Large-size wing assembly jig temperature and thermal deformation prediction method

Abstract

The application discloses a large-size wing assembly jig temperature and deformation prediction method, a jig temperature mathematical model based on a learning rate optimization BP neural network is established by measuring the jig temperature data in a whole year, three schemes for predicting the jig temperature characteristics are provided, the jig temperature of the next working stage can be predicted by the established temperature model, and the thermal deformation characteristics of the jig are analyzed. The method can quickly, accurately and at low cost predict the large-size wing assembly jig temperature and thermal deformation characteristics.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of assembly in the process of aircraft manufacturing, and relates to a large-size wing assembly jig temperature and deformation prediction method. BACKGROUND

[0002] Temperature is an important factor affecting the assembly accuracy and reliability of an aircraft, especially in the assembly process of a new large-size aircraft. Due to the large size of the corresponding assembly tooling, a small temperature change will cause a large thermal deformation, resulting in assembly residual stress or local stress concentration in the internal structure of the aircraft after assembly, which affects the service life of the aircraft. In addition, with the development of digital assembly technology, the ERS system of the assembly jig is also affected by temperature, causing a large system deviation and a large number of inspection problems. Therefore, to ensure the assembly and measurement accuracy of the aircraft, the temperature characteristics of the large-size jig must be accurately characterized.

[0003] At present, in the research on the temperature influence of the aircraft jig, the temperature of the jig is generally regarded as linear, and it is considered that the temperature of the jig is uniformly distributed. However, with the development of large-size aircraft, the size of the assembly jig is also getting larger and larger, and the assembly cycle is long. In terms of space and time dimensions, temperature shows strong nonlinearity. At this time, the temperature at a certain point of the jig is regarded as the uniform temperature of the jig, which will cause a large error. In addition to using the temperature at one point of the jig as the overall temperature characteristics, the second scheme is to directly observe the temperature characteristics of the aircraft jig. The temperature field can be recorded by an infrared thermal imaging camera, and the temperature data has high precision. However, for a large-size jig, complete measurement and recording of the temperature will consume a lot of time and manpower, which is difficult to realize in actual engineering.

[0004] Therefore, how to obtain the temperature distribution characteristics of the assembly jig through certain representative data and accurately predict them is the difficulty and key of the research on the temperature characteristics of the jig. After the temperature characteristics of the jig are accurately analyzed and predicted, the thermal deformation of the jig can be based on the discrete idea, that is, the expansion plate and the frame of the jig are divided into a plurality of measurement units composed of temperature measurement points, and then the thermal deformation characteristics of the jig are obtained. SUMMARY

[0005] The application provides a large-size wing assembly jig temperature and thermal deformation prediction method, which accumulates temperature data throughout the year, establishes a large-size wing assembly jig temperature model based on a BP neural network, predicts the temperature of the jig in a new stage, and analyzes the thermal deformation of the jig caused by temperature.

[0006] To achieve the above purpose, the application adopts the following technical scheme:

[0007] A large-size wing assembly jig temperature and thermal deformation prediction method, comprising the following steps:

[0008] Step one, the large size wing assembly jig includes a frame and an expansion plate, a plurality of temperature measuring points are arranged on the frame and the expansion plate, the temperature of all temperature measuring points is measured and recorded at a fixed time, the arrangement interval of the temperature measuring points is not greater than 3000mm, and the temperature measuring points are located at the ERS points or the wing positioner.

[0009] Step two, according to the temperature data of the temperature measuring points, a temperature distribution model based on a BP neural network optimized by a learning rate is established.

[0010] The temperature distribution model based on the BP neural network optimized by the learning rate is a three-layer BP neural network model, which includes an input layer, a hidden layer and an output layer, the characteristics of the input layer include time, environment temperature and jig structure temperature itself, the characteristics of the output layer are the temperatures of the temperature measuring points to be predicted at the working time, and the establishment of the temperature distribution model needs to determine three principles: hidden layer node number determination principle, learning rate optimization principle and input feature composition principle.

[0011] The hidden layer node setting principle is:

[0012]

[0013] Wherein n is the number of hidden layers, n inp is the number of input layer nodes, n out is the number of output layer nodes, m1 and m2 are input layer and output layer weight coefficients respectively, the values of m1 and m2 are in the range of 1-2, and m c is an adjustment constant in the range of 0-10.

[0014] The optimization learning rate establishment principle is: a larger learning rate is adopted in the early network training process to obtain better global search ability, and a smaller learning rate is adopted in the later period to obtain higher precision local approximation value, and the learning rate optimization formula is as follows:

[0015]

[0016] Wherein lrmax is the maximum learning rate, lrmin is the minimum learning rate, k is the current iteration number, Max_iter is the maximum iteration number, a is a dynamic adjustment factor, and a is

[0017] a = 0.4 * (1-k / Max_iter) 0.25 (3)

[0018] There are three kinds of input feature determination schemes, which are: (1) temperature prediction scheme taking time and temperature of the jig structure itself as input features; (2) temperature prediction scheme taking time and environment temperature as input features, wherein the environment temperature includes daily maximum temperature, minimum temperature and average temperature; (3) temperature prediction scheme taking time, temperature of the jig itself and environment temperature as common input features.

[0019] Step three, obtaining temperature characteristics of the jig to be assembled in the future time through the temperature distribution model;

[0020] Step four, calculating thermal deformation amount of the jig in the future time according to the temperature characteristics.

[0021] The calculation principle of the thermal deformation amount is to divide the deformation of the jig into the deformation of the expansion plate and the deformation of the frame.

[0022] The deformation determination principle of the expansion plate is that the expansion plate is connected to the jig through two groups of slide rails, one end of which is fixed and the other end of which can deform along the direction of the slide rail, which can be regarded as a one-dimensional model, and the thermal deformation amount D of the expansion plate caused by temperature is i which can be described by the following formula:

[0023]

[0024] where i represents a measurement point, L p is the total length of the expansion plate, α p represents the thermal expansion coefficient of the expansion plate, T i represents the temperature of the i-th measurement point, n p represents the number of measurement points of the expansion plate.

[0025] The deformation determination principle of the frame is that the overall structure of the frame can be simplified as a left vertical column, an inclined beam and a right vertical column, wherein the inclined beam is connected to the wing, the bottom ends of the left and right vertical columns are fixed, the two ends of the inclined beam are connected to the left and right vertical columns respectively, the deformation D of the inclined beam caused by temperature change is j which is composed of three parts, and is the superposition of the thermal deformation of the inclined beam and the deformation of the left and right vertical columns, and the specific expression is as follows:

[0026]

[0027] where j represents a measurement point of the frame inclined beam, L f is the length of the frame inclined beam, L l is the height of the left vertical column of the frame, L r is the height of the right vertical column of the frame, α f is the thermal expansion coefficient of the frame, T j is the temperature of the j-th measurement point of the inclined beam, n f represents the number of measurement points of the inclined beam, n l represents the number of measurement points of the left vertical column, n rT kl and T kr are the temperature of the left and right column measurement points, respectively.

[0028] Beneficial effects: The large-size wing assembly jig temperature and thermal deformation prediction method established by the application can quickly, accurately and at low cost predict the temperature and thermal deformation characteristics of the large-size wing assembly jig. Specifically, the method is based on the temperature data of each temperature measurement point of the wing jig, and a BP neural network model with optimized learning rate is established. The expansion plate and frame of the jig are discretized into measurement units composed of temperature measurement points, and the neural network temperature distribution model can accurately analyze and predict the temperature variation characteristics of each temperature measurement point of the jig with space and time, and obtain the thermal deformation characteristics of the expansion plate and frame of the jig.

[0029] The embodiments of the application will be further described in detail below with reference to the drawings and specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 Structure diagram of large-size wing assembly jig

[0031] Figure 2 Temperature measurement point distribution diagram of large-size wing assembly jig

[0032] Figure 3 Temperature prediction results of each measurement point of the large-size jig expansion plate on February 4, 2021

[0033] Figure 4 Temperature prediction results of each measurement point of the large-size jig expansion plate on July 22, 2021

[0034] Figure 5 Thermal deformation curve diagram of each point of the large-size jig expansion plate (P1-P9) on July 22, 2021

[0035] Figure 6 Thermal deformation curve diagram of each point of the large-size jig frame diagonal beam (K1-K19) on July 22, 2021

[0036] Explanation of numbers in the figure: 1, frame; 2, expansion plate; 3, slide rail DETAILED DESCRIPTION

[0037] The following examples are used to illustrate the application, but cannot be used to limit the scope of the application.

[0038] A large-size wing assembly jig temperature and thermal deformation prediction method comprises:

[0039] Step one, the large-size wing assembly jig comprises a frame 1 and an expansion plate 2, the frame 1 and the expansion plate 2 are connected through a slide rail 3, seeFigure 1 , frame and expansion plate set up a plurality of temperature measuring points, at a fixed time on all temperature measuring points temperature measurement and record, here the measurement time interval is 14 days, temperature measuring point spacing is 3000mm, in the mold ERS point and the positioner to increase the corresponding temperature measuring points, expansion plate is arranged with 9 temperature measuring points P1-P9, frame is arranged with 19 temperature measuring points K1-K 19 , as Figure 2 shown;

[0040] Step two, according to the temperature data of temperature measuring points, the temperature distribution model based on learning rate optimization of BP neural network, the temperature distribution model is three layer BP neural network model, including input layer, hidden layer and output layer, the characteristics of input layer include time, environmental temperature and mold structure temperature itself, the characteristics of output layer is the temperature of temperature measuring points of working time to be predicted, need to determine the number of hidden layer nodes, learning rate optimization principle and the feature composition of input layer. Based on the different input layer characteristics, the temperature distribution model determination scheme can be divided into three schemes:

[0041] The first embodiment, time and mold structure temperature of a point as input characteristics of temperature prediction scheme, taking the expansion plate temperature analysis and prediction as an example, taking time and expansion plate temperature of a point as input characteristics, marked as M1;

[0042]

[0043] Where Day n n represents the nth day, T mn is the temperature of the nth day of the mth measuring point,

[0044] The principle of setting the number of nodes in the hidden layer is:

[0045]

[0046] Here n inp is 2, n out is 8, m1 is set to 1.2, m2 is set to 1.6, m c is 9, n is 13,

[0047] The second embodiment, time and environmental temperature as input characteristics of temperature prediction scheme, taking the expansion plate temperature analysis and prediction as an example, taking time and expansion plate temperature of a point as input characteristics, marked as M2;

[0048]

[0049] Where Day n n represents the nth day, T mn is the temperature of the nth day of the mth measuring point, T maxnTn is the maximum temperature of the nth day, T minn Tn is the minimum temperature of the nth day, T aven Tn is the average temperature of the nth day, the maximum temperature, the minimum temperature and the average temperature can be obtained by network query.

[0050] The principle of setting the hidden layer nodes is:

[0051]

[0052] Here n inp is 4, n out is 9, m1 is set to 1.6, m2 is set to 1.9, m c is 10, n is 15,

[0053] The third embodiment, time, expanded plate point temperature and environment temperature as input features, denoted as M3.

[0054]

[0055] Where Day n n represents the nth day, T mn Tnm is the temperature of the nth day of the mth measurement point, T maxn Tn is the maximum temperature of the nth day, T minn Tn is the minimum temperature of the nth day, T aven Tn is the average temperature of the nth day,

[0056] The principle of setting the hidden layer nodes is:

[0057]

[0058] Here n inp is 5, n out is 8, m1 is set to 1.9, m2 is set to 1.9, m c is 10, n is 15,

[0059] After determining the model input scheme, the learning rate is optimized, and the learning rate optimization formula is established as follows:

[0060]

[0061] Where lrmax is the maximum learning rate, lrmin is the minimum learning rate, k is the current iteration number, Max_iter is the maximum iteration number, a is the dynamic adjustment factor, a is

[0062] a = 0.4 * (1-k / Max_iter) 0.25 (12)

[0063] Here, lrmax is the maximum learning rate of 0.1, lrmin is the minimum learning rate of 0.005, and Max_iter is the maximum iteration number of 10000.

[0064] Step three, through the established temperature distribution model, the temperature distribution characteristics of the mold frame in the temperature measurement period are obtained, and the temperature characteristics of the mold frame to be assembled in the future time are predicted. Here, the mold frame temperature measurement time is from March 1, 2020 to March 1, 2021, and the prediction result is shown in Figure 3 、 Figure 4 , Figure 3 In order to predict and compare the temperature of the temperature measurement point in the measurement time range (February 4, 2021), Figure 4 It can be found that when only the time and the temperature of the expansion plate are input as factors affecting the temperature characteristics of the mold frame, the predicted temperature of the expansion plate is close to the true value, but the trend distribution is quite different from the true measurement value; when only the time and the ambient temperature are input as factors affecting the temperature characteristics of the mold frame, the predicted temperature of the expansion plate has a good prediction trend, but there is a certain error with the true value, and when the time, ambient temperature and structure temperature are input as factors affecting the temperature of the expansion plate, the expansion plate temperature model is more accurate, the trend is closer to the true temperature, and the error can be controlled within 0.4℃.

[0065] Step four, according to the predicted temperature characteristics, the thermal deformation amount of the mold frame required for evaluation is calculated. The calculation principle of thermal deformation amount is to divide the deformation of the mold frame into two parts of the expansion plate and the frame.

[0066] The expansion plate is connected to the mold frame by two groups of sliding rails, one end of which is fixed and the other end can deform along the sliding rail direction, which can be regarded as a one-dimensional model, and the thermal deformation amount D i It can be described by the following formula:

[0067]

[0068] Where i represents the measurement point, L p is the total length of the expansion plate, a p represents the thermal expansion coefficient of the expansion plate, T i represents the temperature of the i-th measurement point, and n p represents the number of measurement points of the expansion plate. Here, the material of the expansion plate is aluminum, a p is 2.32e-5 / ℃, and n p is 9.

[0069] The deformation determination principle of the frame is that the overall structure of the frame can be simplified as a left vertical column, an inclined beam, and a right vertical column, wherein the inclined beam is connected with the wing, the bottom ends of the left and right vertical columns are fixed, the two ends of the inclined beam are connected with the left and right vertical columns respectively, and the deformation D of the inclined beam caused by temperature change j The deformation is composed of three parts, which are the superposition of the thermal deformation of the inclined beam and the deformations of the left and right vertical columns, and the specific expression is as follows:

[0070]

[0071] Wherein j represents a measurement point of the inclined beam of the frame, L f represents the length of the inclined beam of the frame, L l represents the height of the left vertical column of the frame, L r represents the height of the right vertical column of the frame, a f represents the thermal expansion coefficient of the frame, T j represents the temperature of the jth measurement point of the inclined beam, n f represents the number of measurement points of the inclined beam, n l represents the number of measurement points of the left vertical column, n r represents the number of measurement points of the right vertical column, T kl and T kr respectively represent the temperatures of the measurement points of the left and right vertical columns. The material of the frame is steel, a f is 1.15e-5 / ℃, n f is 12, n l is 5, and n r is 2. The thermal deformations of the expansion plate and the inclined beam of the frame at the pre-evaluation time (July 22, 2021) are calculated, the thermal deformation of the expansion plate is Figure 5 , and the thermal deformation of the frame is Figure 6 .

Claims

1. A method for temperature and thermal distortion prediction of large size wing assembly jigs, characterized in that The method comprises the following steps: Step one, the large-size wing assembly jig frame comprises a frame and an expansion plate, a plurality of temperature measuring points are arranged on the frame and the expansion plate, the temperature of all the temperature measuring points is measured and recorded at a fixed time; Step two, a temperature distribution model based on a learning rate optimized BP neural network is established according to the temperature data of the temperature measuring points, the temperature distribution model based on the learning rate optimized BP neural network is a three-layer BP neural network model, and comprises an input layer, a hidden layer and an output layer, the features of the input layer include time, ambient temperature and the temperature of the jig frame structure itself, the features of the output layer are the temperature of the temperature measuring point at the working time to be predicted, and the establishment of the temperature distribution model needs to determine three principles: a hidden layer node number determination principle, a learning rate optimization principle and an input feature composition principle. The hidden layer node number determination principle is that: where n is the number of hidden layers, n inp is the number of input layer nodes, n out is the number of output layer nodes, m1, m2 are the weight coefficients of the input layer and the output layer, respectively, the values of m1, m2 range from 1 to 2, m c is an adjustment constant ranging from 0 to 10, The learning rate optimization principle is that a larger learning rate is adopted in the early network training process to obtain better global search capability, and a smaller learning rate is adopted in the later period to obtain higher precision local approximation value, and the learning rate optimization formula is as follows: Wherein, lrmax is the maximum learning rate, lrmin is the minimum learning rate, k is the current iteration number, Max_iter is the maximum iteration number, a is a dynamic adjustment factor, and a is a = 0.4 * (1 - k / Max iter) 0.25 , The input feature composition principle is that the temperature prediction scheme of taking time and the temperature of a point of the jig frame structure itself as input features or the temperature prediction scheme of taking time and ambient temperature as input features, wherein the ambient temperature includes daily maximum temperature, minimum temperature and average temperature, or the temperature prediction scheme of taking time, the temperature of the jig frame itself and ambient temperature as common input features. Step three, the temperature characteristics of the jig frame to be assembled in the future time are obtained through the temperature distribution model. Step four, the thermal deformation amount of the jig frame in the future time is calculated according to the temperature characteristics.

2. The method of claim 1, wherein The temperature measuring point arrangement interval is not greater than 3000mm, and the temperature measuring point is located at an ERS point or a wing positioner.

3. The method of claim 1, wherein The thermal deformation amount calculation principle is that the deformation of the jig frame is divided into the deformation of the expansion plate and the frame.

4. The method of claim 3, wherein The deformation determination principle of the expansion plate is that the expansion plate is connected to the mold frame through two groups of slide rails, one end is fixed, and the other end is deformed along the slide rail direction, which is regarded as a one-dimensional model, and the thermal deformation amount D caused by temperature is i Described by the following formula: where i represents the measurement point, a p represents the thermal expansion coefficient of the expansion plate, T i represents the temperature of the i-th measurement point, n p represents the number of measurement points of the expansion plate, L p is the total length of the expansion plate.

5. The method of claim 3, wherein The deformation determination principle of the frame is that the overall structure of the frame can be simplified as a left vertical column, an inclined beam and a right vertical column, wherein the inclined beam is connected with the wing, the bottom ends of the left and right vertical columns are fixed, the two ends of the inclined beam are connected with the left and right vertical columns respectively, the deformation D of the inclined beam is caused by temperature change j The deformation is composed of three parts, which are the superposition of the thermal deformation of the inclined beam and the deformation of the left and right vertical columns, and the specific expression is as follows: where j represents the frame diagonal beam measurement point, L f is the frame diagonal beam length, L l is the frame left side column height, L r is the frame right side column height, a f is the frame thermal expansion coefficient, T j is the jthdiagonal beam measurement point temperature, n f represents the number of diagonal beam measurement points, n l represents the number of left side column measurement points, n r is the number of right side column measurement points, T kl and T kr are the left side and right side column measurement point temperatures, respectively.

Images