Random forest algorithm-based antenna panel deformation prediction method and system, medium and equipment
Through the antenna panel deformation prediction method based on the random forest algorithm, the problem of difficult measurement and real-time correction of deformation caused by the radio telescope due to temperature and wind is solved, and fast and accurate surface shape detection and automated correction are achieved, which is suitable for telescopes of various sizes. .
Patent Information
- Application Number
- CN202411060658.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-26
AI Technical Summary
The prior art is difficult to quickly and accurately measure and correct the deformation of the antenna panel caused by the radio telescope due to temperature and wind, resulting in a decrease in imaging quality. The traditional method has low automation and requires a lot of manpower to operate.
The antenna panel deformation prediction method based on the random forest algorithm is adopted. By obtaining the reference plane shape and applying displacement actuators to generate deformation data, the random forest algorithm is trained to quickly predict the deformation plane shape distribution, and the detection speed and automation level are improved.
The surface shape detection speed and automation are significantly improved while ensuring accuracy, and can quickly obtain surface shape deformation. It is suitable for large or small and medium-sized telescopes, and the training set and parameters are flexibly adjusted to suit different situations.
Smart Images

Figure CN119026462A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of deformation prediction analysis and real-time correction of antenna main panels, and specifically relates to an antenna panel deformation prediction method, system, medium and equipment based on a random forest algorithm. Background Art
[0002] Gravity, temperature and wind are the main factors that cause structural deformation of radio telescopes. Many radio telescopes are optimized for this at the beginning of their structural design. However, in actual operation, the residual structural deformation of high-precision telescopes may still cause errors such as focusing, pointing and path length to exceed performance specifications, affecting the final image quality and causing inconvenience to astronomical observations.
[0003] Generally, the structural design of large radio telescopes is based on finite element analysis. Combined with the elastic deformation theory of primary mirror materials such as steel, aluminum, and modern carbon fiber materials, the deformation of telescope structures under the influence of gravity, temperature, and wind can be studied and analyzed. The influence of gravity on the deformation of the primary mirror can be analyzed by finite element analysis and other methods to obtain high-precision results and make corrections. Temperature and wind are unstable and variable transient phenomena, and their impact on the structure is relatively difficult to analyze and predict in advance.
[0004] Traditional surface shape detection methods include photogrammetry, holography, laser tracker, etc. The most common is photogrammetry, because it is easier to operate and can achieve higher accuracy, but it takes too long. Obviously, it is difficult for this type of measurement method to accurately measure the transient antenna reflector surface deformation caused by temperature or wind. In particular, when the telescope aperture becomes larger, it takes longer, and the accuracy of the telescope reflector surface needs to be maintained at a high level for a long time, and traditional methods are difficult to achieve real-time measurement.
[0005] This type of traditional method in surface measurement has a low degree of automation and requires manpower to solve separately, which is an independent problem. In order to improve the fluency of the surface measurement and recovery part of the telescope operation, some methods point out that surface recovery is based on geometric optics. By measuring the near-field amplitude, the deformation amplitude equation is derived, and the partial differential equation is solved to obtain accurate deformation recovery. Due to the complexity of the mathematical model, there is also a method that proposes to combine artificial neural networks to solve the equation, improve the efficiency of the algorithm, and finally obtain high-precision surface recovery.
[0006] Random forest is a supervised machine learning algorithm, which is an overall model composed of many decision trees. The prediction is made by averaging the predictions of each decision tree. Each tree in the random forest is training a subset of data. For regression problems, the final output of the random forest is the average of the outputs of all decision trees. The random forest first performs random sampling with replacement on all samples, that is, the bootstrapping sampling method. The selected small sample set uses a CART decision tree, and then randomly selects m features, m < M, for regression analysis; after multiple samplings, n decision trees are formed, and the output of each tree is combined with the arithmetic average strategy as the final output. Due to the random sampling process with replacement, the random forest effectively improves the robustness and accuracy of the algorithm. The random selection of features can ensure the algorithm's rapid processing of high-dimensional data. Summary of the invention
[0007] The present invention aims to improve the detection speed of the telescope primary mirror surface shape, improve the degree of surface shape detection automation, and provide a basis for real-time detection and correction technology of surface shape.
[0008] To achieve the above object, the present invention adopts the following technical solutions:
[0009] The antenna panel deformation prediction method based on the random forest algorithm is characterized by comprising the following steps:
[0010] Obtaining the reference surface shape of the antenna panel;
[0011] The displacement actuator installed on the antenna panel is used to apply multiple sets of displacements to make the antenna panel deform. Taking the reference surface shape as the standard, each set of deformation coordinate data is obtained by photogrammetry. The deformation coordinate data of these known displacement deformation variables are used as samples to provide a data set for the random forest algorithm.
[0012] The random forest algorithm is trained using a data set, with its input being the point coordinates and displacement deformation of the displacement detection point, and its output being the coordinates of the displacement detection point after deformation;
[0013] The point coordinates and displacement deformation of the displacement detection points are input into the trained random forest algorithm, and the surface shape distribution of the antenna panel after deformation is obtained according to the output coordinates of the displacement detection points after deformation.
[0014] To optimize the above technical solutions, the specific measures taken also include:
[0015] Furthermore, the obtaining of the reference surface shape of the antenna panel is specifically as follows:
[0016] After correcting the antenna panel deformation caused only by gravity at the set pitch angle, the photogrammetry method is used to obtain the coordinate information of some points on the antenna panel, and then interpolation is performed to obtain the overall surface shape distribution as the reference surface shape.
[0017] Furthermore, six displacement actuators and corresponding sensors are installed on the antenna panel.
[0018] Furthermore, the random forest algorithm adopts uniform sampling with replacement, the node splitting strategy of the decision tree adopts the Gini index, and the number of decision trees is 800.
[0019] Furthermore, according to the outputted coordinates of the displacement detection points after deformation, the surface shape distribution of the antenna panel after deformation is obtained, specifically:
[0020] The deformed coordinates of the output displacement detection points are interpolated to obtain the deformed surface distribution of the antenna panel.
[0021] Accordingly, the present invention proposes an antenna panel deformation prediction system based on a random forest algorithm, which is characterized by comprising:
[0022] A reference surface shape acquisition module, used to acquire the reference surface shape of the antenna panel;
[0023] The data set construction module is used to use the displacement actuator installed on the antenna panel to apply multiple groups of displacements to deform the antenna panel, and use photogrammetry to obtain each group of deformation coordinate data based on the reference surface shape. The deformation coordinate data of these known displacement deformation variables are used as samples to provide a data set for the random forest algorithm.
[0024] A training module is used to train a random forest algorithm using a data set, so that its input is the point coordinates and displacement deformation of the displacement detection point, and its output is the coordinates of the displacement detection point after deformation;
[0025] The prediction module is used to input the point coordinates and displacement deformation of the displacement detection point into the trained random forest algorithm, and obtain the surface distribution of the antenna panel after deformation according to the output coordinates of the displacement detection point after deformation.
[0026] Accordingly, the present invention proposes a computer-readable storage medium storing a computer program, characterized in that the computer program enables a computer to execute the antenna panel deformation prediction method based on the random forest algorithm as described above.
[0027] Accordingly, the present invention proposes an electronic device, characterized in that it includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the antenna panel deformation prediction method based on the random forest algorithm as described above is implemented.
[0028] The beneficial effects of the present invention are as follows: the present invention effectively improves the speed and automation of surface shape detection while ensuring accuracy. After the data set is prepared in advance, the surface shape deformation can be obtained by obtaining the displacement of several measurement points each time the deformation occurs. The method of the present invention is relatively flexible, and only the measurement point positions need to be determined and the quality of the training set needs to be ensured. If it is necessary to determine the accuracy requirements according to the actual surface shape conditions, the training set can be adjusted or the input and output can be retrained. The method of the present invention has a wide range of applicability and is applicable to large or medium and small telescopes. It only needs to select relevant random forest parameters according to the situation. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 It is a flow chart of the antenna panel deformation prediction method based on the random forest algorithm.
[0030] Figure 2 This is the general flow chart of the random forest algorithm.
[0031] Figure 3a A simulation result diagram showing the reference surface shape using true values.
[0032] Figure 3b A result diagram showing the reference surface shape obtained by the method of the present invention.
[0033] Figure 4a Photogrammetric results showing the shape distribution after deformation using the true value.
[0034] Figure 4b A result diagram showing the shape distribution after deformation obtained by the method of the present invention. DETAILED DESCRIPTION
[0035] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.
[0036] In one embodiment, the present invention proposes a method for predicting antenna panel deformation based on a random forest algorithm, such as Figure 1 The specific design steps are as follows.
[0037] (1) To obtain the distribution of panel deformation, it is actually necessary to obtain the coordinates of each point on the panel after deformation. First, a reference surface shape is required, which can be approximately regarded as an ideal surface shape. Subsequent deformations need to be compared with the reference surface shape and corrected and adjusted accordingly. In practical problems, the reference surface can be selected according to actual needs. For example, the surface shape after correction of gravity deformation at a certain pitch angle can be obtained by using photogrammetry methods to obtain the coordinate information of some points on the panel, and then interpolation is performed to obtain the overall surface shape distribution.
[0038] (2) Since random forest is a machine learning algorithm, multiple sets of displacements are applied by displacement actuators installed on the panel to cause the panel to deform, and photogrammetry is used to obtain each set of deformation data. The panel deformation coordinate data with these known deformation variables are used as samples to provide a data set for the random forest method.
[0039] (3) For panel deformation, the most obvious feature is the change in point coordinates. Therefore, the input of the problem is abstracted into the point coordinates and the displacement deformation of the displacement detection point, and the output is the coordinate of the point after deformation in the z direction. Therefore, the input dimension of the problem is 3+n, that is, the coordinate information of these points and the displacement deformation measured by n measurement points. In the actual experimental environment, the panel has 6 actuators and sensors, so the dimension of the random forest input data is 3+6; the output data dimension is 1, which means the deformation in the z direction at the corresponding coordinate point. The algorithm obtains the deformed coordinates of these points, and uses the same interpolation method to obtain the overall panel deformation distribution. Figure 2 The workflow of the random forest algorithm is shown.
[0040] It should be noted that some parameters and methods in the random forest algorithm are selected. The sampling method adopted in this embodiment is the random selection emphasized by the random forest, that is, uniform sampling with replacement; the node splitting strategy of the decision tree in the forest adopts the Gini index; the number of decision trees needs to be optimized according to the size of the data set, and 800 is used here.
[0041] (4) The random forest will output a set of deformation variables in the z direction, which correspond one-to-one with the input x and y coordinates to form a set of predicted coordinates after the panel is deformed. That is, the coordinate set of the panel after deformation is obtained, and the surface shape distribution can be obtained by interpolation.
[0042] Next, the method proposed in this embodiment is described with reference to specific examples.
[0043] Taking the simulation data of a single panel of a 60m telescope obtained by ANSYS simulation as an example, at a zenith angle of 0 degrees, the panel affected only by gravity is deformed into the initial reference surface. The actuator applies 50 sets of displacements to four points on the panel to obtain 50 sets of different surface shapes; 49 sets are used as training sets and 1 set as a test set, and the coordinate points after deformation are obtained by regression prediction using random forest, and finally the surface shape distribution is obtained by interpolation.
[0044] The true value simulation results of the reference surface shape and the results obtained by using the present invention are as follows: Figure 3a and 3b The results show that the surface error RMS of the true value is 5.55mm, and the surface error obtained by the method proposed in the present invention is 4.16mm, which shows that the method proposed in the present invention has good accuracy and is less different from the true value.
[0045] Taking a single panel of a 13.7m telescope with a 15° aperture angle as an example, under field experimental conditions, the panel deformation affected only by gravity becomes the initial reference surface, which is obtained by photogrammetry. The actuator applies 16 sets of displacements to six points on the panel to obtain 16 different surface shapes; 15 sets are used as training sets and 1 set as test sets, and the coordinate points after deformation are obtained by regression prediction using random forest, and finally the surface shape distribution is obtained by interpolation.
[0046] The true value simulation results of the shape distribution after deformation and the results obtained by using the present invention are as follows: Figure 4a and 4b The results show that the surface error RMS of the true value is 389.05um, and the surface error obtained by the proposed method is 401.09um. Since there are more factors affecting the surface shape in the real environment, the result is slightly worse than that on the simulation data set, but it still has an acceptable error and good accuracy, and the gap with the true value is small.
[0047] In another embodiment, the present invention proposes an antenna panel deformation prediction system based on a random forest algorithm, which corresponds to the antenna panel deformation prediction method based on a random forest algorithm proposed in the first embodiment, and specifically includes:
[0048] A reference surface shape acquisition module, used to acquire the reference surface shape of the antenna panel;
[0049] The data set construction module is used to use the displacement actuator installed on the antenna panel to apply multiple groups of displacements to deform the antenna panel, and use photogrammetry to obtain each group of deformation coordinate data based on the reference surface shape. The deformation coordinate data of these known displacement deformation variables are used as samples to provide a data set for the random forest algorithm.
[0050] A training module is used to train a random forest algorithm using a data set, so that its input is the point coordinates and displacement deformation of the displacement detection point, and its output is the coordinates of the displacement detection point after deformation;
[0051] The prediction module is used to input the point coordinates and displacement deformation of the displacement detection point into the trained random forest algorithm, and obtain the surface distribution of the antenna panel after deformation according to the output coordinates of the displacement detection point after deformation.
[0052] In this embodiment, the working principle and execution process of each module of the system are the same as the steps of the method proposed in Embodiment 1, so they will not be repeated here.
[0053] In another embodiment, the present invention proposes a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the antenna panel deformation prediction method based on the random forest algorithm as described in the first embodiment.
[0054] In another embodiment, the present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the antenna panel deformation prediction method based on the random forest algorithm as described in Example 1 is implemented.
[0055] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.
[0056] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0057] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.
Claims
1. The antenna panel deformation prediction method based on random forest algorithm is characterized by: The steps include: Obtaining the reference surface shape of the antenna panel; The displacement actuator installed on the antenna panel is used to apply multiple sets of displacements to make the antenna panel deform. Taking the reference surface shape as the standard, each set of deformation coordinate data is obtained by photogrammetry. The deformation coordinate data of these known displacement deformation variables are used as samples to provide a data set for the random forest algorithm. The random forest algorithm is trained using a data set, with its input being the point coordinates and displacement deformation of the displacement detection point, and its output being the coordinates of the displacement detection point after deformation; The point coordinates and displacement deformation of the displacement detection points are input into the trained random forest algorithm, and the surface shape distribution of the antenna panel after deformation is obtained according to the output coordinates of the displacement detection points after deformation.
2. The antenna panel deformation prediction method based on random forest algorithm according to claim 1, characterized in that: The obtaining of the reference surface shape of the antenna panel is specifically as follows: After correcting the antenna panel deformation caused only by gravity at the set pitch angle, the photogrammetry method is used to obtain the coordinate information of some points on the antenna panel, and then interpolation is performed to obtain the overall surface shape distribution as the reference surface shape.
3. The antenna panel deformation prediction method based on random forest algorithm according to claim 1, characterized in that: Six displacement actuators and corresponding sensors are installed on the antenna panel.
4. The antenna panel deformation prediction method based on random forest algorithm according to claim 1, characterized in that: The random forest algorithm adopts uniform sampling with replacement, the node splitting strategy of the decision tree adopts the Gini index, and the number of decision trees is 800.
5. The antenna panel deformation prediction method based on random forest algorithm according to claim 1, characterized in that: According to the output coordinates of the displacement detection points after deformation, the surface distribution of the antenna panel after deformation is obtained, which is specifically: The deformed coordinates of the output displacement detection points are interpolated to obtain the deformed surface distribution of the antenna panel.
6. The antenna panel deformation prediction system based on random forest algorithm is characterized by: include: A reference surface shape acquisition module, used to acquire the reference surface shape of the antenna panel; The data set construction module is used to use the displacement actuator installed on the antenna panel to apply multiple groups of displacements to deform the antenna panel, and use photogrammetry to obtain each group of deformation coordinate data based on the reference surface shape. The deformation coordinate data of these known displacement deformation variables are used as samples to provide a data set for the random forest algorithm. A training module is used to train a random forest algorithm using a data set, so that its input is the point coordinates and displacement deformation of the displacement detection point, and its output is the coordinates of the displacement detection point after deformation; The prediction module is used to input the point coordinates and displacement deformation of the displacement detection point into the trained random forest algorithm, and obtain the surface distribution of the antenna panel after deformation according to the output coordinates of the displacement detection point after deformation.
7. A computer-readable storage medium storing a computer program, characterized in that: The computer program enables the computer to execute the antenna panel deformation prediction method based on the random forest algorithm as described in any one of claims 1 to 5.
8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the antenna panel deformation prediction method based on the random forest algorithm as described in any one of claims 1 to 5 is implemented.
