Multispectral temperature measurement method for aero-engine turbine blades based on HPSOGA

Abstract

The present invention discloses a multispectral temperature measurement method for aircraft engine turbine blades based on HPSOGA, belonging to the technical field of turbine blade temperature measurement. The method comprises the following steps: S1: collecting multispectral radiation data and background reflection radiation data of the turbine blades; S2: initializing the HPSOGA algorithm and randomly generating an initial particle swarm; S3: calculating the environmental complexity factor, updating the dynamic penalty function, and calculating the value of the dynamic penalty function; S4: constructing a multi-objective optimization function; S5: calculating the fitness of all particles to obtain individual optimal solutions and global optimal solutions; S6: using the HPSOGA algorithm to update the speed and position of each particle and generate new particles; iteratively executing steps S3-S6 until conditions are met, and outputting the global optimal solution. The present invention overcomes the problems of low measurement accuracy and poor stability when measuring the temperature of turbine blades. The dynamic penalty function is introduced to improve the adaptability and accuracy of the inversion temperature measurement algorithm. The multi-objective optimization strategy is adopted to improve the reliability and stability of temperature and emissivity calculations.

Description

Technical Field

[0001] The present invention relates to the technical field of turbine blade temperature measurement, and in particular to a multi-spectral temperature measurement method for an aero-engine turbine blade based on HPSOGA. Background Art

[0002] When measuring the temperature of equipment components or specific areas, direct contact temperature measurement methods such as temperature sensors often fail to accurately detect the temperature in all directions due to factors such as excessively high, unstable, and uneven regional temperatures. Installing temperature measurement devices is even more inconvenient when dealing with high-speed turbine blades. Since the intensity of an object's radiation wavelength is closely related to its temperature, non-contact multispectral temperature measurement technology can be used for indirect temperature measurement. Multispectral temperature measurement is a high-precision temperature measurement method that inverts the temperature by collecting radiation data from a target object at multiple wavelengths. It is suitable for measurement applications requiring high precision and complex environments, such as aircraft engine turbine blades.

[0003] In multispectral temperature measurement, the emissivity of the measured object varies with wavelength. Common techniques use a predefined emissivity model to predict this, but this approach cannot accurately predict emissivity changes at high temperatures, and measurement performance varies for some materials. This approach is significantly limited, and the temperature solution can easily fall into a local optimum. A common approach to improving detection performance is to transform the underdetermined equation problem of multispectral temperature measurement into an optimization problem using constrained optimization. A hybrid particle swarm optimization and genetic algorithm (HPSOGA) combines the global search capabilities of a particle swarm algorithm with the local search advantages of a genetic algorithm, and holds great promise for multispectral temperature measurement of turbine blades. However, it still suffers from numerous drawbacks, such as its failure to account for the influence of reflected radiation from adjacent blades, high computational complexity, low efficiency, poor stability in the presence of environmental noise and interference, and the need for further optimization and adjustment when dealing with material emissivity changes due to temperature and surface conditions. These limitations limit its widespread application in complex temperature measurement environments, and new technical solutions are urgently needed to address them. Summary of the Invention

[0004] The present invention overcomes the problems of low measurement accuracy and poor stability when measuring the temperature of turbine blades with existing multispectral temperature measurement technology, and provides a multispectral temperature measurement method for aircraft engine turbine blades based on HPSOGA. Taking into account the background radiation influence of turbine blades, a dynamic penalty function based on environmental changes is introduced to improve the adaptability and accuracy of the inversion temperature measurement algorithm. Taking into account the influence of material emissivity changing with temperature, a multi-objective optimization strategy is adopted to improve the reliability and stability of temperature and emissivity calculations.

[0005] In order to achieve the above object, the present invention adopts the following scheme:

[0006] The multi-spectral temperature measurement method of aero-engine turbine blades based on HPSOGA includes the following steps:

[0007] S1: Use a spectrometer to collect multi-spectral radiation data of turbine blades and background reflection radiation data;

[0008] S2: Initializing the parameters of the HPSOGA algorithm and the parameters of the dynamic penalty function, randomly generating an initial particle swarm, wherein the information of each particle in the initial particle swarm includes the temperature estimation value and the emissivity estimation value of the turbine blade;

[0009] S3: Calculate the environment complexity factor using the background reflected radiation data, and update the penalty intensity coefficient and tolerance threshold of the dynamic penalty function according to the environment complexity factor, calculate the theoretical reflected radiation value corresponding to the temperature estimate value of each particle according to Planck's law, and use the penalty intensity coefficient, tolerance threshold and theoretical reflected radiation value to describe the dynamic penalty function;

[0010] S4: Calculate the error between the estimated temperature value and the estimated emissivity value of each particle, and the error between the preset temperature value and the preset emissivity value, and construct a multi-objective optimization function;

[0011] S5: combining the dynamic penalty function and the multi-objective optimization function to calculate the fitness of all particles, and obtaining the individual optimal solution and the global optimal solution of each particle in the particle swarm according to the fitness;

[0012] S6: Using the HPSOGA algorithm to update the speed and position of each particle in the particle swarm according to the individual optimal solution and the global optimal solution, generating a new particle and adding it to the particle swarm; iteratively executing steps S3-S6 until a preset number of iterations is reached or the change in the fitness of the particle is lower than a preset threshold, and outputting the temperature estimation value and the emissivity estimation value of the particle of the global optimal solution as the temperature estimation value and the emissivity estimation value of the turbine blade.

[0013] Preferably, in step S3, the environment complexity factor EFC(gen) is calculated using the following formula:

[0014]

[0015] in, is the background radiation data R collected in the past M times m The average value of .

[0016] As a preference, the penalty intensity coefficient α(gen) and tolerance threshold θ(gen) of the dynamic penalty function are updated as follows:

[0017]

[0018]

[0019] Among them, α0 is the initial penalty intensity coefficient, θ0 is the initial tolerance threshold, γ is the sensitivity coefficient, γ>0, and δ is the adjustment coefficient, δ>0.

[0020] Preferably, the dynamic penalty function P(T,gen) is calculated as follows:

[0021]

[0022] Among them, R e (T) is the theoretical reflected radiation value corresponding to temperature T obtained by algorithm inversion.

[0023] Preferably, step S4 specifically includes the following steps:

[0024] Calculate the temperature error of the particle , calculate the particle emission rate error value , where T m is the estimated temperature of the current particle, T0 is the preset temperature value, ε m is the estimated emission rate of the current particle, and ε0 is the preset emission rate;

[0025] The average temperature error value is obtained by statistically calculating the temperature error values ​​and emissivity error values ​​calculated for the most recent times. and the average emissivity error ;

[0026] Calculate the dynamic adjustment target weights for temperature and emissivity separately:

[0027]

[0028]

[0029] Among them, ω 1,0 is the initial base weight for temperature, ω 2,0 is the initial weight for emissivity, μ is the adjustment sensitivity parameter, μ>0, and τ is a small amount to avoid the division by zero problem;

[0030] Construct a multi-objective optimization function:

[0031] .

[0032] Preferably, step S4 further comprises defining a synergy factor: , adjust the multi-objective optimization function to:

[0033]

[0034] Where λ≥0 is the synergy factor weight.

[0035] Preferably, in step S5, the fitness of the particle is: , where J(ε m , T m , gen) is a multi-objective optimization function, P(T, gen) is a dynamic penalty function; the fitness of each particle is calculated, and the individual optimal solution for a single particle in the particle swarm and the global optimal solution for all particles are obtained according to the fitness.

[0036] Preferably, step S6 includes the following steps:

[0037] Use the PSO algorithm to update the speed and position of each particle in the particle swarm:

[0038]

[0039]

[0040]

[0041]

[0042] in, is the current velocity corresponding to the particle temperature, is the current velocity corresponding to the emissivity, is the updated velocity corresponding to the particle temperature, is the updated velocity corresponding to the emissivity, is the current position corresponding to the particle temperature, is the current position corresponding to the emissivity, is the updated position corresponding to the particle temperature, is the updated position corresponding to the emission rate; ω is the inertia weight of the PSO algorithm, c1 and c2 are learning factors, r1 and r2 are balance coefficients, and pbest T and pbest ε are the individual optimal solutions for temperature and emissivity, gbest T and gbest ε are the global optimal solutions for temperature and emissivity respectively; T m is the estimated temperature of the particle, ε m is the estimated emission rate of the particle;

[0043] The GA algorithm is used to combine partial solutions through selection, crossover, and mutation to generate new particles, merge the newly generated particles into the particle swarm, and delete the same number of particles as the newly generated particles in order of fitness from low to high to maintain the size of the particle swarm unchanged.

[0044] The present invention has at least the following beneficial effects: (1) a penalty function is added to the optimization process, and an environmental complexity factor is introduced. The penalty parameter is dynamically adjusted according to the degree of fluctuation of the background reflected radiation data, so that the penalty intensity coefficient and the tolerance threshold change adaptively. The dynamic penalty function allows the inversion algorithm to dynamically adjust the calculation parameters according to the changes in environmental factors such as background radiation, so that it can maintain a high-accuracy temperature measurement effect when facing different operating conditions and external interference; (2) the algorithm optimizes the emissivity and temperature at the same time, tracks the average value of the temperature error and the emissivity error in several consecutive iterations, and dynamically adjusts the weights of the temperature and emissivity targets according to the error situation to construct a final multi-objective optimization function, which can not only accurately estimate the temperature, but also consider the influence of the material emissivity changing with temperature, thereby improving the algorithm's adaptability to changes in the physical properties of the material, and thus improving the accuracy and reliability of the overall temperature measurement; (3) by introducing a synergistic factor into the multi-objective optimization function, the problem that the algorithm may tend to unilaterally and quickly reduce the error of a certain target while ignoring the zero-one target, resulting in an unbalanced optimization result, is avoided, so that the temperature and emissivity can be more balanced and synergistically optimized to ensure the stability of the particle mass and the accuracy of the temperature measurement results. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The figure is a schematic diagram of the principle of the multi-spectral temperature measurement method of aero-engine turbine blades based on HPSOGA of the present invention. DETAILED DESCRIPTION

[0046] The present invention will be described in further detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0047] like Figure 1 As shown, the multi-spectral temperature measurement method of an aircraft engine turbine blade based on HPSOGA provided by the present invention includes the following steps:

[0048] S1: Use a spectrometer to collect multispectral radiation data of turbine blades and background reflected radiation data.

[0049] As a key instrument, the spectrometer accurately captures turbine blade radiation information at multiple wavelengths. This radiation data contains important characteristics of turbine blade temperature. Because turbine blades operate in high-temperature environments, they emit electromagnetic radiation with varying energies. The intensity of radiation at different wavelengths is physically correlated with blade temperature. By collecting multispectral radiation data within the specific wavelength bands used for analysis, the spectrometer can obtain real-time spectral information about the turbine blade radiation, providing data support for subsequent temperature inversion calculations. In actual measurement environments, ambient light and other factors can generate reflected radiation from the turbine blade surface. This background reflected radiation can interfere with the analysis of the blade's own radiation data. For example, light from other heat-generating components and lighting sources within the engine compartment can reflect off the turbine blades, and reflected radiation from the blades can also interact. Failure to account for this background reflected radiation can introduce significant errors in subsequent temperature calculations. Accurately capturing background reflected radiation data allows for effective analysis and correction of environmental influences in subsequent algorithm processing, thereby improving the accuracy of temperature measurements.

[0050] S2: Initialize the parameters of the HPSOGA algorithm and the parameters of the dynamic penalty function, and randomly generate an initial particle swarm. The information of each particle in the initial particle swarm includes the temperature estimation value and the emissivity estimation value of the turbine blade.

[0051] Initializing the HPSOGA algorithm's parameters sets the basic conditions for its operation. These parameters, including the particle swarm size, inertia weight, and learning factor, influence the algorithm's search capabilities and convergence speed. For example, the particle swarm size determines the number of particles involved in the optimization search. A large swarm size can increase the computational workload, while a small swarm size may prevent a full search of the solution space. The inertia weight influences the degree to which particles inherit their historical velocities. A reasonable setting can balance the algorithm's global and local search capabilities. The learning factor controls the learning step size of particles towards the individual and global optimal solutions, influencing the algorithm's convergence.

[0052] A dynamic penalty function is used to handle constraints. In this temperature measurement method, the initial setting of its parameters affects the severity of the penalty for solutions that fail to meet the constraints. For example, the initial penalty intensity coefficient and tolerance threshold determine the degree to which particles that fail to meet basic requirements (such as those with large deviations from theoretical reflected radiation values) are screened out early in the algorithm. This guides the algorithm toward a reasonable solution space, thereby increasing solution efficiency and accuracy.

[0053] Each particle in the randomly generated initial particle swarm contains an estimate of the turbine blade's temperature and emissivity. In multispectral thermometry, temperature and emissivity are interrelated and critical parameters that must be solved simultaneously. Although the particles are randomly generated, the initial estimate combination is confined to a reasonable range that encompasses the possible variations in the actual turbine blade temperature and emissivity. This allows the algorithm to begin searching across a broad solution space, increasing the likelihood of finding the global optimal solution. For example, the temperature estimate may be randomly selected within the common operating temperature range of the turbine blade, while the emissivity estimate is set based on the material properties. As the algorithm runs, the particles continuously update their positions and velocities, gradually approaching the actual temperature and emissivity values.

[0054] S3: Use the background reflected radiation data to calculate the environmental complexity factor, and update the penalty intensity coefficient and tolerance threshold of the dynamic penalty function according to the environmental complexity factor, calculate the theoretical reflected radiation value corresponding to the temperature estimate value of each particle according to Planck's law, and calculate the dynamic penalty function using the penalty intensity coefficient, tolerance threshold and theoretical reflected radiation value.

[0055] The core of the dynamic penalty function is to adjust the penalty intensity of the inverted temperature to synchronize with environmental changes. When the measured background reflected radiation data differs significantly from the predicted value, the penalty function increases the penalty for the inverted temperature, prompting the algorithm to avoid selecting temperature values ​​that may increase error due to environmental changes. Leveraging this dynamic penalty function, the improved HPSOGA algorithm of this invention can adaptively adjust its optimization direction in the face of environmental perturbations, strengthening the penalty for unreliable solutions. This allows it to maintain high temperature measurement accuracy under conditions such as high temperature and high background radiation. The penalty intensity coefficient can be adjusted according to actual application requirements, and the tolerance threshold defines the acceptable error range between the background radiation data and the predicted value. The environmental complexity factor (ECF(gen)) measures the complexity and uncertainty of the environmental conditions under the current iteration, gen, and is used to dynamically adjust the penalty intensity coefficient and tolerance threshold. A larger ECF(gen) value indicates a more dramatic change in the background reflected radiation, which has a greater impact on the results. As the amount of reflected radiation increases, the penalty intensity coefficient is increased to strengthen the penalty, while the tolerance threshold is decreased to tighten the tolerance, thereby imposing stricter constraints on mismatched temperature solutions. It is defined by statistically analyzing the fluctuations in background radiation measurements over the most recent generations. Planck's law describes the relationship between the energy distribution of blackbody radiation and temperature. In this method, this law is combined with the particle's temperature estimate to calculate the theoretical reflected radiation value. This theoretical value is used in subsequent algorithms to measure the degree of deviation between the particle's temperature estimate and the actual value.

[0056] S4: Calculate the error between the estimated temperature and the estimated emissivity of each particle, as well as the error between the preset temperature and the preset emissivity, and construct a multi-objective optimization function. This method incorporates multi-objective optimization into the inversion optimization process of the HPSOGA algorithm, optimizing both emissivity and temperature simultaneously. When the error temperature is relatively high, the weight of the temperature objective is increased, while when the error temperature is relatively high, the weight of the emissivity objective is increased. This achieves adaptive adjustment and effectively balances the optimization of these two parameters to ensure the accuracy and reliability of the final temperature measurement results.

[0057] S5: The dynamic penalty function and the multi-objective optimization function are combined to calculate the fitness of all particles, and the individual optimal solution and the global optimal solution of each particle in the particle swarm are obtained according to the fitness.

[0058] Combining the dynamic penalty function with the multi-objective optimization function to calculate fitness comprehensively evaluates each particle's performance in satisfying environmental constraints and approximating the true temperature and emissivity. For example, if a particle's temperature and emissivity estimates result in a small dynamic penalty function and a small error term in the multi-objective optimization function, its fitness is relatively high, indicating that the particle is more likely to be close to the true solution. If the current particle's fitness is better than that of its previous best position, the particle's position is updated to the individual best solution. Simultaneously, the particle with the highest fitness is searched within the entire particle swarm and designated as the global best solution. These best solutions serve as important references for subsequent particle updates. For example, in the PSO algorithm's formula for updating particle velocity and position, the individual best solution (pbest) and the global best solution (gbest) guide particles toward more optimal directions, allowing the particle swarm to continuously approach the true temperature and emissivity values ​​during the iterative process, thereby improving temperature measurement accuracy. By repeatedly repeating this process, the algorithm can gradually select the temperature and emissivity estimates that best reflect the actual situation, enabling accurate temperature measurement of aircraft engine turbine blades.

[0059] S6: Using the HPSOGA algorithm to update the speed and position of each particle in the particle swarm according to the individual optimal solution and the global optimal solution, generating a new particle and adding it to the particle swarm; iteratively executing steps S3-S6 until a preset number of iterations is reached or the change in the fitness of the particle is lower than a preset threshold, and outputting the temperature estimation value and the emissivity estimation value of the particle of the global optimal solution as the temperature estimation value and the emissivity estimation value of the turbine blade.

[0060] When using the HPSOGA algorithm, the particle swarm optimization (PSO) component enables particles to adjust their positions in the solution space based on their historical speed, individual optimal solutions, and information about the global optimal solution, searching in a more optimal direction and accelerating convergence to the optimal solution. Simultaneously, a genetic algorithm (GA) combines partial solutions to generate new particles through selection, crossover, and mutation. The selection operation selects outstanding individuals based on their fitness, increasing the probability that particles with high fitness will be selected for subsequent operations and ensuring the inheritance of excellent genes. The crossover operation exchanges and combines partial information of selected individuals to increase the diversity of solutions. The mutation operation randomly changes certain genes of particles with a certain probability to prevent the algorithm from falling into a local optimum.

[0061] A test example for the above method:

[0062] Prepare the temperature measurement environment, install thermocouples at the corresponding locations on the turbine blades, and heat the turbine blades to 1100K. Use a spectrometer to collect multispectral radiation data from the turbine blades in the 1100nm-1600nm band. Use the spectrometer to obtain background reflected radiation data, and then perform multiple measurements to obtain a record of background radiation data for the past M times. Filter and smooth the spectral radiation data to minimize the impact of random noise on the temperature inversion. Calculate the blackbody radiation intensity at any temperature T using Planck's law.

[0063] Initialize the HPSOGA algorithm parameters: set the particle swarm size N, PSO inertia weight ω, learning factors c1 and c2, crossover rate and mutation rate of the genetic algorithm, and the maximum number of iterations Gmax.

[0064] Initialize the dynamic penalty function parameters: set the basic penalty intensity α0 and basic threshold θ0, sensitivity coefficients γ, δ.

[0065] Initialize multi-objective optimization parameters: set initial weight ω 1，0 ,ω 2，0 As well as the sensitivity parameter μ, the synergy factor weight λ, and the small amount τ.

[0066] An initial particle swarm is randomly generated, each particle contains a temperature estimate and an emissivity estimate, and their initial values ​​are uniformly distributed within a reasonable range.

[0067] The environmental complexity factor (ECF) is calculated, and the penalty function coefficient is updated based on ECF(gen). The theoretical reflected radiation value is calculated for the current particle's temperature. A dynamic penalty function is calculated, and the particle's temperature error and emissivity error are calculated. The average value of the error statistics for the most recent generations is calculated, and the objective weights are dynamically adjusted based on the average value to construct a multi-objective optimization function. The multi-objective optimization function is combined with the dynamic penalty function to obtain the particle's fitness. The fitness is calculated for all particles in the swarm, and the individual optimal solution pbest and the global optimal solution gbest are determined based on the fitness. Each particle has its own individual optimal solution. Initially, each particle's current position is set to its individual optimal position, and the corresponding fitness value is set to the individual optimal fitness value. In each iteration, the fitness value of the particle's current position is calculated. If the fitness value of the current position is better than the previously recorded individual optimal fitness value, the individual optimal position and individual optimal fitness value are updated. The global optimal solution is the best solution among the individual optimal solutions of all particles in the entire particle swarm. Initially, the individual optimal fitness values ​​of all particles are compared, and the best one is selected as the global optimal solution. During each iteration, the individual optimal fitness values ​​of all particles are also compared. If the individual optimal fitness value of a particle is better than the current global optimal fitness value, the global optimal solution is updated. The PSO formula is used to update the speed and position of the particles. The genetic algorithm operation is applied to the population every several generations: partial solutions are combined through selection, crossover, and mutation to generate new particles, which are then merged with the population updated by PSO. The merged solution set is sorted from best to worst based on fitness, and the top N particles with better fitness are retained as the new generation of particle swarms. Repeat the iteration, and update the ECF(gen), dynamic penalty coefficient, and dynamic weight after each generation. The iteration stops when the maximum number of iterations Gmax is reached or the fitness change is lower than the preset threshold. The temperature estimate and emissivity estimate of the gbest particle at the end of the iteration are output as the final inversion result, that is, the true blade temperature and corresponding emissivity estimated by the inversion algorithm.

[0068] This method adds a dynamic penalty function to the optimization process and introduces an environmental complexity factor. The penalty parameters are dynamically adjusted according to the fluctuation degree of the background reflected radiation data, so that the penalty intensity coefficient and the tolerance threshold change adaptively. The dynamic penalty function allows the inversion algorithm to dynamically adjust the calculation parameters according to the changes in environmental factors such as background radiation, so that it can maintain a high-accuracy temperature measurement effect in the face of different operating conditions and external interference; the algorithm simultaneously optimizes the emissivity and temperature, tracks the average value of the temperature error and emissivity error in several consecutive iterations, and dynamically adjusts the weights of the temperature and emissivity targets according to the error situation to construct the final multi-objective optimization function. It can not only accurately estimate the temperature, but also consider the influence of the material emissivity changing with temperature, thereby improving the algorithm's adaptability to changes in the physical properties of the material, thereby improving the accuracy and reliability of the overall temperature measurement.

[0069] In another technical solution, in step S3, the environment complexity factor EFC(gen) is calculated using the following formula:

[0070]

[0071] in, is the background radiation data R collected in the past M times m The average value of .

[0072] The way to update the penalty intensity coefficient α(gen) and tolerance threshold θ(gen) of the dynamic penalty function is:

[0073]

[0074]

[0075] Among them, α0 is the initial penalty intensity coefficient, θ0 is the initial tolerance threshold, γ is the sensitivity coefficient, γ>0, and δ is the adjustment coefficient, δ>0.

[0076] The dynamic penalty function P(T,gen) is calculated as follows:

[0077]

[0078] Among them, R e (T) is the theoretical reflected radiation value corresponding to the temperature T obtained by the algorithm inversion. T represents the temperature obtained by inversion, R m It is the background radiation data measured and actually calculated from the spectrometer.

[0079] In another technical solution, step S4 specifically includes the following steps:

[0080] Calculate the temperature error of the particle , calculate the particle emissivity error value , where T m is the estimated temperature of the current particle, T0 is the preset temperature value, ε m is the estimated emission rate of the current particle, and ε0 is the preset emission rate;

[0081] The average temperature error value is obtained by statistically calculating the temperature error values ​​and emissivity error values ​​calculated for the most recent times. and the average emissivity error :

[0082]

[0083]

[0084] M is the evaluation window length of the most recent generations.

[0085] Calculate the dynamic adjustment target weights for temperature and emissivity separately:

[0086]

[0087]

[0088] Among them, ω 1,0 is the initial base weight for temperature, ω 2,0 is the initial weight for emissivity, μ is the sensitivity adjustment parameter, μ>0, τ is a small amount to avoid the zero division problem; when When it is relatively large, will increase accordingly, causing the algorithm to pay more attention to the accuracy of temperature; when When it is relatively large, Increased to emphasize the optimization of emissivity.

[0089] Construct a multi-objective optimization function:

[0090] .

[0091] Step S4 also includes defining the synergy factor: , adjust the multi-objective optimization function to:

[0092]

[0093] Where λ ≥ 0 is the synergy factor weight. When the temperature error and emissivity error differ significantly and are unbalanced, S(gen) increases to reduce this imbalance. A synergy term is added to the optimization objective and weighted with an adjustable coefficient λ to promote simultaneous improvement in temperature and emissivity. This avoids the problem of the algorithm tending to unilaterally and rapidly reduce the error of a single objective while ignoring the zero-one objective, leading to unbalanced optimization results. This allows for a more balanced synergistic optimization of temperature and emissivity, thereby ensuring the stability of particle mass and the accuracy of temperature measurement results.

[0094] In step S5, the fitness of the particle is: , where J(ε m , T m , gen) is a multi-objective optimization function, P(T, gen) is a dynamic penalty function; the fitness of each particle is calculated, and the individual optimal solution for a single particle in the particle swarm and the global optimal solution for all particles are obtained according to the fitness.

[0095] In another technical solution, step S6 includes the following steps:

[0096] Use the PSO algorithm to update the speed and position of each particle in the particle swarm:

[0097]

[0098]

[0099]

[0100]

[0101] in, is the current velocity corresponding to the particle temperature, is the current velocity corresponding to the emissivity, is the updated velocity corresponding to the particle temperature, is the updated velocity corresponding to the emissivity, The current position of the corresponding particle temperature, is the current position corresponding to the emissivity, is the updated position corresponding to the particle temperature, is the updated position corresponding to the emission rate; ω is the inertia weight of the PSO algorithm, c1 and c2 are learning factors, r1 and r2 are balance coefficients, and pbest T and pbest ε are the individual optimal solutions for temperature and emissivity, gbest T and gbest ε are the global optimal solutions for temperature and emissivity respectively; T m is the estimated temperature of the particle, ε m is the estimated emission rate of the particle;

[0102] The GA algorithm combines partial solutions through selection, crossover, and mutation to generate new particles. Proportional selection is used when selecting partial solutions. The probability of selection is determined by the particle's fitness. Particles with higher fitness have a greater probability of selection. The probabilities of all particles form a probability wheel. The wheel is rotated Z times, and the Z particles pointed to by the pointer are selected as partial solutions to participate in the crossover and mutation operations of the GA algorithm. The size of Z is determined by the size of the particle swarm. For example, if the particle swarm size is 100, the recommended value of Z is between 3 and 7. Newly generated particles are merged into the particle swarm, and particles equal to the number of newly generated particles are deleted in descending fitness order to maintain the particle swarm size. This ensures the diversity of the particle swarm while preventing the population from expanding indefinitely, enabling the algorithm to continuously and efficiently search for the optimal solution with limited resources.

[0103] During the optimization process, the HPSOGA algorithm leverages the strengths of both the PSO and GA algorithms to achieve both global search and local improvement. The PSO rapidly searches the global solution space through velocity and position updates, while the GA's selection, crossover, and mutation operations increase population diversity and avoid local optima. The introduction of a dynamic penalty function and multi-objective optimization strategy throughout the iterative process enables the algorithm to adaptively adjust its optimization direction and focus based on real-time environmental conditions and optimization progress. As environmental complexity increases, the dynamic penalty function strengthens the constraints on suboptimal solutions, ensuring that the algorithm can still obtain reliable inversion results under high-perturbation conditions. When a significant imbalance in improvement occurs between temperature and emissivity, the dynamic multi-objective optimization strategy uses dynamic weights and synergy factors to guide the algorithm to simultaneously improve the accuracy of both, ensuring that the final temperature measurement results are both accurate and stable. This improved optimization strategy enables the algorithm to ensure high accuracy in temperature estimation while maintaining the stability of emissivity estimation, effectively improving the robustness and reliability of the entire temperature measurement process.

[0104] It should be noted that although the steps are described above in a specific order, this does not necessarily mean that the steps must be performed in this specific order. In fact, some of these steps can be performed concurrently or even in a different order, as long as the required functions can be achieved. The number of devices and processing scales described here are intended to simplify the description of the present invention. Applications, modifications, and variations of the present invention will be apparent to those skilled in the art.

[0105] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. The multi-spectral temperature measurement method of aircraft engine turbine blades based on HPSOGA is characterized by: The following steps are involved: S1: Use a spectrometer to collect multi-spectral radiation data of turbine blades and background reflection radiation data; S2: Initializing the parameters of the HPSOGA algorithm and the parameters of the dynamic penalty function, randomly generating an initial particle swarm, wherein the information of each particle in the initial particle swarm includes the temperature estimation value and the emissivity estimation value of the turbine blade; S3: Calculate the environment complexity factor using the background reflected radiation data, and update the penalty intensity coefficient and tolerance threshold of the dynamic penalty function according to the environment complexity factor, calculate the theoretical reflected radiation value corresponding to the temperature estimate value of each particle according to Planck's law, and use the penalty intensity coefficient, tolerance threshold and theoretical reflected radiation value to describe the dynamic penalty function; S4: Calculate the error between the estimated temperature value and the estimated emissivity value of each particle, and the error between the preset temperature value and the preset emissivity value, and construct a multi-objective optimization function; S5: combining the dynamic penalty function and the multi-objective optimization function to calculate the fitness of all particles, and obtaining the individual optimal solution and the global optimal solution of each particle in the particle swarm according to the fitness; S6: using the HPSOGA algorithm to update the speed and position of each particle in the particle swarm according to the individual optimal solution and the global optimal solution, generating a new particle and adding it to the particle swarm; iteratively executing steps S3-S6 until a preset number of iterations is reached or the change in the fitness of the particle is lower than a preset threshold, and outputting the temperature estimation value and the emissivity estimation value of the particle of the global optimal solution as the temperature estimation value and the emissivity estimation value of the turbine blade; The step S6 includes the following steps: Use the PSO algorithm to update the speed and position of each particle in the particle swarm: in, is the current velocity corresponding to the particle temperature, is the current velocity corresponding to the emissivity, is the updated velocity corresponding to the particle temperature, is the updated velocity corresponding to the emissivity, is the current position corresponding to the particle temperature, is the current position corresponding to the emissivity, is the updated position corresponding to the particle temperature, is the updated position corresponding to the emission rate; ω is the inertia weight of the PSO algorithm, c1 and c2 are learning factors, r1 and r2 are balance coefficients, and pbest T and pbest ε are the individual optimal solutions for temperature and emissivity, gbest T and gbest ε are the global optimal solutions for temperature and emissivity respectively; T m is the estimated temperature of the particle, ε m is the estimated emission rate of the particle; The GA algorithm is used to combine partial solutions through selection, crossover, and mutation to generate new particles, merge the newly generated particles into the particle swarm, and delete the same number of particles as the newly generated particles in order of fitness from low to high to maintain the size of the particle swarm unchanged.

2. The multi-spectral temperature measurement method for aircraft engine turbine blades based on HPSOGA according to claim 1, characterized in that: In step S3, the environment complexity factor EFC(gen) is calculated using the following formula: in, is the background radiation data R collected in the past M times m The average value of .

3. The multi-spectral temperature measurement method for aircraft engine turbine blades based on HPSOGA according to claim 2, characterized in that: The way to update the penalty intensity coefficient α(gen) and tolerance threshold θ(gen) of the dynamic penalty function is: Among them, α0 is the initial penalty intensity coefficient, θ0 is the initial tolerance threshold, γ is the sensitivity coefficient, γ>0, and δ is the adjustment coefficient, δ>0.

4. The multi-spectral temperature measurement method for aircraft engine turbine blades based on HPSOGA according to claim 3, characterized in that: The method for updating the parameters of the dynamic penalty function P(T,gen) is: Among them, R e (T) is the theoretical reflected radiation value corresponding to temperature T obtained by algorithm inversion.

5. The multi-spectral temperature measurement method for aircraft engine turbine blades based on HPSOGA according to claim 1, characterized in that: Step S4 specifically includes the following steps: Calculate the temperature error of the particle , calculate the particle emission rate error value , where T m is the estimated temperature of the current particle, T0 is the preset temperature value, ε m is the estimated emission rate of the current particle, and ε0 is the preset emission rate; The average temperature error value is obtained by statistically calculating the temperature error values ​​and emissivity error values ​​calculated for the most recent times. and the average emissivity error ; Calculate the dynamic adjustment target weights for temperature and emissivity separately: Among them, ω 1,0 is the initial base weight for temperature, ω 2,0 is the initial weight for emissivity, μ is the adjustment sensitivity parameter, μ>0, and τ is a small amount to avoid the division by zero problem; Construct a multi-objective optimization function: 。 6. The multi-spectral temperature measurement method for aircraft engine turbine blades based on HPSOGA according to claim 5, characterized in that: Step S4 also includes defining the synergy factor: , adjust the multi-objective optimization function to: Where λ≥0 is the synergy factor weight.

7. The multi-spectral temperature measurement method for aircraft engine turbine blades based on HPSOGA according to claim 6, characterized in that: In step S5, the fitness of the particle is: , where J(ε m , T m , gen) is a multi-objective optimization function, P(T, gen) is a dynamic penalty function; the fitness of each particle is calculated, and the individual optimal solution for a single particle in the particle swarm and the global optimal solution for all particles are obtained according to the fitness.

Images