Intelligent processing method for injection hole of liquid rocket engine

Abstract

The present application belongs to the field of mechanical engineering, and discloses an intelligent processing method for injection holes of a liquid rocket engine. The system used in the intelligent processing method comprises a five-axis numerical control processing machine tool configured with a readable data information system, a line laser sensor capable of measuring object height information, a force measuring tool holder capable of reading force signals in real time during processing, etc. The rocket engine injection disc is fixed on the numerical control machine tool workbench through a clamp to prevent position changes during processing. A high-precision tool is used to connect the line laser sensor with the tool holder, and the injection hole three-dimensional point cloud information is obtained by scanning the workpiece. Processing parameter solving is performed through point cloud reconstruction. The force measuring tool holder is used for processing, and the processing process is regulated according to the real-time force signals. The intelligent processing method proposed by the present application eliminates the problem that the processing parameters of the workpiece are difficult to determine in the traditional processing method, and can significantly improve the processing precision and reduce the labor intensity of workers.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of mechanical engineering and relates to an intelligent machining method for injection holes of a liquid rocket engine. BACKGROUND

[0002] As a key component of a liquid rocket engine, the injection disc has hundreds of injection holes arranged in a ring array. The machining precision of the injection holes can directly affect the fuel atomization combustion efficiency and the stability of the rocket flight attitude. The injection holes are formed by brazing a copper base and a steel nozzle. The machining precision of the array holes of the copper base is low, the consistency of the assembly posture of the copper base and the steel nozzle is poor, and the solder flows in the gap between the base and the steel pipe during the brazing process. Due to the above factors, the spatial pose of the injection holes presents a differential irregular deviation from the theoretical coordinates, and the heights of the chamfers to be machined of the steel pipe parts of the injection holes are different, so the conventional numerical control machining method based on the design model to generate machining codes cannot be used. The chamfer machining of the injection holes is generally completed by manual operation of a radial drilling machine, which requires a high level of experience of workers, has a long machining period, and it is difficult to ensure the consistency of products. Therefore, there is an urgent need for an intelligent machining method for the injection holes of a rocket engine that can improve the production efficiency and the overall machining quality of the injection disc.

[0003] Zhejiang University discloses in the invention patent "A fine machining method for an outer wing butt joint intersection hole", CN113000881A, a fine machining method for an outer wing butt joint intersection hole. Although this method can provide high measurement accuracy, it has complex operation steps, more manual operation steps, and low processing speed. China Architecture Eighth Engineering Corporation Limited discloses in the invention patent "Chamfer machining device and chamfer machining method for connecting holes on connecting plates", CN118060632A, a method for chamfer machining of connecting holes on connecting plates. The spatial coordinates of the connecting holes are strictly arranged in an array, and the chamfer machining amount of each hole is a constant value. Xi'an Qunjian Aviation Precision Manufacturing Co., Ltd. discloses in the invention patent "Array hole machining method", CN113449393A, a method for obtaining first and second coordinate values corresponding to each array hole, determining the respective offset values, and then machining the array holes. This method actually machined according to the theoretical position and did not consider the pose deviation of the hole position.

[0004] The above researches do not mention an intelligent machining method for the injection holes of a liquid rocket engine. SUMMARY

[0005] The present application mainly aims at the problems of strong artificial dependence, long processing period and low precision caused by individual differences of injection holes in the existing liquid rocket engine injection hole processing process, and proposes an intelligent processing method for liquid rocket engine injection holes. The method uses a line laser sensor to obtain the actual spatial pose of the injection holes of the injection disc, calculates the to-be-processed feature parameters of the injection holes based on a point cloud reconstruction method, and realizes the automatic generation of processing G code relying on the secondary development function module of the numerical control system; a cutting force trend prediction model is trained to assist in realizing real-time feedback control of the cutting process.

[0006] The technical scheme of the present application:

[0007] Firstly, the workpiece, i.e. the liquid rocket engine injection disc, is fixed on the workbench surface of the five-axis numerical control machine tool by the self-adaptive centering clamp, and the relative positional relationship between the workpiece coordinate system and the machine tool coordinate system is determined;

[0008] Secondly, the injection disc is scanned at multiple angles by using a line laser sensor to ensure the integrity of the three-dimensional point cloud data of the injection holes, which is not limited by the structure of the injection disc and is not blocked;

[0009] Subsequently, the spatial geometric features of the to-be-processed part are accurately analyzed by point cloud reconstruction technology, and the actual center coordinates of the injection holes and the to-be-processed chamfer height information are obtained;

[0010] Thirdly, based on the calculated processing feature parameters, the automatic generation of processing G code is realized by means of the secondary development interface of the numerical control system;

[0011] Finally, the cutting force signal is collected in real time during the processing by using a force measuring tool holder, and the cutting force data is dynamically analyzed by combining a deep learning model, the processing parameters are adaptively adjusted to optimize the processing state, and the intelligent monitoring and fine control of the whole process are realized.

[0012] The specific steps are as follows:

[0013] First step: use the self-adaptive centering clamp for workpiece positioning and clamping

[0014] The self-adaptive centering clamp mainly consists of two semicircular structures of fixed end and movable end; during the workpiece clamping process, first, the outer wall surface of the liquid rocket engine injection disc is tightly fitted against the fixed end, and then the movable end is closed and tightened by screws;

[0015] Second step: line laser scanning measurement

[0016] The line laser sensor establishes communication with the machine tool numerical control system through a network cable, and is integrated at the end of the machine tool spindle by using a tool holder; after completing the pose calibration of the line laser sensor, the machine tool spindle is locked by executing the "SPOS" instruction; in order to ensure the integrity of the three-dimensional point cloud data of each injection hole, a multi-directional line laser scanning measurement trajectory is designed;

[0017] Step 3: Solving for relevant processing quantities

[0018] Select the smallest bounding box Ω that can completely cover the geometric features of a single injection hole, and extract the point cloud data within the bounding box to construct a sub-point cloud N; group the coordinate points within the sub-point cloud N according to the different laser scanning lines they belong to, and number them P1, P2…P i …P n , where n is the total number of laser scanning lines contained within the acquired workpiece point cloud, from P i Starting with every 3 consecutive laser scan lines, data P i+1 P i+2 P i+3 Form a data matrix D i The data matrix D after convolution is obtained by performing a convolution operation using a 3×3 kernel K. i ';

[0019] D i '=D i *K (1)

[0020] Several data points on the upper surface of the steel nozzle with injection holes were manually labeled as labels H, and end-to-end training was performed using a multilayer perceptron; the input data was a data matrix D' = (D'1, D'2, ..., D'') obtained from processing different injection holes. i ,..D' n-2 The output data is the data point set of the upper surface of the injection hole steel nozzle. Data point set The amount of data is consistent with the label H; the mean square error function value (MSE) is used as the standard for calculating the loss function and training parameters of the neural network;

[0021]

[0022] Where i is the data point set The number of data points corresponding to label H; using the trained neural network, obtain the data point set P = {P} on the upper surface of a single injection hole steel nozzle. j (x j ,y j ,z j ), j∈[1,n]}, calculate the average height of all points as the height h of the upper surface of the injection hole. top ;

[0023]

[0024] Based on the data point set P on the upper surface of the injection hole steel nozzle, find the outermost edge point set S of the injection hole steel nozzle. side Extending it outwards, we obtain the cone-shaped point set S of the injection hole copper substrate. cone ;

[0025] S1: solving the center coordinates of the injection hole

[0026] In the outermost circle edge point set S of the injection hole steel nozzle side Randomly select three points s1(x1, y1), s2(x2, y2), s3(x3, y3) to form an initial sample, and use the least square fitting algorithm for circle fitting. The circle fitting equation is:

[0027]

[0028] Where (x c ,y c ) is the center coordinates of the circle to be fitted, and R is the radius of the circle to be fitted.

[0029] Calculate the error of other points in the outermost circle edge point set S of the injection hole steel nozzle side with the fitted circle. For any point s i (x i ,y i ), if the error is within the preset radius threshold δ of the fitted circle center, it is considered as an inner point. The judgment basis is as follows:

[0030]

[0031] e i <δ (6)

[0032] Repeat the above steps several times, select the fitting circle with the most inner points as the final result, and get the center coordinates o(x0, y0) as the center coordinates of the injection hole.

[0033] S2: solving the chamfer height of the injection hole

[0034] Take the upper surface point cloud of the injection hole steel nozzle and use the least square method to fit the first straight line. Use the least square method to fit the second straight line with the side surface point cloud. Get the third straight line by taking the intersection point of the first straight line and the second straight line perpendicular to the horizontal plane. Take the points on the copper base cone surface and use the least square method to fit the fourth straight line. The intersection point of the fourth straight line and the third straight line is the lowest point, and the lowest point height is the lower surface height h bottom . The injection hole nozzle height is represented as:

[0035] h=h top -h bottom (7)

[0036] Assume that the data set of all injection hole center coordinates and the chamfer height to be processed is D={(x i ,y i ,h i}, i∈1, 2, …, N, where N is the number of injection holes, then for each data point, it needs to verify whether it meets the following conditions:

[0037]

[0038] where X ci , Y ci , H ci are the theoretical coordinates and the height to be processed of the injection hole, σ x , σ y , σ h are the standard deviations of the difference between the theoretical coordinates and the actual coordinates of the injection hole and the height distribution to be processed.

[0039] Fourth step: processing G code generation

[0040] Use the secondary development interface of the numerical control system to write a universal processing G code template, design the height coordinate value of the injection hole to be processed and the center coordinate of the injection hole as dynamic variables; in the G code template, the initial value of the parameter is the theoretical coordinate and the processing height designed for the workpiece; combine the data information of all the center coordinates of the injection hole and the height of the chamfer to be processed calculated in the third step, import the G code template through dynamic assignment and real-time adjustment, realize the dynamic and personalized of the parameters; the processing process code structure of the i-th injection hole is as follows:

[0041]

[0042] where G90 is the instruction to execute the operation in the machine tool coordinate system, G01 is the instruction of straight line motion, κ is the safety distance to improve, to prevent collision;

[0043] After data processing, store the parameters to be written into the template, such as power failure or system abnormality, you can recover the last stable state; through the combination of dynamic adjustment of the template and fault tolerance logic, realize the intelligent generation and optimization of processing G code;

[0044] Fifth step: intelligent processing based on force measuring tool holder

[0045] Use the force measuring tool holder to collect the cutting force signal in the processing process, because the dynamic change of the cutting force signal is fast, the sampling frequency must meet the Nyquist sampling theorem to capture the complete details; the sampling frequency should meet the following formula:

[0046]

[0047] where s is the spindle speed, z is the number of teeth of the tool, f v is the natural frequency of the processing system, f sThe cutting force frequency; due to the data acquisition contains noise or outliers, the signal needs to be filtered; due to the cutting process belongs to high frequency force signal, therefore, low pass filter is adopted, the formula is as follows:

[0048] f cutoff <0.5f s (11)

[0049] Wherein, f cutoff is the cut-off frequency;

[0050] Due to the time error of system feedback, a deep learning framework based on long short-term memory network and one-dimensional convolutional neural network is designed to realize the prediction of cutting force signal and the extraction of key features; the cutting force signal is collected as time series data: represented as:

[0051] F(t)={F1,F2,...,F T} (12)

[0052] The actual machining data of multiple injection holes in the injection disc is used as the training sample, and the long short-term memory neural network is used to capture the time dependence of the cutting force signal; the one-dimensional convolutional neural network is used to extract the machining features in the predicted cutting force signal, and the expressions of the long short-term memory neural network and the one-dimensional convolutional neural network model are respectively:

[0053] h t ,C t =LSTM(x t ,h t-1 ,C t-1 ) (13)

[0054]

[0055] Wherein, h t is the hidden state of the current time step, C t is the cell state of the current time step, x t is the input of the current time step, h t-1 is the hidden state of the previous time step, C t-1 is the cell state of the previous time step; z t,k represents the output of the kth convolution kernel at time step t, K represents the size of the convolution kernel, that is, the number of input sequence elements involved in the convolution operation, w k,j represents the weight of the jth position in the kth convolution kernel, x t+j-1 represents the value of the input sequence at time step t+j-1, b k represents the bias term of the kth convolution kernel;

[0056] When the height value obtained by the feature prediction result is lower than the calculation result of the geometric feature when the machining is completed, the numerical control system is used to skip the current line instruction, interrupt the current injection hole machining in advance and continue the subsequent operation, and the machining geometric parameters are backed up in the document, and the data and machining quality are rechecked after the overall machining task is completed to judge the machining completion degree.

[0057] The present application has the advantages of improving the machining precision, reducing the labor intensity of workers, and providing an advanced solution for efficient and intelligent machining of the injection hole of a liquid rocket engine. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 It is a schematic diagram of the overall machining system.

[0059] Figure 2 It is a schematic diagram of the distribution of injection holes on the injection disc.

[0060] Figure 3 It is a schematic diagram of the structure of the injection hole.

[0061] Figure 4 It is an edge fitting effect diagram of the injection hole scanning point cloud.

[0062] Figure 5 It is a schematic diagram of the height solving of the injection hole.

[0063] Figure 6 It is a flow chart of the intelligent machining method of the rocket engine injection hole.

[0064] In the figure, 1 is a machine tool main shaft connecting flange, 2 is an injection disc, 3 is a self-centering fixture, 4 is a machine tool workbench, 5 is a line laser sensor tooling, and 6 is a line laser sensor. DETAILED DESCRIPTION

[0065] The specific embodiments of the present application will be further described in combination with the drawings and technical solutions.

[0066] The machining object of the embodiment of the present application is a certain type of rocket engine injection disc with a diameter of 430mm and a height of 300mm, the line laser sensor used is a Keyence LJX-8200 sensor, the machine tool is a five-axis numerical control machining center, the numerical control system is a SINUMERIK ONE system, and the force measuring tool holder is a Spike mobile tool holder type force sensor. Figure 6 The overall flow chart of the large-scale point cloud simplification method based on the improved feature descriptor is as follows:

[0067] First step: using a self-centering fixture to position and clamp the workpiece

[0068] The self-adaptive centering fixture is composed of two semicircular parts of fixed end and movable end. In order to ensure the coaxiality of the workpiece and the center of rotation of the machine tool worktable, the fixed end arc surface is first installed and adjusted. During the workpiece clamping process, first, the outer wall surface of the injection disc is tightly attached to the fixed end, and then the movable end is closed and tightened through the screw.

[0069] Second step: line laser scanning measurement

[0070] Considering the high sensitivity of the line laser sensor to the posture, after the sensor is clamped and fixed, the high-precision profile surface is adjusted. After completing the pose calibration of the sensor, execute the "SPOS" instruction to lock the machine tool spindle to avoid the change of the sensor pose in the subsequent operation and introduce systematic errors. Design an S-shaped scanning trajectory to cover the area to be processed and scan the workpiece.

[0071] Third step: solving the machining related quantities

[0072] Select the smallest bounding box Ω that can completely cover the geometric features of a single injection hole, extract the point cloud data in the bounding box to establish a sub-point cloud N. According to the different scanning lines of the line laser sensor, the coordinate points in the sub-point cloud N are grouped and numbered as P1, P2…P i …P n , P i+1 , P i+2 , P i+3 , and P i form a data matrix D i . Use a convolution kernel K of size 3x3 to perform convolution operation to obtain the data matrix D i ' after convolution.

[0073] D i ' = D i *K (15)

[0074] A number of data points on the upper surface of the steel injection pipe of the injection hole are manually labeled as labels H, and a multilayer perception machine is used for end-to-end training. The input data is the data matrix D' obtained by processing different injection holes D'1, D'2,...D' n-2 , and the output is the data point set j on the upper surface of the steel injection pipe of the injection hole. The data quantity in the point set j is consistent with the label H. The mean square error function value (MSE) is used as the loss function of the neural network to calculate and train the parameters.

[0075]

[0076] Where i is the data quantity of the point set and the label H. The neural network obtained by training is used to obtain the data point set P on the upper surface of the single injection hole steel injection pipe P j(x j ,y j ,z j ), j∈[1, n]}. The average value of all point z values is 0.2232 mm as the upper surface height h top of the injection hole.

[0077] According to the upper surface data point set P, the outermost edge point set S side of the steel pipe is found, which is expanded outward to obtain the injection hole copper base conical surface point set S cone .

[0078] S1: Injection hole center coordinate solving

[0079] Three points s1(x1, y1), s2(x2, y2), and s3(x3, y3) are randomly selected in S side to form an initial sample, and a circle fitting algorithm is used for circle fitting. The circle fitting equation is:

[0080]

[0081] where (x c , y c ) is the center coordinate of the fitted circle. The error of other points in S side from the fitted circle is calculated. If the error is within the preset radius threshold 0.1 of the fitted circle center, the point is considered an inner point. The judgment basis is as follows:

[0082]

[0083] The above steps are repeated multiple times, and the parameters of the fitted circle with the most inner points are selected as the final result. The center coordinate o(x0, y0) is the injection hole center coordinate. The calculated center coordinate of the first injection hole is (-0.494039, 0.2888). The process is repeated to obtain all center coordinates.

[0084] S2: Injection hole chamfer height solving

[0085] The least squares method is used to fit the straight line z1 = 0.023y + 0.03506 from the injection hole steel pipe upper surface point cloud, and the least squares method is used to fit the straight line z2 = -0.71y + 5.82 from the steel pipe side surface point cloud. The intersection point of the straight lines z1 and z2 is solved to obtain the straight line y = 8.18. The least squares method is used to fit the straight line z4 = -0.71y + 5.82 from the injection hole copper base conical surface points, and the intersection point 13 of the straight lines 12 and 11 is the lowest point, i.e., the lower surface height 0.0122 mm.

[0086] A point set L is formed by taking a point on the generatrix of the first injection hole copper base cone, and a straight line equation z is obtained by using the least square method for linear fitting. The slope straight line equation is z2=0.64y-5.22, and the intersection point of the straight line z and the outer wall straight line z1 of the steel pipe, that is, z1=-0.71y+5.82, is the lowest point, and the lower surface height value h can be obtained bottom . Further, the nozzle height of the first injection hole is 0.211 mm. Repeat the process to obtain the nozzle height of all injection holes.

[0087] In order to eliminate the error and noise in the point cloud data analysis process, improve the stability and reliability of the machining process, trigger alarm and stop when the parameters exceed the safe machining range, and reduce the risk of accidents caused by abnormal data. Assuming that the data set of all injection hole center coordinates and the height of the chamfer to be machined is D={(x i ,y i ,h i}, i∈1, 2,..., N, then for each data point, it needs to verify whether it meets the following conditions:

[0088]

[0089] Where X ci , Y ci , H ci are the theoretical coordinates and the height to be machined of the injection hole, and σ x , σ y , σ h are the standard deviations of the difference between the theoretical coordinates and the actual coordinates of the injection hole and the height distribution. After threshold judgment, the machining height of the first injection hole in the machine tool coordinate system is-311.025 mm.

[0090] Fourth step: processing G code generation

[0091] A general machining G code template is written using the secondary development interface of the numerical control system, and the height coordinate value and the injection hole center coordinate are designed as dynamic variables. In the G code template, the initial value of the parameter is the theoretical coordinate and the machining height of the workpiece design. Combined with the workpiece geometric information calculated in the third step, the actual calculated height and center coordinates and other machining parameters are imported into the G code template through dynamic assignment and real-time adjustment, realizing the dynamic and personalized parameters. The machining process code structure of the first injection hole is as follows:

[0092] G90 G01 X-0.494039Y0.2888Z-301.025F500

[0093] G90 G01 X-0.494039Y0.2888Z-311.025F10

[0094] G90 G01X-0.494039Y0.2888Z-301.025F500

[0095] Fifth step: intelligent machining based on force measuring tool holder

[0096] The cutting force signal in the machining process is collected by using the force measuring tool holder. Since the dynamic change of the cutting force signal is fast, the sampling frequency meets the Nyquist sampling theorem. In this embodiment, the spindle speed s = 100, the number of teeth of the tool z = 4, and the natural frequency of the machining system f v = 1 kHz. To meet:

[0097]

[0098] The cutting force sampling frequency f s = 2.5 kHz. Since the data acquisition contains noise or outliers, low-pass filtering is used to process the signal, and the cutoff frequency f cutoff is set to f v = 1.25 kHz

[0099] Due to the time error of system feedback, a deep learning framework based on long short-term memory network and one-dimensional convolutional neural network is designed to realize the prediction of cutting force signal and the extraction of key features. The cutting force signal is collected as time series data: F(t) = {F1, F2,..., F

[0100] F(t) = {F1, F2,..., F T} (21)

[0101] The actual machining data of multiple injection holes in the injection disc are used as training samples to train the long short-term memory network to capture the time dependence of the cutting force signal. The model expression is:

[0102] h t ,C t = LSTM(x t ,h t-1 ,C t-1 ) (22)

[0103] Where h t is the hidden state of the current time step, C t is the cell state of the current time step, x t is the input of the current time step, h t-1 is the hidden state of the previous time step, and C t-1 is the cell state of the previous time step.

[0104] One-dimensional convolutional network is used to extract the machining features in the cutting force signal to prevent overcutting. The specific formula is:

[0105]

[0106] When the height value of the machining completion feature prediction result is lower than the geometric feature calculation result, the numerical control system skips the current line instruction, interrupts the current injection hole machining in advance, and continues the subsequent operation. The machining geometry parameters are backed up in a special document, and the data and machining quality are rechecked after the overall machining task is completed to judge the machining completion degree. Other prediction results represent the safety of the machining process, and the machining can be performed normally according to the code.

[0107] The above specific implementation cases further illustrate the purposes, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific implementation case of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A smart machining method for injection holes in a liquid rocket engine, characterized in that, The steps are as follows: First, the workpiece, namely the liquid rocket engine injection disk, is fixed on the worktable of a five-axis CNC machine tool using an adaptive centering fixture to determine the relative positional relationship between the workpiece coordinate system and the machine tool coordinate system. The adaptive centering fixture is used for workpiece positioning and clamping, and it mainly consists of two semi-circular structures: a fixed end and a movable end. During workpiece clamping, the outer wall of the liquid rocket engine injection disk is first pressed against the fixed end, and then the movable end is closed and tightened with screws. Secondly, a line laser sensor is used to scan the injection plate from multiple angles to ensure the integrity of the 3D point cloud data of the injection holes, without being limited by the structure of the injection plate or any obstructions. The line laser scanning measurement is as follows: the line laser sensor establishes communication with the machine tool CNC system through a network cable and is integrated into the end of the machine tool spindle using a tool holder-type fixture; after the pose calibration of the line laser sensor is completed, the "SPOS" command is executed to lock the machine tool spindle; to ensure the integrity of the 3D point cloud data of each injection hole, a multi-directional line laser scanning measurement trajectory is designed. Subsequently, point cloud reconstruction technology was used to accurately analyze the spatial geometric features of the area to be processed, obtaining the actual center coordinates of the injection hole and the chamfer height information to be processed; the specific solution of the processing-related quantities is as follows: the smallest bounding box that can completely cover the geometric features of a single injection hole is selected. Extract point cloud data within the bounding box to construct sub-point cloud N; group the coordinate points within sub-point cloud N according to their corresponding laser scan lines, numbering them P1, P2…P… i …P n , where n is the total number of laser scanning lines contained within the acquired workpiece point cloud, from P i Starting with every 3 consecutive laser scan lines, data P i+1 P i+2 P i+3 Form a data matrix D i The convolution operation is performed using a 3×3 kernel K to obtain the convolutioned data matrix. ; ； Several data points on the upper surface of the injection nozzle steel nozzle were manually marked as labels. End-to-end training was performed using a multilayer perceptron; the input data consisted of data matrices obtained from processing different injection orifices. The output data is a set of data points on the upper surface of the injection hole steel nozzle. Data point set Data volume and tags Maintain consistency; use the mean squared error function (MSE) as the standard for calculating the loss function and training parameters of the neural network; ； Where i is the data point set With tags The amount of data; using the trained neural network to obtain the data point set of the upper surface of a single injection hole steel nozzle. Calculate the average height of all points as the height h of the upper surface of the injection hole. top ; ； Based on the data point set P on the upper surface of the injection hole steel nozzle, find the outermost edge point set S of the injection hole steel nozzle. side Extending it outwards, we obtain the cone-shaped point set S of the injection hole copper substrate. cone ; S1: Solving for the coordinates of the injection hole center Points at the outermost edge of the injection hole steel nozzle Three points were randomly selected from the middle. , , The initial samples are used for circle fitting using the least squares fitting algorithm. The circle fitting equation is: ； in, Let R be the coordinates of the center of the circle to be fitted, and R be the radius of the circle to be fitted. Calculate the outermost edge point set of the injection nozzle steel nozzle The error between other points and the fitted circle, for any point If the error is within the preset radius threshold of the fitted circle center If a point is inside a point, then that point is considered an interior point; the criteria for this determination are as follows: ； ； Repeat the above steps multiple times, select the parameters of the fitted circle with the most interior points as the final result, and obtain the coordinates of the circle center. That is, the coordinates of the center of the injection hole; S2: Solving for the chamfer height of the injection hole The first straight line is obtained by fitting the point cloud on the upper surface of the steel nozzle using the least squares method. The second straight line is obtained by fitting the point cloud on the side of the steel nozzle using the least squares method. The third straight line is obtained by solving the problem perpendicular to the horizontal plane through the intersection of the first and second straight lines. The fourth straight line is obtained by fitting the points on the conical surface of the copper substrate of the nozzle using the least squares method. The intersection of the fourth and third straight lines is the lowest point, and the height of the lowest point is the height h of the lower surface. bottom The nozzle height of the injection orifice is expressed as: ； Assume the dataset of all injection hole center coordinates and the chamfer height to be processed is as follows: Where N is the number of injection holes, then for each data point, it is necessary to verify whether it meets the following conditions: ； in, , , These are the theoretical coordinates and the height to be processed corresponding to the injection hole. , , These represent the difference between the theoretical and actual coordinates of the injection hole and the standard deviation of the height distribution to be processed, respectively. Furthermore, based on the calculated machining feature parameters, the machining G-code is automatically generated using the secondary development interface of the CNC system; the specific details of the machining G-code generation are as follows: A generalized machining G-code template was written using the secondary development interface of the CNC system, designing the height coordinates and center coordinates of the injection holes to be machined as dynamic variables. In the G-code template, the initial parameter values ​​are the theoretical coordinates and machining height of the workpiece. Combining the data information of all injection hole center coordinates and the chamfer height obtained in the third step, the parameters are dynamically assigned and adjusted in real time and imported into the G-code template to achieve parameter dynamism and personalization. The machining process is then completed. The code structure for the machining process of each injection hole is as follows: ； Among them, G90 is the command to perform operations in the machine tool coordinate system, and G01 is the command for linear motion. To increase the safety distance and prevent collisions; After data processing, the parameters to be written to the template are stored. In the event of a power outage or system malfunction, the system can recover to the previous stable state. Through the combination of dynamic template adjustment and fault-tolerant logic, intelligent generation and optimization of the processing G-code are achieved. Finally, a force-measuring tool holder is used to collect cutting force signals in real time during the machining process. Combined with a deep learning model, the cutting force data is dynamically analyzed, and machining parameters are adaptively adjusted to optimize the machining state, achieving intelligent monitoring and precise control throughout the entire process. The specific details of intelligent machining based on the force-measuring tool holder are as follows: A force-measuring tool holder is used to collect cutting force signals during machining. Because the cutting force signal changes rapidly, the sampling frequency must satisfy the Nyquist sampling theorem to capture complete details. The sampling frequency should conform to the following formula: ； in, Main spindle speed The number of teeth on the cutting tool. The inherent frequency of the processing system, The cutting force frequency is used. Since the data acquisition includes noise or outliers, the signal needs to be filtered. Because the cutting process involves high-frequency force signals, a low-pass filter is used, as shown in the following formula: ； in, The cutoff frequency; Due to time errors in the system feedback, a deep learning framework based on Long Short-Term Memory (LSTM) networks and one-dimensional convolutional neural networks was designed to predict cutting force signals and extract key features. The cutting force signals were collected as time-series data, represented as follows: ； Using actual machining data from multiple injection holes in the injection plate as training samples, a long short-term memory neural network is employed to capture the time dependence of the cutting force signal. A one-dimensional convolutional neural network is used to extract machining features from the predicted cutting force signal. The model expressions for the long short-term memory neural network and the one-dimensional convolutional neural network are as follows: ； ； in, The hidden state at the current time step. This represents the cell state at the current time step. For the current time step input, This is the hidden state from the previous time step. This represents the cell state at the previous time step; This represents the output of the k-th convolutional kernel at time step t, where K represents the size of the convolutional kernel, i.e., the number of input sequence elements involved in the convolution operation. This represents the weight at position j in the k-th convolutional kernel. This represents the value of the input sequence at time step t+j−1. This represents the bias term of the k-th convolutional kernel; When the height value obtained from the feature prediction result after machining is lower than the geometric feature calculation result, the CNC system skips the current line instruction, interrupts the current injection hole machining in advance, and continues the subsequent operation. The geometric parameters to be machined are backed up in the document. After the overall machining task is completed, the data and machining quality are reviewed to determine the degree of machining completion.

2. The intelligent machining method for the injection hole of a liquid rocket engine according to claim 1, characterized in that, To ensure the coaxiality of the workpiece and the rotation center of the machine tool table, the fixed end arc surface of the self-adaptive centering fixture needs to be aligned using a dial indicator during the initial installation.

Images