Method for measuring icing meteorological parameters

Abstract

The invention discloses a method for measuring icing meteorological parameters, which comprises the following steps of: measuring pressure data under various wind speeds and wind directions by using an icing meteorological parameter measuring device, and constructing a first database through the data; training a BP neural network N1 optimized based on a particle swarm algorithm according to the first database to obtain a first artificial neural network for calculating the wind speed and the wind direction according to the pressure intensity; an icing meteorological parameter measuring device is used for measuring icing ice shapes on the detection points under various icing conditions, data of effective ice thickness and icing rate are obtained, and a second database is established; a BP neural network N2 optimized based on a dung beetle algorithm is trained according to the second database, and a second artificial neural network for calculating icing meteorological parameters, namely the average volume diameter of water drops and the liquid water content according to the icing thickness, the wind speed and the wind direction is obtained; according to the method, complex light path construction is avoided, and the method can be used for real-time measurement of icing meteorological parameters during an engine icing test and airplane airworthiness verification flight.

Description

Technical Field

[0001] The present invention relates to the field of aerospace, and particularly relates to a method for measuring icing meteorological parameters. Background Art

[0002] The existing measurement of the mean volumetric diameter (MVD) of water droplets and the liquid water content (LWC) is mostly divided into the rotating multi-cylinder method and the optical measurement method. Among them, the rotating multi-cylinder method is one of the important methods for calibration in an icing wind tunnel. In this method, multiple cylinders with different diameters are installed on a concentric shaft and rotated at a low speed under icing conditions. After a period of time, by measuring the icing thickness and density of each cylinder, the LWC and MVD of the oncoming flow are calculated. This method has reliable measurement accuracy, but it can only be used for calibrating icing environmental parameters and cannot perform real-time measurement. The basic principle of optical measurement is to use a Charge Coupled Device (CCD) to record the fluctuations of laser scattering and interference fringes caused by droplets of different diameters, and calculate the average diameter of the droplets. For MVD measurement, optical instruments such as a Forward Scattering Spectrometer Probe (FSSP), an Optical Array Processor (OAP), a Malvern Particle Size Analyzer (MPSA), and a Phase Doppler Particle Analyzer (PDPA) can be used. However, most of these above measurement methods cannot meet the requirements of real-time measurement, and the optical devices that can achieve real-time measurement have great limitations in use. For example, when it is necessary to measure the icing meteorological parameters behind the fan in an engine, it is difficult to arrange the optical path. Summary of the Invention

[0003] To solve the above problems, the present invention provides a solution that avoids complex optical path construction and can be used for real-time measurement of icing meteorological parameters during engine icing tests and aircraft airworthiness verification flights.

[0004] To achieve the above object, the present invention provides a method for measuring icing meteorological parameters, including the following steps:

[0005] Step 1: Measure data under various wind speeds and wind directions using the above-mentioned icing meteorological parameter measurement device, and construct a first database of the oncoming flow wind speed and wind direction corresponding to the pressure based on this data;

[0006] Step 2: Train the BP neural network N1 optimized by the particle swarm algorithm according to the first database to obtain a first artificial neural network for calculating the oncoming flow wind speed and wind direction based on the pressure;

[0007] Step 3: Measure the ice shapes at the detection points under various icing conditions using the above-mentioned icing meteorological parameter measurement device, obtain data on the effective ice thickness and icing rate, and establish a second database;

[0008] Step 4: Train the BP neural network N2 optimized based on the dung beetle algorithm according to the second database to obtain a second artificial neural network for calculating the average volume diameter of water droplets and the liquid water content of icing meteorological parameters based on the icing thickness and wind speed and direction.

[0009] Optionally, the icing meteorological parameter measuring device includes: a disc-shaped base, the edge of the base bulges towards one side to form an open cavity, the center of the cavity is connected to one end of a connecting rod, and the other end of the connecting rod is connected with a hemispherical accommodating cavity. A plurality of icing ultrasonic sensors for measuring the ice layer thickness are arranged inside the base, and a pressure sensor is arranged in the accommodating cavity.

[0010] Optionally, the step 2 includes:

[0011] The input condition of the BP neural network N1 is the pressure corresponding to different positions in the first database, and the output is the corresponding oncoming flow wind speed and direction;

[0012] Train the BP neural network N1 using the particle swarm algorithm;

[0013] During the training process, use the particle swarm algorithm to optimize the initial threshold and weight matrix of the BP neural network N1 to determine the optimal threshold and weight;

[0014] Use the trained BP neural network N1 to predict the oncoming flow wind speed and direction.

[0015] Optionally, training the BP neural network N1 using the particle swarm algorithm includes:

[0016] Randomly generate a cluster and initialize the particles and particle velocities in the cluster;

[0017] Obtain the initial weights and thresholds of the BP neural network N1 according to the individuals. After training the BP neural network N1 with the training data, predict the system output. Respectively take the absolute sum of the errors between the predicted output and the expected output as the individual fitness value, the maximum value of the errors between the predicted output and the expected output as the individual fitness value, and the sum of the maximum value and the average value of the errors between the predicted output and the expected output as the individual fitness value;

[0018] Adopt three fitness functions and compare the optimization accuracy of the algorithm under different fitness functions;

[0019] ;

[0020] ;

[0021] ;

[0022] Where: n is the number of output nodes of the network; is the expected output of the i-th node of the BP neural network; is the predicted output of the i-th node; l is a coefficient;

[0023] In each iteration process, the particle updates its velocity and position through the individual extreme value and the global extreme value, and the update formula is:

[0024] ;

[0025] ;

[0026] In the formula: is the inertia weight; ; is the current iteration number; is the velocity of the particle; and are non-negative constants; and are random numbers distributed between ; is the position of the particle, is the individual extreme value, is the global extreme value.

[0027] When the fitness value of the particle is less than the individual extreme value, the new fitness value is replaced by the individual extreme value; when the individual extreme value of the particle is less than the global extreme value of the population, the individual extreme value is replaced by the global extreme value of the population, and the update formula is:

[0028] ;

[0029] In the formula: is the fitness value of the new particle.

[0030] Optionally, the step 3 includes:

[0031] Setting multiple icing conditions according to the relationship between the average volume diameter of water droplets, the liquid water content and the temperature;

[0032] Calculating the ice shape under the corresponding icing conditions according to the icing conditions;

[0033] Selecting appropriate measuring point positions and numbers on the above-mentioned icing meteorological parameter measuring device according to the different ice shapes, obtaining the effective icing thickness at the measuring points, and establishing a second database.

[0034] Optionally, the second database is established through the following steps:

[0035] Obtaining the icing thickness L and the corresponding icing parameters and the oncoming flow parameter vector Y through numerical calculation or experiment, and the icing thickness L is measured by an icing detection ultrasonic sensor;

[0036] The Kriging algorithm is used to construct a mathematical surrogate model. The icing parameters and the incoming flow parameter vector Y include temperature, average volume diameter of water droplets, liquid water content, and incoming flow wind speed and direction.

[0037] The mathematical surrogate model takes the temperature, incoming flow wind speed and direction, and icing thickness L in Y as inputs, and the average volume diameter of water droplets and liquid water content as outputs, and generates a second database for training the neural network N2.

[0038] Optionally, step 4 includes:

[0039] The input conditions of the BP neural network N2 are the effective icing thickness, ambient temperature, wind speed, and direction in the icing database, and the output conditions are the meteorological parameters of the average volume diameter of water droplets and liquid water content.

[0040] The dung beetle algorithm is used to train the BP neural network N2.

[0041] During the training process, the dung beetle algorithm is used to optimize the initial threshold and weight matrix of the BP neural network N2 to determine the optimal threshold and weight values.

[0042] After using the trained BP neural network N2, the meteorological parameters of the average volume diameter of water droplets and liquid water content are predicted.

[0043] Optionally, using the dung beetle algorithm to train the BP neural network N2 includes:

[0044] The position update method of the rolling ball dung beetle is as follows:

[0045] ;

[0046] ;

[0047] where t represents the current iteration number, represents the position information of the i-th dung beetle at the t-th iteration, is a natural coefficient indicating whether to deviate from the original direction, assigned as -1 or 1 according to the probability method, k ∈ (0, 0.2) represents the deflection coefficient, b ∈ (0, 1) represents a constant, and k and b are set to 0.1 and 0.3 respectively, represents the global worst position, used to simulate the change of light intensity;

[0048] The formula for updating the position of the dancing rolling ball dung beetle is defined as follows:

[0049] ;

[0050] where ∈ [0, π] represents the deflection angle, at When it is equal to 0, π / 2 or π, the position of the dung beetle will not be updated;

[0051] The brood ball adopts a boundary selection strategy to simulate the egg-laying area of the female dung beetle. The definition of the egg-laying area is:

[0052] ;

[0053] ;

[0054] In the formula, represents the current local best position, and represent the lower and upper limits of the egg-laying area respectively, and R = 1 - t / T max , T max represents the maximum number of iterations, and Lb and Ub represent the lower and upper limits of the optimization problem respectively;

[0055] During the iteration process, the position of the brood ball is dynamically changing, and its definition is:

[0056] ;

[0057] In the formula, is the position information of the i-th brood ball at the t-th iteration. b1 and b2 represent two independent random vectors of size 1×D, and D represents the dimension of the optimization problem;

[0058] An optimal foraging area is established to guide the larval dung beetles to find food and simulate their foraging behavior. The definition of the optimal foraging area is:

[0059] ;

[0060] ;

[0061] In the formula, represents the current local best position, and represent the lower and upper limits of the best foraging area respectively;

[0062] The position of the small dung beetle is updated as follows:

[0063] ;

[0064] In the formula, is the position information of the i-th small dung beetle at the t-th iteration. C1 represents a random number following a normal distribution, and C2 ∈ (0, 1) represents a random vector;

[0065] The position update method of the thief dung beetle is as follows:

[0066] ;

[0067] Wherein, g represents a random vector of size 1×D following a normal distribution, S represents a constant value, and X b is the optimal position for food competition.

[0068] The beneficial effects of the present invention compared with the prior art are as follows: The local wind speed and direction are obtained by resolving the pressure sensors arranged in the hemispherical cavity of the device, and then the ice thickness is measured by the ice detection ultrasonic sensors installed in the disc-shaped base. Combining the local wind speed and direction, the local icing meteorological parameters, namely the average volume diameter of water droplets and the liquid water content, are calculated in real time. This device avoids the complex optical path construction and can be used for the real-time measurement of icing meteorological parameters during engine icing tests and aircraft airworthiness verification flights. Description of the Drawings

[0069] Figure 1 is a structural diagram of an icing meteorological parameter measurement device provided by an embodiment of the present invention;

[0070] Figure 2 is a flowchart of an icing meteorological parameter measurement method provided by an embodiment of the present invention. Detailed Embodiments

[0071] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0072] Referring to Figure 1 , this embodiment provides an icing meteorological parameter measurement device, including: a disc-shaped base 1, with the edge of the base 1 protruding towards one side to form an open cavity. The center of the cavity is connected to one end of a connecting rod 3, and the other end of the connecting rod 3 is connected to a hemispherical accommodation cavity 4. A plurality of ice detection ultrasonic sensors 5 for measuring the ice layer thickness are arranged inside the disc-shaped base, and a pressure sensor for measuring the wind speed and direction (not shown in the figure) is arranged in the accommodation cavity 4.

[0073] In this embodiment, the diameter of the base 1 is 2 cm and the thickness is 0.5 cm, and the diameter of the accommodation cavity 4 is 0.5 cm.

[0074] The number and arrangement scheme of the pressure sensors and the ice detection ultrasonic sensors. Five pressure sensors are arranged on the accommodation cavity 4; twenty ice detection ultrasonic sensors are arranged on the disc-shaped base. It should be noted that the arrangement method and the number can be adjusted to be denser or sparser according to the required accuracy.

[0075] The working principle of the device is as follows:

[0076] Under different incoming flow wind speeds and directions, the pressure distribution on the surface of the hemispherical accommodating cavity 4 of this device is different. Therefore, several pressure sensors arranged in the hemispherical accommodating cavity can measure the pressure at corresponding positions on the hemispherical surface. By using the pressure at specific points on the surface as the input and the incoming flow wind speed and direction as the output to establish a mathematical model N1 in advance, the incoming flow wind speed and direction can be calculated in real time according to the surface pressure measured by the pressure sensors in real time. In addition, an ice removal device is installed in the hemispherical accommodating cavity 4 to prevent the hemispherical surface from icing and affecting pressure measurement. Under different incoming flow wind speeds, directions and icing meteorological conditions, the ice shapes at the base 1 and the protrusion 2 of this device are different. Therefore, several icing detection ultrasonic sensors can be arranged at these two places. By using the ice thickness at specific points on the surface and the incoming flow conditions (incoming flow wind speed, direction and temperature) as the input and the icing meteorological parameters (LWC, MVD) as the output to establish a mathematical model N2 in advance, the incoming flow meteorological parameters (LWC and MVD) can be calculated in real time according to the local temperature, the ice thickness at specific points measured by the icing ultrasonic sensors in real time, and the incoming flow wind speed and direction obtained by solving N1. The edge of the base 1 protrudes towards the protrusion 2 on one side to form an open cavity to prevent the loss of overflow water and the decline of LWC calculation accuracy. This device can realize the real-time measurement of icing meteorological parameters without building a complex optical path. Therefore, it can solve the problem that it is impossible to build an optical path inside the engine for measurement in the engine icing test in an icing wind tunnel, or it can be installed on the surface of an aircraft to measure the icing meteorological parameters during flight in real time.

[0077] Refer to Figure 2 , this embodiment also provides a method for measuring icing meteorological parameters, including the following steps:

[0078] Step S10: Use the icing meteorological parameter measurement device described in the above embodiment to measure data under multiple wind speeds and directions, and construct a first database of the incoming flow wind speed and direction corresponding to the pressure through this data.

[0079] Specifically, select appropriate pressure measurement point positions and numbers on the icing meteorological parameter measurement device described in the above embodiment, and use the pressure vectors P at each pressure measurement point obtained by numerical calculation or experiment and the corresponding atmospheric data vectors Y. The Kriging algorithm is used to construct a mathematical surrogate model. The atmospheric data vectors Y include wind speed and direction. The mathematical surrogate model takes Y as the input and P as the output, and generates the first library data for the training of the neural network N1.

[0080] Step S20: Train the BP neural network N1 optimized by the particle swarm algorithm according to the first database to obtain a first artificial neural network for calculating the wind speed and direction according to the pressure.

[0081] Specifically, the particle swarm algorithm is used to optimize the initial weights and thresholds of the BP neural network, so that the optimized first artificial neural network can better predict the function output. The elements of optimizing the BP neural network by the particle swarm algorithm include population initialization, fitness function, particle velocity and position update, and individual and population extreme value update operations.

[0082] (1) Population initialization: Randomly generate a cluster, initialize the particles and particle velocities in the cluster. Each particle in the cluster consists of four parts: the connection weights between the input layer and the hidden layer, the hidden layer threshold, the connection weights between the hidden layer and the output layer, and the output layer threshold. The particle contains all the weights and thresholds of the neural network. Given a known network structure, a neural network with a determined structure, weights, and thresholds can be formed.

[0083] (2) Fitness function: According to the individual, obtain the initial weights and thresholds of the BP neural network. After training the BP neural network with the training data to predict the system output, take the absolute sum of the errors between the predicted output and the expected output as the individual fitness value, the maximum value of the errors between the predicted output and the expected output as the individual fitness value, and the sum of the maximum value and the average value of the errors between the predicted output and the expected output as the individual fitness value; Three fitness functions are adopted and the optimization accuracy of the algorithm under different fitness functions is compared.

[0084]

[0085]

[0086]

[0087] In the formula: n is the number of output nodes of the network; is the expected output of the i-th node of the BP neural network; is the predicted output of the i-th node; l is a coefficient.

[0088] (3) Particle velocity and position update: In each iteration process, the particle updates its velocity and position through the individual extreme value and the population extreme value. The update formula:

[0089]

[0090]

[0091] In the formula: is the inertia weight; ; is the current iteration number; is the velocity of the particle; and are non-negative constants; and are distributed in A random number between is the position of the particle, is the individual extreme value, is the global extreme value.

[0092] (4) Update of individual extreme value and population extreme value: When the fitness value of the particle is less than the individual extreme value, the new fitness value replaces the individual extreme value; if the individual extreme value of the particle is less than the population extreme value, the individual extreme value replaces the population extreme value. The update formula is:

[0093]

[0094] In the formula: The fitness value of the new particle.

[0095] The trained BP neural network N1 is used to predict the incoming flow wind speed and direction; after optimizing the BP neural network N1 using the particle swarm algorithm, the BP neural network N1 is used to predict the wind speed and direction, providing conditions for the subsequent prediction of icing meteorological parameters.

[0096] Step S30: Use the icing meteorological parameter measurement device described in the above embodiment to measure the ice shapes at the detection points under various icing conditions, obtain data on the effective ice thickness and icing rate, and establish a second database.

[0097] Specifically, according to the relationship between the mean volume diameter (LWC), liquid water content (MVD) and temperature of the intermittent maximum icing state in Appendix C of the "Airworthiness Standards for Transport Category Aircraft in China [CCAR - 25 - R4]", as well as the incoming flow parameters, different icing conditions are set. The icing thickness L and the corresponding icing parameters and incoming flow parameter vector Y are obtained by numerical calculation or experiment. The icing thickness L is measured by icing measurement at the detection point, and a mathematical surrogate model is constructed using the Kriging algorithm. The icing parameters and incoming flow parameter vector Y include temperature, MVD, LWC, incoming flow velocity, and angle of attack. The mathematical surrogate model takes the temperature, incoming flow wind speed and direction, and icing thickness L in Y as inputs, and MVD and LWC as outputs, generating the second - library data for training the neural network N2.

[0098] Step S40: Train the BP neural network N2 optimized by the dung beetle algorithm according to the second database to obtain a second artificial neural network for calculating the mean volume diameter of water droplets and liquid water content of icing meteorological parameters based on the icing thickness and wind speed and direction.

[0099] Specifically, the input conditions of the BP neural network N2 are the effective icing thickness, environmental temperature, wind speed, and wind direction in the icing database, and the output conditions are the meteorological parameters of the average volume diameter of water droplets and liquid water content. The BP neural network N2 is trained using the dung beetle algorithm; it should be noted that the dung beetle population in the dung beetle algorithm is divided into four parts, namely, the ball-rolling dung beetle, the brood ball, the small dung beetle, and the thief dung beetle. The ball-rolling dung beetle dynamically adjusts its movement direction by simulating the natural environment (such as light, wind direction, etc.) to initially explore the safe foraging area; the brood ball is placed in a known safe area for hatching and finally develops into an adult; the small dung beetle is the adult hatched from the brood ball and concentrates on local search in the current optimal foraging area; the thief dung beetle obtains resources by stealing the discoveries of other dung beetles or directly using the information of the optimal foraging area.

[0100] (1) Ball-rolling dung beetle: The ball-rolling dung beetle uses the sun as a navigation to ensure that the dung ball rolls on a straight path. Natural factors such as light intensity and wind will affect the movement route of the ball-rolling dung beetle. The position update method of the ball-rolling dung beetle is as follows:

[0101] ;

[0102] ;

[0103] In the formula, t represents the current iteration number, represents the position information of the i-th dung beetle at the t-th iteration, is a natural coefficient indicating whether to deviate from the original direction, assigned as -1 or 1 according to the probability method, k ∈ (0, 0.2) represents the deflection coefficient, b ∈ (0, 1) represents a constant, and k and b are set to 0.1 and 0.3 respectively, represents the global worst position, is used to simulate the change in light intensity;

[0104] The formula for updating the position of the ball-rolling dung beetle by dancing is defined as follows:

[0105] ;

[0106] In the formula ∈ [0, π] represents the deflection angle. When is equal to 0, π / 2, or π, the position of the dung beetle will not be updated.

[0107] (2) Brood ball: The brood ball adopts a boundary selection strategy to simulate the egg-laying area of female dung beetles. The definition of the egg-laying area is:

[0108] ;

[0109] ;

[0110] In the formula, Represents the current local best position, and represent the lower and upper limits of the spawning area respectively, where \(R = 1 - t / T\) max , \(T\) max represents the maximum number of iterations, and \(Lb\) and \(Ub\) represent the lower and upper limits of the optimization problem respectively;

[0111] During the iteration process, the position of the brood ball is dynamically changing, and its definition is:

[0112] ;

[0113] In the formula, is the position information of the \(i\)-th brood ball at the \(t\)-th iteration, and \(b1\) and \(b2\) represent two independent random vectors of size \(1\times D\), where \(D\) represents the dimension of the optimization problem.

[0114] (3) Small dung beetle: Establish an optimal foraging area to guide the larvae dung beetles to find food and simulate their foraging behavior, where the optimal foraging area is defined as:

[0115] ;

[0116] ;

[0117] In the formula, represents the current local best position, and represent the lower and upper limits of the best foraging area respectively;

[0118] The position update of the small dung beetle is as follows:

[0119] ;

[0120] In the formula, is the position information of the \(i\)-th small dung beetle at the \(t\)-th iteration, \(C1\) represents a random number following a normal distribution, and \(C2\in(0,1)\) represents a random vector.

[0121] (4) Thief dung beetle: The position update method of the thief dung beetle is as follows:

[0122] ;

[0123] In the formula, \(g\) represents a random vector of size \(1\times D\) following a normal distribution, \(S\) represents a constant value, and \(X\) b is the best position for food competition.

[0124] Finally, after using the trained BP neural network \(N2\), the meteorological parameters of the average volume diameter of water droplets and the liquid water content are predicted.

[0125] This measurement method trains neural network N1 with the pressure data provided at the front end of the measurement device, and combines the predicted wind speed and direction with the ice thickness data at the back end of the device to train neural network N2. The icing parameters are measured in real time through the measurement device. It is known from experimental measurements that the measurement error of MVD is less than 3μm. Thus, it can be seen that this measurement method not only realizes the measurement of oncoming flow parameters and meteorological parameters on the same device, but also has high measurement accuracy.

[0126] The above has described an exemplary flowchart for implementing the icing meteorological parameter measurement method according to the embodiments of the present invention with reference to the accompanying drawings. It should be noted that a large number of details included in the above description are only exemplary descriptions of the present invention, rather than limitations on the present invention. In other embodiments of the present invention, the method may have more, fewer or different steps, and the order, inclusion, function and other relationships between the steps may be different from those described and illustrated.

Claims

1. A method for measuring icing meteorological parameters, characterized in that, It includes the following steps: Step 1: Use an icing meteorological parameter measurement device to measure pressure data under various wind speeds and wind directions, and construct a first database of the incoming flow wind speed and wind direction corresponding to the pressure through this data; Step 2: Train the BP neural network N1 optimized based on the particle swarm algorithm according to the first database to obtain a first artificial neural network for calculating the wind speed and wind direction according to the pressure; Step 3: Use an icing meteorological parameter measurement device to measure the ice shapes on the detection points under various icing conditions, obtain data on the effective ice thickness and icing rate, and establish a second database; Step 4: Train the BP neural network N2 optimized based on the dung beetle algorithm according to the second database to obtain a second artificial neural network for calculating the icing meteorological parameters of the average volume diameter of water droplets and liquid water content according to the icing thickness and wind speed and wind direction.

2. The measurement method according to claim 1, characterized in that The icing meteorological parameter measurement device includes: a disc-shaped base, the edge of the base bulges towards one side to form an open cavity, the center of the cavity is connected to one end of a connecting rod, the other end of the connecting rod is connected with a hemispherical accommodating cavity, multiple icing ultrasonic sensors for measuring the ice layer thickness are arranged inside the base, and a pressure sensor is arranged in the accommodating cavity.

3. The measuring method according to claim 1, characterized in that The said Step 2 includes: The input condition of the BP neural network N1 is the pressure corresponding to different positions in the first database, and the output is the corresponding incoming flow wind speed and wind direction; Use the particle swarm algorithm to train the BP neural network N1; During the training process, use the particle swarm algorithm to optimize the initial threshold and weight matrix of the BP neural network N1 to determine the optimal threshold and weight; Use the trained BP neural network N1 to predict the incoming flow wind speed and wind direction.

4. The measuring method according to claim 3, characterized in that, Using the particle swarm algorithm to train the BP neural network N1 includes: Randomly generate a cluster and initialize the particles and particle velocities in the cluster; According to the individual, obtain the initial weights and thresholds of the BP neural network N1, train the BP neural network N1 with the training data and then predict the system output. Respectively, take the absolute sum of the errors between the predicted output and the expected output as the individual fitness value, the maximum value of the errors between the predicted output and the expected output as the individual fitness value, and the sum of the maximum value and the average value of the errors between the predicted output and the expected output as the individual fitness value; Adopt three fitness functions and compare the optimization accuracy of the algorithm under different fitness functions; ； ； ； Where: n is the number of output nodes of the network; is the expected output of the i-th node of the BP neural network; is the predicted output of the i-th node; l is a coefficient; In each iteration process, the particles update their velocities and positions through the individual extreme value and the global extreme value, and the update formula: ； ； Where: is the inertia weight; ; is the current iteration number; is the velocity of the particle; and are non - negative constants; and are random numbers distributed between ; is the position of the particle, is the personal best, is the global best; when the fitness value of the particle is less than the personal best, the new fitness value replaces the personal best; when the personal best of the particle is less than the global best, the personal best replaces the global best. The update formula is: ； In the formula: Fitness value of the new particle.

5. The measurement method according to claim 2, wherein The said Step 3 includes: According to the relationship between the average volume diameter of water droplets, liquid water content and temperature, and the incoming flow wind speed and wind direction, set multiple icing conditions; Calculate the ice shapes under the corresponding icing conditions according to the icing conditions; According to the differences in the ice shapes, select appropriate measurement point positions and numbers on the icing meteorological parameter measurement device to obtain the effective icing thickness at the measurement points and establish a second database.

6. The measurement method according to claim 5, characterized in that, The second database is established through the following steps: Use numerical calculation or experiment to obtain the icing thickness L and the corresponding icing parameters and incoming flow parameter vector Y. The icing thickness L is measured by the icing detection ultrasonic sensor; The Kriging algorithm is used to construct a mathematical surrogate model. The icing parameters and the incoming flow parameter vector Y include temperature, mean volume diameter of water droplets, liquid water content, and incoming flow wind speed and direction. The mathematical surrogate model takes the temperature, incoming flow velocity, angle of attack, and icing thickness L in Y as inputs, and the mean volume diameter of water droplets and liquid water content as outputs, and generates a second database for training the neural network N2.

7. The measuring method according to claim 2, characterized in that Step 4 includes: The input conditions of the BP neural network N2 are the effective icing thickness, environmental temperature, wind speed, and wind direction in the icing database, and the output conditions are the meteorological parameters of the mean volume diameter of water droplets and liquid water content. The BP neural network N2 is trained using the dung beetle algorithm. During the training process, the dung beetle algorithm is used to optimize the initial threshold and weight matrix of the BP neural network N2 to determine the optimal threshold and weights. After using the trained BP neural network N2, the meteorological parameters of the mean volume diameter of water droplets and liquid water content are predicted.

8. The measuring method according to claim 7, wherein Training the BP neural network N2 using the dung beetle algorithm includes: The position update method of the rolling ball dung beetle is as follows: ； ； where t represents the current iteration number, represents the position information of the i-th dung beetle at the t-th iteration, is a natural coefficient indicating whether to deviate from the original direction, assigned as -1 or 1 according to the probability method, k ∈ (0, 0.2) represents the deflection coefficient, b ∈ (0, 1) represents a constant, and k and b are set to 0.1 and 0.3 respectively, represents the globally worst position, is used to simulate the change in light intensity; The formula for updating the position of the dancing rolling ball dung beetle is defined as follows: ； where ∈[0, π] represents the deflection angle. When is equal to 0, π / 2 or π, the position of the dung beetle will not be updated; The brood ball adopts a boundary selection strategy to simulate the egg-laying area of the female dung beetle. The definition of the egg-laying area is: ； ； In the formula, represents the current local best position, and represent the lower and upper limits of the spawning area respectively, and \(R = 1 - t / T\) max , \(T\) max represents the maximum number of iterations, and \(Lb\) and \(Ub\) represent the lower and upper limits of the optimization problem respectively; During the iteration process, the position of the brood ball is dynamically changing, and its definition is: ； wherein, is the position information of the i-th brood sphere at the t-th iteration, b1 and b2 represent two independent random vectors of size 1×D, and D represents the dimension of the optimization problem; An optimal foraging area is established to guide the larval dung beetles to find food and simulate their foraging behavior, where the optimal foraging area is defined as: ； ； In the formula, represents the current local best position, and represent the lower and upper limits of the best foraging area respectively; The position update of the small dung beetle is as follows: ； In the formula, is the position information of the i-th dung beetle at the t-th iteration, C1 represents a random number following a normal distribution, and C2 ∈ (0, 1) represents a random vector; The position update method of the thief dung beetle is as follows: ； where, g represents a random vector of size 1×D following a normal distribution, S represents a constant value, and X b is the optimal position for food competition.

Images