Method and system for predicting acceleration of wedge-shaped body entering water

Abstract

The embodiment of the present invention discloses a method and system for predicting acceleration of a wedge-shaped body entering water impact, which relates to the field of water load design of surface aircraft, and can improve the problems of long test cycle and high test cost of a wedge-shaped body entering water impact in the current scheme. The present invention includes: a server receives initial data sent by a client, and establishes a wedge-shaped body entering water impact response model. The original test data of the wedge-shaped body entering water impact is collected, sample data is obtained using the original test data, and training sample data and test sample data are extracted from the sample data. The initial data is input into the wedge-shaped body entering water impact response model, and the wedge-shaped body entering water impact response model is trained using the training sample data. The test sample data is input into the trained wedge-shaped body entering water impact response model and an output result is obtained, and then the model accuracy is detected according to the output result.

Description

Technical Field

[0001] The present invention relates to the field of water load design of surface aircraft, and in particular to a method and system for predicting the impact acceleration of a wedge-shaped body entering water, which can be realized by artificial intelligence technology. Background Art

[0002] When a surface vehicle lands on water or a ship is hit by wind and waves, the structure will be subjected to a huge impact at the moment of entering the water. The huge impact load generated by the structure entering the water, including pressure and acceleration, is the most concerned issue in hull structural design and strength verification, and is also the main purpose of the two-dimensional wedge-shaped body water entry test. The wedge-shaped body water entry test uses different two-dimensional wedge-shaped body frames to drop from different heights for impact water entry measurement, so as to obtain the water entry load and structural response characteristics of the model, and study the influence of different parameters on the structural response.

[0003] The research methods of water slamming can be divided into three categories: theoretical methods, numerical methods, and experimental methods. In terms of water load theory, the approximate plate theory of Von Karman and Wagner assumes that the energy of the fluid system is conserved at the initial stage of water entry. According to the law of conservation of momentum, the inertial force of the additional mass is used to replace the impact of the fluid on the structure, and the impact load of the structure when entering the water is calculated. However, the theoretical method is limited to the study of rigid body entry into water, and many assumptions are made on the flow field. Compared with the actual value, the deviation is large and it cannot reflect the essential characteristics and physical mechanism of the slamming phenomenon. The numerical calculation method is not mature enough, and the human factor is relatively large. Its calculation results must also be verified by experiments. At present, the experimental method is still one of the effective and most reliable means to study the slamming problem. But at the same time, the experimental method also has its inevitable defects. One test requires huge manpower and material resources, the test cycle is long, and the cost is high; and the test results are completely dependent on the set test conditions. Once the test conditions change slightly, the test results will also change, and the test needs to be repeated. The test results are low in cost performance. However, the problem reflected in specific practical applications is that the research and development cycle of surface aircraft is long, and due to the large number of re-tests, the material cost of the test process is relatively high, the server or supercomputer time is also long, and the computing resources occupied are also relatively large.

[0004] Therefore, how to improve the problems of long cycle and high test cost of the wedge-shaped body water impact test in the current plan has become an urgent problem to be solved in the process of further improving the efficiency of my country's surface aircraft research and development. Summary of the invention

[0005] The embodiments of the present invention provide a method and system for predicting the acceleration of a wedge-shaped body entering water impact, which can improve the problems of long test cycle and high test cost of a wedge-shaped body entering water impact test in the current scheme, and further improve the efficiency of my country's research and development of surface aircraft.

[0006] To achieve the above object, the embodiments of the present invention adopt the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a method comprising:

[0008] S1. The server receives initial data sent by the client and establishes a wedge-shaped body water entry impact response model, wherein the initial data at least includes: the water entry state parameters of the wedge-shaped body set by the client, and the water entry state parameters of the wedge-shaped body at least include: the half width of the wedge-shaped body, the bottom ramp angle, the structural elastic modulus, the water entry height and the structural weight;

[0009] S2, collecting original test data of the wedge-shaped body entering water and impacting, using the original test data to obtain sample data, and extracting training sample data and test sample data from the sample data;

[0010] S3, inputting the initial data into the wedge-shaped body water entry impact response model, and using the training sample data to train the wedge-shaped body water entry impact response model;

[0011] S4. Input the test sample data into the trained wedge-shaped body water impact response model and obtain an output result, and then detect the model accuracy according to the output result. When the model accuracy does not reach the preset model accuracy, update the training sample data and repeat steps S3 to S4 until the model accuracy reaches the preset model accuracy.

[0012] In a second aspect, an embodiment of the present invention provides a system comprising: a server, a client, and an experimental device;

[0013] The client is used to record initial data according to the user's operation and send the initial data to the server;

[0014] The server is used to receive the initial data and establish a wedge-shaped body water entry impact response model, wherein the initial data at least includes: the wedge-shaped body water entry state parameters and the preset model accuracy set by the client, and the wedge-shaped body water entry state parameters at least include: the half width of the wedge-shaped body, the bottom ramp angle, the structural elastic modulus, the water entry height and the structural weight;

[0015] The experimental device is used to record the original test data of the wedge-shaped body entering the water, and the server collects the original test data, and uses the original test data to obtain sample data, and extracts training sample data and test sample data from the sample data;

[0016] The server is also used to input the initial data into the wedge-shaped body water entry impact response model, and use the training sample data to train the wedge-shaped body water entry impact response model; then, input the test sample data into the trained wedge-shaped body water entry impact response model and obtain an output result, then detect the model accuracy based on the output result and transmit it to the client.

[0017] The method and system for predicting the impact acceleration of a wedge-shaped body entering water provided by the embodiment of the present invention can mine the existing wedge-shaped body entering water impact test data, avoid unnecessary duplication of work, deeply abstract and summarize the internal laws of the data, and predict the impact acceleration characteristics of the wedge-shaped body entering water under any falling water state parameters at the fastest speed; the Gaussian process model used can obtain good model accuracy through model training, and play an engineering application role in the design of the impact strength of the structure entering water; and the Gaussian process is an artificial intelligence model, which can ignore the influence of random errors in the test on the test results to a certain extent, and has a certain anti-interference ability. Thereby, the problems of long cycle and high test cost of the wedge-shaped body entering water impact test in the current scheme can be improved, which has become a further improvement in the efficiency of my country's research and development of surface aircraft. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A schematic diagram of variable parameters of a wedge-shaped body falling into water test provided by an embodiment of the present invention;

[0020] Figure 2 A flow chart of the design of a wedge-shaped body water impact acceleration response model test provided by an embodiment of the present invention;

[0021] Figure 3 A diagram of the water entry process of a wedge-shaped body water entry test provided by an embodiment of the present invention;

[0022] Figure 4 A discrete schematic diagram of a wedge-shaped body water impact acceleration response curve provided by an embodiment of the present invention;

[0023] Figure 5A schematic diagram of original sample data of a wedge-shaped body water impact acceleration response curve provided by an embodiment of the present invention;

[0024] Figure 6 A schematic diagram of sample data of a newly constructed acceleration response curve of a wedge-shaped body entering water impact provided by an embodiment of the present invention;

[0025] Figure 7a and Figure 7b A comparison diagram of the kernel function and hyperparameter training errors of the wedge-shaped body water impact time model provided by an embodiment of the present invention;

[0026] Figure 8a and Figure 8b A comparison diagram of the kernel function and hyperparameter training errors of the wedge-shaped body water impact acceleration model provided by an embodiment of the present invention;

[0027] Fig. 9 A comparison chart of the absolute errors between the real value and the predicted value of the impact time and acceleration of the wedge-shaped body entering water provided by the embodiment of the present invention;

[0028] Fig.10 A comparison diagram of a predicted curve and a real curve of the acceleration response of a wedge-shaped body entering water impact provided by an embodiment of the present invention;

[0029] Fig.11a and Fig.11b A comparison diagram of relative error and root mean square error of the wedge-shaped body water impact acceleration response curve model provided by an embodiment of the present invention;

[0030] Fig.12 A schematic diagram of a method flow chart provided by an embodiment of the present invention;

[0031] Fig.13 A schematic diagram of a system architecture provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention will be described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention. It can be understood by those skilled in the art that, unless specifically stated, the singular forms "one", "one", "said" and "the" used herein may also include plural forms. It should be further understood that the term "including" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or their groups. It should be understood that when we say that an element is "connected" or "coupled" to another element, it can be directly connected or coupled to other elements, or there may also be intermediate elements. In addition, the "connection" or "coupling" used here may include wireless connection or coupling. The term "and / or" used herein includes any unit and all combinations of one or more associated listed items. It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as generally understood by those of ordinary skill in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless defined as herein.

[0033] The purpose of this embodiment is to design an artificial intelligence-based wedge-shaped body water impact acceleration prediction method, which is suitable for the rapid prediction and evaluation of water loads on surface aircraft. In view of the long cycle of the wedge-shaped body water impact test, high test cost, and the need to repeat the test when changing the water-falling conditions or the shape of the wedge-shaped body, based on the machine learning method, the existing wedge-shaped body water impact test data is reused and learned, and a wedge-shaped body water impact test model with good prediction effect and practical engineering significance is established.

[0034] The design concept of this embodiment can be roughly divided into four steps: Figure 2 As shown, it includes: 1. Design of the model test of the acceleration response curve of a wedge-shaped body entering water impact; 2. Simulation of the model test sample of the wedge-shaped body entering water impact response curve; 3. Training of the model of the acceleration response curve of a wedge-shaped body entering water impact; 4. Testing and accuracy evaluation of the model of the wedge-shaped body entering water impact response curve.

[0035] Specifically, an embodiment of the present invention provides a method for predicting the impact acceleration of a wedge-shaped body entering water, such as Fig.12 As shown, including:

[0036] S1. The server receives the initial data sent by the client and establishes a wedge-shaped body water entry impact response model.

[0037] The initial data at least includes: the falling state parameters of the wedge-shaped body set by the client, and the falling state parameters of the wedge-shaped body at least include: the half width, bottom slope angle, structural elastic modulus, falling height and structural weight of the wedge-shaped body.

[0038] S2. Collecting original test data of the wedge-shaped body entering water and impacting it, using the original test data to obtain sample data, and extracting training sample data and test sample data from the sample data.

[0039] S3, inputting the initial data into the wedge-shaped body water entry impact response model, and using the training sample data to train the wedge-shaped body water entry impact response model.

[0040] S4, inputting the test sample data into the trained wedge-shaped body water impact response model to obtain an output result, and then detecting the model accuracy according to the output result.

[0041] When the model accuracy does not reach the preset model accuracy, the training sample data is updated and steps S3 to S4 are repeatedly executed until the model accuracy reaches the preset model accuracy.

[0042] During the design process of this embodiment, an experimental design is carried out on the wedge-shaped body water impact test, a reasonable wedge-shaped body water impact mathematical model is established, and appropriate parameters are selected to represent the wedge-shaped body falling into water and the acceleration characteristics of the water impact. The wedge-shaped body falling into water parameters include: wedge-shaped body half-width b, bottom slope angle α, structural elastic modulus E, falling height H, structural weight M and other five parameters.

[0043] Specifically, in this embodiment, step S1 includes: the established wedge-shaped body water entry impact response model is expressed as:

[0044] The input parameters are consistent with the parameters of the wedge-shaped body falling into the water, including: b represents the half width of the wedge-shaped body, α represents the bottom slope angle, E represents the structural elastic modulus, H represents the falling height, and M represents the structural weight, such as Figure 1 As shown, it is recorded as X = [x1, x2, ..., x n ] T , x i =[b i , α i , E i , H i , M iThe output parameters include: water contact time T1, acceleration peak time T2, acceleration constant time T3, total time T4, acceleration value a at time j j , denoted as Y = [y1, y2, ..., y n ] T ,y i =[T 1i , T 2i , T 3i , T 4i , a 1i , a 2i , ..., a ji ], j represents the time value.

[0045] Furthermore, in step S2, the collected original test data of the wedge-shaped body entering water impact includes: the slamming acceleration response of the wedge-shaped body entering water impact process, wherein the slamming acceleration response is represented by a continuous curve recorded in real time, and the continuous curve is divided into T1, T2 and T3 segments. After collecting the original test data, the continuous curve of the slamming acceleration response is discretized, wherein the discrete points from segment T1 to segment T2 are encrypted, and the time averaging method is used to discretize the segments T2 to T3, wherein, since the physical laws from segment T1 to segment T2 are complex and more sample data are needed to describe the physical laws, the discrete points from segment T1 to segment T2 are encrypted, that is, the number of discrete points is increased, and the time averaging method is used to discretize the segments T2 to T3. . Specifically, the water entry process of the wedge-shaped body entering water impact test is as follows: Figure 3 As shown in the figure, the slamming acceleration response obtained from the wedge-shaped body water impact test is a continuous curve recorded in real time. Considering the computational cost of the computer, the fewer the output items, the higher the computational efficiency. When designing the experiment, the acceleration response curve is discretized in sections by using the rule that the acceleration of the wedge-shaped body is basically unchanged before it touches the water, and the acceleration increases first and then gradually decreases after entering the water, as shown in the figure. Figure 4 As shown. Assume that the time of water contact is T1, the time of acceleration peak is T2, the time of constant acceleration is T3, and the total time is T4. Since the slamming acceleration of the wedge before entering the water is equal to the acceleration of gravity, the magnitude of the acceleration remains basically unchanged. Therefore, two discrete points are taken in the T1 time period to simulate the magnitude of the slamming acceleration of the wedge before entering the water; in addition, after touching the water, the slamming acceleration of the wedge quickly reaches the peak value under the action of the water surface, and then the acceleration gradually decreases to a stable state. Therefore, the discrete points of the T1-T2 segment are encrypted to enhance the accuracy of its discreteness, and the time averaging method is used to discretize the slamming acceleration curve in the T2-T3 segment.

[0046] In this embodiment, in step S2, the process of obtaining sample data using the original test data includes:

[0047] The original test data is interpolated and constructed to obtain the sample data, wherein the sample data consisting of input and output pairs can be obtained. In the process of interpolating and constructing the original test data, random noise is added to simulate actual errors.

[0048] The interpolation construction process includes: x i represents the input of sample data obtained by interpolation, y i represents the output of the sample data obtained by interpolation, x0 represents the minimum value of the interpolation interval in the original input, y0 represents the minimum value of the interpolation interval in the original output, i represents the subscript of the sample data obtained by interpolation, x n Indicates the maximum value of the interpolation interval in the original input, y n represents the maximum value of the interpolation interval in the original output, and n represents the subscript of the maximum value of the interpolation interval in the original sample. The ways to add random noise include: Among them, ε i represents the random noise, Represents the output of the sample data after adding random noise, y i represents the output of the sample data to which random noise is to be added, i represents the subscript of the sample data, and N represents Gaussian distribution.

[0049] In this embodiment, the samples required for the wedge-shaped body water impact model test can be processed based on the test process and test results, and a sample space corresponding to the wedge-shaped body falling into water parameters and the water impact acceleration characteristic data can be constructed to provide initial samples for the Gaussian regression model and perform a preliminary analysis on the initial samples. The original test data of the wedge-shaped body water impact is relatively small. In order to appropriately increase the number of samples, the initial samples are interpolated to construct new samples. The interpolation method is as shown in formula (1), and appropriate random noise is added to simulate the actual error to obtain new sample data.

[0050]

[0051] Specifically, before using the Gaussian regression model to construct the wedge-shaped body water impact model, the obtained test samples are simulated first. Figure 5 is the original test data of the wedge-shaped body's impact response when entering water. Since the original data is small in number, in order to appropriately increase the number of samples, the initial samples are interpolated to construct new samples. The interpolation method is as shown in formula (1), and appropriate random noise is added to simulate the actual error. The new sample data is obtained as shown in the attached figure. Figure 6 shown.

[0052] In this embodiment, after obtaining the required sample data, a random sampling method can be used to sample samples in the sample space, and some samples can be extracted as training samples of the slamming response model of the wedge-shaped body water impact model, and the remaining samples are reserved as test samples of the model, so as to construct a sample space corresponding to the wedge-shaped body falling into water parameters and the water impact acceleration characteristic data.

[0053] Specifically, in step S2, the process of extracting training sample data and test sample data from the sample data includes: after obtaining the sample data, establishing a sample space, wherein the sample space includes: the correspondence between the falling state parameters of the wedge body and the acceleration characteristics of the water impact. Randomly sampling the samples in the sample space to obtain the training sample data and the test sample data.

[0054] The random sampling of samples in the sample space includes: after determining the sampling scale n0, in the sample space, the variable x of each dimension is i The domain interval [l i ,u i ] is divided into n0 equal parts, so that the sample space is n0×n0 hypercubes. A n0-order matrix U can be randomly generated, where U includes all permutations of random numbers {1, 2, ... n}. Then, a sample point is randomly selected from each row in the matrix U, and a total of n0 sample points are obtained. In each dimension, the coordinate point on the dimension is used as the variable x i , l i is the lower limit of the dimension, u i is the upper limit of the dimension, and i is the dimension subscript.

[0055] Specifically, after obtaining the required amount of sample data, some of the samples are extracted as training samples of the wedge-shaped body water impact acceleration response curve model, and the remaining samples are reserved as test samples of the model. Monte Carlo sampling method is used to randomly sample samples in the sample space.

[0056] Assume that the dimension of the experimental space is n, the sampling size is n0, and the coordinate point x on a certain dimension is i ∈[l i ,u i ], where l i This is the lower limit of the dimension, u i As the upper limit, the main steps are as follows: 1) Determine the sampling scale n0. 2) Set each dimensional variable x i The domain interval [l i ,u i] is divided into n0 equal parts, so the entire test space is divided into n0×n0 hypercubes. 3) Randomly generate an n0-order matrix U in this test space, where U is a full permutation of random numbers {1, 2,,…n}. Randomly select a sample point from each row of the matrix U to get n0 sample points. For example: Based on the existing partial wedge impact water test data, the sample space of the test design is given as shown in Table 1, and all training and testing are carried out in this sample space.

[0057] Table 1 Sample space

[0058] Design variables Value range b(m) <![CDATA[[b MIN ,b MAX ]]]> α(°) <![CDATA[[α MIN ,a MAX ]]> E(MPa) <![CDATA[[E MIN ,AND MAX ]]]> H(m) <![CDATA[[H MIN ,H MAX ]]]> M(kg) <![CDATA[[M MIN ,M MAX ]]]>

[0059] In the preferred solution of this embodiment, the Monte Carlo sampling method is used to select 80% of the samples as training samples to build the model, and 20% of the samples are used as test samples to evaluate the model accuracy. The water-falling state of 80% of the training samples is used as input parameters, and the slamming acceleration curve parameters are used as output parameters to train the slamming acceleration response curve model of the wedge-shaped body entering the water.

[0060] In this embodiment, the wedge-shaped body water-falling state parameters are used as the input parameters of the test sample, and the test results of the water-entry impact acceleration characteristics are used as the output parameters, and the wedge-shaped body water-entry impact response model is trained using a machine learning method based on a Gaussian process regression model. Specifically, in step S3, the process of using the training sample data to train the wedge-shaped body water-entry impact response model includes: training through a Gaussian process regression model, and in each training iteration, comparing the error of the wedge-shaped body water-entry impact response model until the error is minimized.

[0061] The error is the error between the impact acceleration value predicted by the wedge-shaped body water impact response model and the actual acceleration value. The kernel function, kernel function type and hyperparameters of the Gaussian process regression model are updated before each training iteration. For example, a Gaussian process regression model with different types of kernel functions and initial values ​​of hyperparameters is selected to train sample data, and the error between the impact acceleration value predicted by the model and the actual acceleration value is compared. Through continuous training, the kernel function type and hyperparameters that minimize the error of the Gaussian process acceleration model are obtained, such as Figure 7a and Figure 7b The same method can be used to train the time model, simulate the time changes on the impact acceleration curve under different falling water conditions, and find the kernel function and hyperparameters that minimize the error of the Gaussian process time model prediction results, as shown in Figure 8a and Figure 8b shown.

[0062] Furthermore, the method of predicting the wedge-shaped body water entry impact response model includes:

[0063]

[0064] In the prediction formula of the wedge-shaped body impact response based on the Gaussian process regression model, X represents the water-falling state data in the training sample data, Y represents the water-falling impact response data in the training sample data, and X * Indicates the water-falling state data in the test sample data, Y * represents the water impact response data in the test sample data, K represents the covariance matrix of the water-falling state data, represents the variance value of the Gaussian distribution of the falling water state data, I represents the unit matrix, N represents the Gaussian distribution, and the covariance matrix K(X,X)={k(x p ,x q )}, its constituent elements k(x p ,x q ) is the kernel function of the Gaussian process regression model, which is related to the spatial distribution of data points and has a decisive influence on the prediction accuracy of the Gaussian process regression model.

[0065] In the kernel function:

[0066]

[0067] Among them, x p Represents sample data p, x q represents sample data q, σ f represents a hyperparameter, λ represents a hyperparameter, and the value of the kernel function θ={σ f ,λ}. The kernel function of the Gaussian process regression model is also called the SEISO kernel function. When the input parameter X is given, the hyperparameter σ f , λ determines the value of the kernel function, denoted as θ={σ f ,λ}.

[0068] Assume x p =x q ,but

[0069]

[0070] Assuming that the mean of the Gaussian distribution is 0, then remember Then the observed value y i The probability density function p(y i ) is expressed as:

[0071]

[0072] According to the given input vector and the corresponding output value, we get n random variables y1, y2, ..., y n Observed value According to the maximum likelihood estimation method, we can get:

[0073]

[0074]

[0075] Assuming the noise variance but The unknown number r in the formula is an empirical value and needs to be determined according to the actual problem. n ] T Given, assuming |x p -x q | 2 The mean value is c 2 =(ΣΣ|x p -x q | 2 ) / n 2 , any two input quantities x p 、x q The kernel function k(x p ,x q ) has a mean of 0.5σ f 2 , that is, exp(-0.5c 2 / λ 2 )=0.5, then the initial value of the hyperparameter is as follows:

[0076]

[0077]

[0078]

[0079] After setting the initial value of the hyperparameter, the gradient method is used to obtain the optimal solution of the hyperparameter. The objective function is differentiated with respect to each parameter, and the function gradient value is continuously iterated so that it gradually decreases until it converges. Finally, the converged hyperparameters are solved and a Gaussian process regression model is established.

[0080] In this embodiment, in step S4, the process of detecting the accuracy of the model according to the output result includes: using absolute error, relative error and root-mean-square error to evaluate the accuracy of the wedge-shaped body water entry impact response model.

[0081] After the accuracy of the wedge-shaped body water impact response model is evaluated, a result comparison curve is generated and output to the client, wherein the result comparison curve records the acceleration response curve of the output result and the curve of the actual result. Specifically, the on-site personnel can determine whether the model accuracy reaches the preset model accuracy based on the result comparison curve displayed by the client. If it reaches the preset model accuracy, the client can be operated to input confirmation information, and the client sends the confirmation information to the server. After the server confirms, it stops updating the training sample data and stops executing steps S3 to S4, and the process terminates.

[0082] Specifically, in step S4, the wedge-shaped body water impact response model is tested and tested to determine its model accuracy. If the model accuracy is not reached, new learning samples are added to the training set again, and the learning process is repeated until a satisfactory accuracy is obtained. It can be understood that the model accuracy reaches the preset model accuracy. For example, the wedge-shaped body water impact response curve model test and accuracy evaluation method can be: the remaining 20% ​​of the samples in the test samples are used as test samples (the test samples are different from the training samples used when building the model), and the water-falling state parameters of the selected test samples are used as input items of the wedge-shaped body water impact acceleration curve model, and then the output items of the acceleration curve model are obtained through the trained Gaussian process model, and the absolute error, relative error and root-mean-square error are used to evaluate the accuracy of the acceleration response curve model.

[0083] Absolute error is the difference between the measured value (single measured value or the average of multiple measured values) and the true value. It can intuitively reflect the error between the measured value and the true value. The absolute error formula is as follows:

[0084] absolute error=|yy * | (12)

[0085] Where y represents the measured value, y * Represents a true value.

[0086] Relative error is the ratio of absolute error to true value. Compared with absolute error, relative error reflects the deviation of measurement relative to true value. The specific relative error formula is as follows:

[0087]

[0088] The root-mean-square error is the square root of the ratio of the square of the deviation between the observed value and the true value to the number of observations n in a finite number of measurements. It describes the deviation between the observed value and its simulated value. When the simulated value and the observed value differ greatly, the root-mean-square error is large. Therefore, the root-mean-square error is very sensitive to very large or very small errors in the group of measurements. The root-mean-square error formula is as follows:

[0089]

[0090] The absolute error comparison of each parameter of the acceleration curve is as follows: Fig. 9 The result comparison curve of a representation form is shown. It can be seen that the data points composed of the model's predicted value and the true value basically fall on y=x, and the predicted value and the true value are very close. And observe Fig.10 Another form of result comparison curve is shown. The acceleration response curve measured by the test sample and the actual result are basically on the same curve. The model prediction accuracy is high, indicating that the acceleration response curve model trained by the training sample can well predict the changes of each point on the acceleration response curve.

[0091] In addition, if Fig.11a and Fig.11b As shown in the figure, the relative error and root mean square error of the time model and acceleration model in the acceleration response curve are evaluated. In general, under 10% noise interference, the acceleration response curve model constructed by the Gaussian process model still obtains the test results very well. Both the deviation between the predicted value and the true value and the error caused by the extremely large predicted value are small, and the accuracy of the acceleration response curve model is high.

[0092] This embodiment also provides a system for predicting the acceleration of a wedge-shaped body entering water. Fig.13 As shown, it includes: server, client and experimental device.

[0093] The client is used to record initial data according to user operations and send the initial data to the server.

[0094] The server is used to receive the initial data and establish a wedge-shaped body water entry impact response model, wherein the initial data at least includes: the water-falling state parameters of the wedge-shaped body set by the client and the preset model accuracy, and the water-falling state parameters of the wedge-shaped body at least include: the half-width of the wedge-shaped body, the bottom ramp angle, the structural elastic modulus, the water-falling height and the structural weight.

[0095] The experimental device is used to record the original test data of the wedge-shaped body entering the water, and the server collects the original test data, and uses the original test data to obtain sample data, and extracts training sample data and test sample data from the sample data.

[0096] The server is further configured to input the initial data into the wedge-shaped body water entry impact response model, and train the wedge-shaped body water entry impact response model using the training sample data. Afterwards, the test sample data is input into the trained wedge-shaped body water entry impact response model to obtain an output result, and then the model accuracy is detected according to the output result, and transmitted to the client.

[0097] The solution provided in this embodiment can mine the existing wedge-shaped body water impact test data, avoid unnecessary duplication of work, deeply abstract and summarize the internal laws of the data, and predict the impact acceleration characteristics of the wedge-shaped body entering the water under any falling water state parameters at the fastest speed; the Gaussian process model proposed in this embodiment can obtain good model accuracy through model training, and play an engineering application role in the design of the water impact strength of the structure; the Gaussian process proposed in this embodiment is an artificial intelligence model, which can ignore the influence of random errors occurring in the test on the test results to a certain extent, and has a certain anti-interference ability.

[0098] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The above is only a specific implementation method of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or replacements that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for predicting the impact acceleration of a wedge-shaped body entering water, characterized in that: include: S1. The server receives initial data sent by the client and establishes a wedge-shaped body water entry impact response model, wherein the initial data at least includes: the water entry state parameters of the wedge-shaped body set by the client, and the water entry state parameters of the wedge-shaped body at least include: the half width of the wedge-shaped body, the bottom ramp angle, the structural elastic modulus, the water entry height and the structural weight; S2, collecting original test data of the wedge-shaped body entering water and impacting, using the original test data to obtain sample data, and extracting training sample data and test sample data from the sample data; S3, inputting the initial data into the wedge-shaped body water entry impact response model, and using the training sample data to train the wedge-shaped body water entry impact response model; S4, inputting the test sample data into the trained wedge-shaped body water impact response model and obtaining an output result, and then detecting the model accuracy according to the output result, wherein, when the model accuracy does not reach the preset model accuracy, updating the training sample data and repeating steps S3 to S4 until the model accuracy reaches the preset model accuracy.

2. The method according to claim 1, characterized in that: Step S1 includes: The established wedge-shaped body water impact response model is expressed as: The input parameters are consistent with the parameters of the wedge falling into the water, including: b represents the half width of the wedge, α represents the bottom slope angle, E represents the structural elastic modulus, H represents the height of the falling water, and M represents the structural weight. The output parameters include: the time of touching the water T1, the time of peak acceleration T2, the time of constant acceleration T3, the total time T4, and the acceleration value a at time j. j , j represents the time value.

3. The method according to claim 2, characterized in that In step S2, the collected original test data of the wedge-shaped body entering water impact includes: the slamming acceleration response of the wedge-shaped body entering water impact process, wherein the slamming acceleration response is represented by a continuous curve recorded in real time, and the continuous curve is divided into T1, T2 and T3 segments.

4. The method according to claim 1, characterized in that: In step S2, the process of obtaining sample data using the original test data includes: Performing interpolation construction processing on the original test data to obtain the sample data, wherein the interpolation construction processing includes: Among them, x i represents the input of sample data obtained by interpolation, y i represents the output of the sample data obtained by interpolation, x0 represents the minimum value of the interpolation interval in the original input, y0 represents the minimum value of the interpolation interval in the original output, i represents the subscript of the sample data obtained by interpolation, x n Indicates the maximum value of the interpolation interval in the original input, y n represents the maximum value of the interpolation interval in the original output, and n represents the subscript of the maximum value of the interpolation interval in the original sample; In the process of performing interpolation construction processing on the original test data, it also includes: adding random noise to simulate actual errors, wherein the manner of adding random noise includes: Among them, ε i represents the random noise, Represents the output of the sample data after adding random noise, y i represents the output of the sample data to which random noise is to be added, i represents the subscript of the sample data, and N represents Gaussian distribution.

5. The method according to claim 1, characterized in that In step S2, the process of extracting training sample data and test sample data from the sample data includes: After acquiring the sample data, a sample space is established, wherein the sample space includes: a corresponding relationship between the falling water state parameters of the wedge-shaped body and the water impact acceleration characteristics; The samples in the sample space are randomly sampled to obtain the training sample data and the test sample data.

6. The method according to claim 5, characterized in that The samples in the sample space are randomly sampled, including: After determining the sampling size n0, in the sample space, the variable x of each dimension is i The domain interval [l i ,u i ] is divided into n0 equal parts, so that the sample space is n0×n0 hypercubes, where the coordinate point on the dimension is used as the variable x in each dimension. i , l i is the lower limit of the dimension, u i is the upper limit of the dimension, and i is the dimension subscript; Randomly generate an n0-order matrix U, where U includes all permutations of random numbers {1,2,…n}; Afterwards, a sample point is randomly selected from each row in the matrix U, and a total of n0 sample points are obtained.

7. The method according to claim 2, characterized in that In step S3, the process of using the training sample data to train the wedge-shaped body water entry impact response model includes: The Gaussian process regression model is used for training, and in each training iteration, the error of the wedge-shaped body water entry impact response model is compared until the error is minimized, wherein the error is the error between the impact acceleration value predicted by the wedge-shaped body water entry impact response model and the actual acceleration value, and the kernel function, kernel function type and hyperparameters of the Gaussian process regression model are updated before each training iteration.

8. The method according to claim 7, characterized in that The method of predicting the wedge-shaped body water entry impact response model includes: Among them, X represents the water-falling state data in the training sample data, Y represents the water-entry impact response data in the training sample data, and X * Indicates the falling water state data in the test sample data, Y * represents the water impact response data in the test sample data, K represents the covariance matrix of the water-falling state data, represents the variance value of the Gaussian distribution of the falling water state data, I represents the unit matrix, N represents the Gaussian distribution, and the covariance matrix K(X,X)={k(x p ,x q )},k(x p ,x q ) is the kernel function of the Gaussian process regression model; In the kernel function, x p Represents sample data p, x q represents sample data q, σ f represents a hyperparameter, λ represents a hyperparameter, and the value of the kernel function θ={σ f ,λ}.

9. The method according to claim 8, characterized in that In step S4, the process of detecting the accuracy of the model according to the output result includes: using absolute error, relative error and root-mean-square error to evaluate the accuracy of the wedge-shaped body water entry impact response model; After the accuracy of the wedge-shaped body water entry impact response model is evaluated, a result comparison curve graph is generated and output to the client, wherein the result comparison curve graph records the acceleration response curve of the output result and the curve of the actual result.

10. A system for predicting the acceleration of a wedge-shaped body entering water, characterized in that: include: Server, client, and experimental setup; The client is used to record initial data according to the user's operation and send the initial data to the server; The server is used to receive the initial data and establish a wedge-shaped body water entry impact response model, wherein the initial data at least includes: the wedge-shaped body water entry state parameters and the preset model accuracy set by the client, and the wedge-shaped body water entry state parameters at least include: the half width of the wedge-shaped body, the bottom ramp angle, the structural elastic modulus, the water entry height and the structural weight; The experimental device is used to record the original test data of the wedge-shaped body entering the water, and the server collects the original test data, and uses the original test data to obtain sample data, and extracts training sample data and test sample data from the sample data; The server is also used to input the initial data into the wedge-shaped body water entry impact response model, and use the training sample data to train the wedge-shaped body water entry impact response model; then, input the test sample data into the trained wedge-shaped body water entry impact response model and obtain an output result, then detect the model accuracy based on the output result and transmit it to the client.

Images