSTM in series. The current data is input into CNN to obtain a low-dimensional local feature representation, and then the time-dependent feature n1 in the sequence is extracted through LSTM in the time dimension. Then, a new feature n2 is obtained by combining n1 with key processing parameters and curvature along the V-direction streamline using a stitching layer. Pass n2 through a fully connected layer with 10 neurons for preliminary prediction to obtain the first feature result n3; After the residual height is subjected to a nonlinear transformation operation through a DNN network with three hidden layers, the second feature result n4 is obtained; the first feature result n3 and the second feature result n4 are concatenated through a concatinate layer to obtain n5; n5 is calculated through a fully connected layer with 10 neurons to obtain the final result.

3. The method according to claim 2, characterized in that The number of neurons in the three hidden layers are 100, 100, and 10 respectively.

4. The method according to claim 1, wherein When dividing the data set into training sets and training the neural network model, the root mean square error RMSE is used as the loss function. Among them, RMSE is the root mean square error, y predi is the position value of the i-th detection point predicted by the neural network model, y truei is the actual position value of the i-th detection point, and n is the number of detection points; Draw the loss curve during training and observe the changing trend of the loss value to determine whether the model has converged and determine the appropriate number of training rounds.

5. The method according to claim 1, wherein The prediction accuracy of the neural network model is: MSE is 0.0016, MAE is 0.

025.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 can be implemented.

Images