Compost material airflow detection method and system

Abstract

The present invention discloses a compost material airflow detection method and system, comprising: randomly generating different airflow state distribution data for the compost material area according to airflow patterns; using the generated airflow state distribution data as input to a finite element simulation model to simulate the airflow state distribution data; obtaining observation values through sensors, wherein the observation values include gas concentration data and temperature data for the compost material area, and comparing and evaluating the observation values with the simulation results; under the guidance of the evaluation results, regenerating new airflow state distribution data through an optimization algorithm, and simulating, comparing, and evaluating the regenerated airflow state distribution data, repeating the above process until the error between the observation value and the simulation result is less than a threshold value, thereby obtaining the optimal airflow state distribution data that conforms to the current area. Through the above technical solution, the airflow of the compost material can be effectively detected.

Description

Technical Field

[0001] The present invention belongs to the technical field of airflow detection, and in particular relates to a compost material airflow detection method and system. Background Art

[0002] Monitoring compost airflow is a critical step in the composting process, directly impacting its efficiency, quality, and environmental impact. However, in practice, this process presents a number of challenges that not only hinder the smooth progress of composting but can also negatively impact the quality of the resulting compost.

[0003] First of all, the composting area cannot be completely sealed due to the construction materials or processes, which may cause the influence of other airflows besides the vents. At the same time, the airflow may produce turbulence and other phenomena. Simulation models are generally simulated in a fixed environment or by applying ideal external factors. However, the fixed environment or the application of ideal external factors under simulation cannot effectively simulate the actual influencing factors and airflow conditions, resulting in inaccurate simulation results.

[0004] In the composting material airflow monitoring environment, the airflow sensor cannot work effectively due to the influence of the high temperature or pH value of the environment, so direct measurement cannot be performed based on the airflow sensor.

[0005] Furthermore, the spatial distribution of gas emissions during composting is also non-uniform, with significant variations in gas concentrations and flow rates at different locations. If sampling and testing are performed only at individual locations, using direct mean or linear calculations will only reflect the overall airflow within the area, and not every location within the entire area. Consequently, the resulting monitoring data may not represent the true condition of the entire compost. Summary of the Invention

[0006] In order to solve the above technical problems, the present invention proposes a compost material airflow detection method and system to solve the problems existing in the above prior art.

[0007] To achieve the above objectives, the present invention provides a compost material airflow detection method and system, comprising:

[0008] Randomly generate different airflow state distribution data in the compost material area according to the airflow law;

[0009] The airflow state distribution data is simulated by a finite element simulation model, wherein the simulation results include simulated gas concentration data and temperature data;

[0010] Obtaining observation values, including gas concentration data and temperature data in the composting material area, and comparing and evaluating simulation results based on the observation values;

[0011] According to the evaluation results, the airflow state distribution data is iteratively adjusted through the optimization algorithm, and simulation and evaluation are performed based on the iteratively adjusted airflow state distribution data until the error between the observed value and the simulation result in the iterative adjustment is less than the threshold value, thereby obtaining the optimal airflow state distribution data.

[0012] Optionally, the airflow state distribution data is generated by a generator in a generative adversarial network, wherein the generator is used to simulate the law of the airflow state distribution data, wherein the generative adversarial network is trained by training samples, wherein the training samples include airflow data measured in the experiment and airflow data generated by simulation.

[0013] Optionally, the generative adversarial network includes a generator and a discriminator, wherein the generator is used to generate airflow state distribution data based on latent variables and auxiliary variables, and different airflow state distribution data are obtained by adjusting the latent variables; the discriminator is used to score the airflow state distribution data based on training samples and auxiliary variables; the generator is optimized according to the score of the discriminator; and during the optimization process, the training samples are used to optimize the generator.

[0014] Optionally, auxiliary variables include gas concentration data and temperature data.

[0015] Optionally, in a finite element simulation model, a simulation model is constructed for the compost material area and the compost material, and the compost material area is used as a boundary constraint and the compost material is used as an emission source, and simulation calculations are performed using a fluid mechanics calculation model and a thermodynamics calculation model;

[0016] The fluid mechanics calculation model and the thermodynamics calculation model are modified according to the collected airflow state distribution data and temperature and gas concentration data.

[0017] Optionally, the airflow state distribution data is iteratively adjusted according to the evaluation results through an optimization algorithm, wherein, in a single adjustment of the airflow state distribution data, the latent variable is taken as an individual, and several individuals are taken as a population, wherein the airflow state distribution data is generated according to the latent variable; the individuals are adjusted according to the evaluation results, wherein the latent variables corresponding to the minimum error between several observation values and simulation results in the evaluation results are taken as local optimal solutions, the population is updated according to the local optimal solution combined with the inertia weight, and after the update, the updated population is updated again based on a random step size, and the corresponding airflow state distribution data is regenerated according to the individuals in the updated population.

[0018] Optionally, the goal of the optimization algorithm is to minimize the weighted sum of the differences between the gas concentration data and temperature data of the composting material area and the simulation results.

[0019] On the other hand, the present invention provides a compost material airflow monitoring system for use in the above method.

[0020] Compared with the prior art, the present invention has the following advantages and technical effects:

[0021] Through the above technical solution, the present invention does not need to obtain the overall regional airflow data through the mean of each environmental position or direct linear calculation, but calculates the airflow distribution data of each position in the entire area through preferentially arranged sensors. Since different observation positions are different, the data of the observation position is affected by multiple data in its airflow distribution. At the same time, the data at different positions of the airflow distribution also have a certain correlation or distribution. Therefore, through limited observation points, most of the airflow state data in the area can be inverted. The airflow in the area affected by the outside world may change, which in turn leads to changes in the observed values. The airflow data in the area is directly generated, and the generated airflow is simulated according to the observation. Value evaluation can directly analyze the airflow data based on the correlation between the airflow and the observed value, without considering the external influence, so as to reduce the external influence. The present invention pre-generates the airflow state distribution data by generating a network, first of all, it satisfies the mutual correlation or certain distribution between the airflow state distribution data, and simulates the airflow state distribution data to generate corresponding environmental data on this basis, and evaluates the simulated environmental data according to the observed environmental data to meet the correlation of the airflow state distribution data with the environmental data. Through the above two correlation constraints, the airflow state distribution data that conforms to the real scene is generated, and the airflow state distribution data contains the airflow size and direction at different positions to solve the above-mentioned ending problem. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0023] Figure 1 Schematic diagram of a method flow in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0025] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0026] like Figure 1 As shown, this embodiment provides a compost material airflow detection method and system, including:

[0027] Generate different airflow state distribution data in the composting material area, build a simulation model, and simulate the airflow state distribution data through the finite element simulation model;

[0028] Acquiring gas concentration data and temperature data of a composting material area, and evaluating simulation results of the airflow state distribution data based on the gas concentration data and temperature data;

[0029] The airflow state distribution data is re-acquired according to the evaluation results, and the simulation and evaluation are re-performed according to the re-acquired airflow state distribution data until a certain accuracy or number of cycles is reached, the simulation is terminated, and the optimal airflow state distribution data is output.

[0030] As some embodiments, the gas concentration data and temperature data are for the gas concentration data and temperature data at different locations in the composting material area, wherein the gas concentration data include CO2 gas concentration data and ammonia gas concentration data, and the gas concentration data and temperature data are collected by corresponding sensors, wherein there is no specific restriction on the layout position of the sensors. In the layout of the sensors, the distance between the two sensors is greater than a certain threshold to ensure monitoring of different areas. At the same time, the sensors are responsible for data monitoring at different locations. In the process of data monitoring, the data or environmental data collected by the above-mentioned sensors are used as observation values, and the airflow state distribution data is continuously adjusted to simulate the airflow. After the simulation is completed, the environmental data in the simulation is compared with the observation data, i.e., evaluated, and the evaluation result is used as the basis for judging the airflow state distribution data.

[0031] As some embodiments, different airflow state distribution data are generated through a generation method, and the generated airflow state distribution data is only a simulated airflow state distribution data, rather than data directly converted through other concentration or temperature data. For the simulated airflow state distribution data, it is in a matrix form, which serves as the applied data in different grids in the subsequent finite element simulation model, wherein each element in the matrix corresponds to a grid according to the grid arrangement of the finite element simulation model, and in the grid, the numerical value in the element represents its applied property, wherein the airflow state distribution data includes airflow size distribution data and airflow direction distribution data, both of which are represented by different numerical values to represent their size or direction.

[0032] In the generation method, the technical solution of the present invention adopts a generator to generate the above-mentioned airflow state distribution data, wherein the generator adopts a generative adversarial network. The generative adversarial network includes a generator and an adversary, wherein the adversary is used for training the generator.

[0033] Specifically, the input data of the generator is a latent variable, where the latent variable serves as the independent variable data related to the airflow state distribution data. By continuously adjusting the form of the input latent variable, different airflow state distribution data can be obtained, where the output data of the generator is the airflow state distribution data.

[0034] As some embodiments, the input data of the generator includes, in addition to potential variables, auxiliary variables, wherein the auxiliary variables are environmental data of the surrounding range and central position of the monitoring area, such as environmental data collected by sensors arranged at fixed intervals around the compost monitoring area and placed at the middle position, wherein the environmental data includes gas concentration data and temperature data, and the gas concentration data and temperature data are photographed and arranged at the positions to generate a gas concentration and temperature data matrix, wherein the matrix elements correspond to the actual values of the gas concentration data and the temperature data. There is no strict restriction on the above position for a monitoring area, and its initial fixed position and subsequent fixed position as an auxiliary variable can remain unchanged.

[0035] For generative adversarial networks, the generative adversarial network mainly uses the generator to generate different gas concentration distribution data based on latent variables and auxiliary variable data. Before use, the generator needs to be trained, and the discriminator is needed to evaluate the data generated by the generator to determine whether the training is completed.

[0036] As a specific implementation method, the generator: the generator network accepts auxiliary variables and latent variables as input, and generates airflow state distribution data that conforms to the corresponding auxiliary variables and latent variables. The generator adopts a structure of several fully connected layers, several deconvolution layers connected in sequence, and a Bessel output layer to ensure that the generated gas state distribution data has reasonable smooth data between grids. The input of its generator: auxiliary parameters include different environmental parameters, expressed in matrix numerical form; latent variables are used to control different airflow condition distribution data, and different latent variables and auxiliary parameters in the latent space correspond to different airflow condition distribution data samples. Output: The generator generates airflow condition distribution data, including airflow size distribution data and airflow direction distribution data.

[0037] Discriminator: The discriminator network D is responsible for judging whether the airflow condition distribution data generated by the generator conforms to the actual airflow condition distribution data distribution. The discriminator uses a combination structure of several convolutional layers and several fully connected layers connected in sequence to extract the airflow condition distribution data layer by layer and output a score. The score is used to judge whether the generated airflow condition distribution data conforms to the normal airflow condition distribution. Input: The airflow condition distribution data includes the airflow condition distribution data generated by the generator and the actual airflow condition distribution data. The discriminator also receives auxiliary parameters as input to ensure that the airflow condition distribution data corresponds to the monitored environment as much as possible. Output: The discriminator outputs a score to indicate whether the input airflow condition distribution data comes from the actual data distribution.

[0038] As a specific embodiment, during the training process of the generative adversarial network, the training goal is to enable the CGAN to generate diverse airflow condition distribution data based on the input auxiliary variables and latent variables through adversarial training of the generator and the discriminator, thereby providing a basis for subsequent simulation.

[0039] The training samples are obtained through relevant experiments or simulations. The relevant experiments include generating a stable airflow in the area through relevant airflow generating equipment, and collecting actual airflow data of the entire area through deployed sensors or mobile sensors; simulating the airflow data in a certain space through simulation methods, and after the simulation is completed, the simulated airflow data in the simulated space is counted. After collecting actual airflow data and simulated airflow data multiple times, the generator is first trained with the simulated airflow data to make initial adjustments to the parameters of the generation network, and after training with the simulated airflow data, it is trained with the actual airflow data. Based on the initial adjustment, fine-tuning is performed to enable the generator to learn the laws of the airflow data itself. The training is mainly to fit the correlation or distribution between the corresponding airflow states, such as the continuity of the airflow values at adjacent positions and the fact that the direction of the intermediate airflow cannot produce sudden changes, such as the direct collision of the airflow directions between adjacent grids. In other experimental scenarios, the airflow laws are the same, and different airflow phenomena, such as turbulence, also conform to the above laws. In order to better conform to the above phenomena, different airflow phenomena can be simulated in the actual airflow data. Therefore, samples obtained in other scenarios can also be used as training samples for the generator.

[0040] During the training process, the latent variables and the corresponding auxiliary variables are input into the generator to generate the corresponding simulated airflow state distribution data. The simulated airflow state distribution data and the corresponding auxiliary variables are input into the discriminator, and the training samples and the corresponding auxiliary variables are input into the discriminator. The discriminator will identify the probability of whether the simulated airflow state distribution data belongs to the real training sample according to the distribution of the training samples, and on this basis, guide the adjustment of the weight parameters in the generator to optimize the generator with the training goal as the goal.

[0041] The training objectives are:

[0042]

[0043] in, G(z|t) It represents the airflow condition distribution data generated when random noise that obeys the real distribution is input under the auxiliary parameter t condition. D(x|t) It represents the probability that the discriminator identifies the sample that follows the distribution under the auxiliary parameter t as the real airflow condition distribution data, D(G(x|t)|t) represents the probability that the airflow condition distribution data generated by the discriminator identification generator is the real airflow condition distribution data, E Represents mathematical expectation, G represents the generator, D represents the discriminator, z represents the latent variable, x represents the real sample, and L represents the loss value.

[0044] In some embodiments, after generating relevant airflow distribution data using a generator using potential variables and auxiliary variables, simulation is performed in finite element simulation software using the airflow distribution data. It should be noted that the airflow distribution data generated by the above process is only the airflow data in the area, or the airflow data in the area after the influence of external factors has been completed. At this time, it is not necessary to consider the impact of external factors on the airflow; only the airflow distribution data in the area is considered. Among them, by using Abaqus or Fluent analysis models, a simulation model consisting of the compost material monitoring area and the compost material is constructed in the pre-processing interface, wherein relevant attribute parameters are set for the compost material, and the compost material is set as an emission source to emit gas in real time, wherein the emission rate and temperature of the gas are obtained based on theoretical calculations of the material, temperature, and area of the material itself. At the back end, the back end grid data is calculated and adjusted using a fluid diffusion calculation model and a thermodynamic model, the compost material detection area is gridded, and the gridded area is subjected to airflow factors based on different airflow distribution data to simulate and obtain gas concentration data and temperature data at different locations in the compost material monitoring area.

[0045] In some embodiments, the latent variables are continuously adjusted. During this adjustment process, the latent variables are optimized using relevant optimization methods to determine the error between observed and simulated values at different locations within the compost material monitoring area. By minimizing this error, the airflow state distribution data obtained is guaranteed to conform to the current conditions within the compost material monitoring area. During the optimization process, the primary optimization goal is to minimize the difference between the environmental data collected by the sensor, including gas concentration data and temperature data, and the gas concentration and temperature data at the corresponding locations simulated by the airflow state distribution data, and this is used as the basis for optimization.

[0046] As some embodiments, during the detection of airflow state distribution data, the airflow state distribution data is searched by combining an optimization algorithm with environmental data. In this embodiment, the optimization method used includes:

[0047] Different latent variables are used as the initial population, which contains different individuals, and each individual contains latent variables. The goal is to minimize the weighted sum of the squares of the differences between the gas concentration data, temperature data, and airflow state distribution data at the corresponding locations. The objective function is:

[0048] f(x)=min∑a(A nm -A nm ') 2 +b(B m -B m ')2

[0049] Among them, A nm A represents the gas observation value of the nth type and mth position transmitted by the sensor, nm ' represents the simulated value of the gas at the mth position of the nth type; B nm Indicates the temperature observation value of the nth type and mth position transmitted by the sensor, B nm ' represents the simulated temperature value at the mth location of the nth species. Constraints include the boundary conditions of the location. Set the population update method and the iteration stop condition (meeting the maximum number of iterations or the objective function is less than a certain threshold). After setting the above, adjust the airflow state distribution data.

[0050] As some embodiments, the specific steps of the optimization algorithm are:

[0051] 1) Initialize the population, where different latent variables are used as the initial population. The population contains different individuals, and the individuals contain latent variables. The positions selected based on manual experience or random selection are used as initialization individuals, and the emissions are randomly generated using random numbers.

[0052] Set the relevant objective function, constraints and corresponding fitness, where the objective function f is:

[0053] f=min∑a(A nm -A nm ') 2 +b(B m -B m ') 2

[0054] A nm A represents the gas observation value of the nth type and mth position transmitted by the sensor, nm ' represents the simulated value of the gas at the mth position of the nth type; B nm Indicates the temperature observation value of the nth type and mth position transmitted by the sensor, B nm ' represents the simulated temperature value at the mth position of the nth type. Constraints include ensuring that the gas concentration, temperature, and airflow conditions are within the set range. The fitness function is the inverse of the objective function.

[0055] 2) In the optimization method, individuals need to update and iterate around the local optimal solution. Initially, the number of local optimal solutions is set and determined. At the same time, during the iteration process, the number of local optimal solutions needs to be adaptively reduced. The adaptive reduction mechanism formula is as follows:

[0056] FN=round(Nt*(N-1) / T)

[0057] Where: N is the number of initial local optimal solutions; t is the current number of iterations; T is the maximum number of iterations, and FN represents the number of local optimal solutions in the tth iteration process.

[0058] 3) The fitness of different individuals is calculated according to the objective function, and the fitness is set to the inverse of the objective function: its fitness is g = 1 / f. When the fitness is greater, the quality of the individual is better. During the calculation process, the data corresponding to the individual, i.e., the latent variable, is input into the generation network to generate the airflow state distribution data. After the generation, the airflow is applied to different grids in the compost material simulation model according to the airflow state distribution data, and a simulation is performed after the application to calculate the gas concentration information and temperature data of the sensor setting point.

[0059] The fitness is sorted from large to small. After sorting, the corresponding number of local optimal solutions is selected according to the sorting result, and the individuals in the fitness sorting are taken as local optimal solutions. The number of local optimal solutions is calculated according to the above-mentioned fitness reduction mechanism formula, and the current best local optimal solution is stored. The best local optimal solution is the individual with the highest number of flames in the fitness sorting.

[0060] 4) According to the above individual update method, during the update process, the initial position update method is first used for updating. In the initial position update method, the above individual position update method is adaptively updated by the inertia weight:

[0061] Mi=Di·e bθ ·cos(2πθ)+ω·Fj

[0062] Among them, b is a constant used to define the logarithmic spiral shape; θ is a random number between [-1,1]; Mi represents the i-th individual; Fj represents the j-th local optimal solution; Di represents the distance between the i-th individual and the j-th local optimal solution, and Di is the constant of the logarithmic spiral shape; the formula is: Di = |Fj-Mi|, ω is the inertia weight, which draws on the inertia weight idea of the particle swarm algorithm. The larger ω is, the greater the search intensity of the algorithm in the global range, and the smaller ω is, the higher the search accuracy of the algorithm in the local range.

[0063]

[0064] Where t is the current iteration number; T is the maximum iteration number; a and d are constants.

[0065] 5) After the update, the individual position is updated again:

[0066]

[0067] Among them, MI represents the label of the individual to be updated again, rand() represents a random number, and Levy represents the corresponding Levy random step size, where:

[0068]

[0069] Among them, β is a constant, which has been set to 1.5, and the distribution number μ~N(0,σ 2 ), σ 2 represents the upper limit; random distribution number v~N(0,1);

[0070]

[0071] Where Г() represents the standard gamma function.

[0072] 6) Determine whether there is only one local optimal solution left. If there is only one left, directly output the final result;

[0073] 7) Determine whether the maximum number of iterations has been reached or the objective function is less than a certain threshold. If not, repeat the above steps 2)-6). If reached, output the optimal individual, that is, the airflow state distribution data finally found, and use the current airflow state distribution data as the airflow state distribution data at the corresponding moment of the observation value.

[0074] As some embodiments, for the airflow data and actual temperature and gas concentration in the composting area, there is a large error between the actual simulation calculation and the on-site data due to the influence of discrete grids, algorithms or other factors. Based on the comparison between the simulated values and the observed values, the calculation model involved in the simulation process is corrected in reverse.

[0075] During the correction process, the observable area is used as the basis, wherein the area in the observable area where the above-mentioned airflow state distribution data, temperature and gas concentration data can be effectively collected is not limited to the composting area. This method mainly corrects the calculation model in the above-mentioned Abaqus or fluent analysis model. The calculation model represents the relationship between different types of data in space and will not change due to the adjustment of the area type. During the correction process, the fixed parameters that may be involved in the calculation model, such as specific heat capacity, gas density and other fixed parameters, are first determined, and the parameters that may be involved in the calculation model that need to be corrected, such as proportional parameters, weight coefficients and other parameters are adjusted. At the same time, the grid parameters in the model are adjusted, such as the size and division method of the grid; the airflow state distribution data is used as the input factor, and the temperature and gas concentration are used as the observation values to adjust the calculation model and grid parameters.

[0076] Specifically, based on the above-mentioned optimization algorithm, the fixed parameters of the calculation model are first determined and substituted into the corresponding calculation model. The parameters that need to be modified among the above parameters are used as adjustable variables. The adjustable variables are used as individuals in the optimization algorithm, and the individuals are updated and optimized through the optimization algorithm. The samples relied on by the optimization algorithm include multiple sets of airflow data, namely airflow distribution data, temperature data, and gas concentration data. The algorithm specifically includes the following contents:

[0077] 1. Initialize the population, take the variables that need to be adjusted, including the parameters that need to be corrected and the grid parameters, as individuals, and use the parameter values that need to be corrected and the grid size values in the calculation model as initial values to initialize the individuals.

[0078] 2. Set relevant goals. Minimize the weighted sum of the differences between the actual temperature data and gas concentration data and the simulation data as the final goal. Select the above objective function f and construct a fitness function, where the fitness function can select the above g.

[0079] 3. Input the airflow data in the sample into the simulation model, simulate the airflow data by substituting the calculation model with the parameter values to be corrected and the grid using the grid parameters to obtain simulation data, wherein the simulation data includes temperature data and gas concentration data under different grids;

[0080] 4. Substitute the simulated data and the corresponding measured temperature data and gas concentration data in the sample into the fitness function to calculate its fitness.

[0081] 5. After the fitness calculation is completed, use steps 2)-6) in the above optimization algorithm to update the individual. After the update, substitute the updated individual into the calculation model and grid division parameters in the simulation model, and input the data in the sample into the simulation model after the individual for re-simulation.

[0082] 6. Repeat the above process 3-5 until the error between the measured data and the simulation data is less than a certain threshold or the maximum number of iterations is reached. The optimal individual with the smallest error is used as the optimal calculation model of the parameter values to be corrected and the grid parameters to be divided, and is substituted into the simulation model to realize the correction of the simulation model.

[0083] Through the above technical solution, the present invention does not need to obtain the overall regional airflow data through the mean of each environmental position or direct linear calculation, but calculates the airflow distribution data of each position in the entire area through preferentially arranged sensors. Due to the differences in different observation positions, the data of the observation position is affected by multiple data in its airflow distribution. At the same time, the data at different positions of the airflow distribution also have a certain correlation or distribution. Therefore, through limited observation points, most of the airflow state data in the area can be inverted. The present invention pre-generates the airflow state distribution data through a generation network, first satisfying the mutual correlation or certain distribution between the airflow state distribution data, and simulating the airflow state distribution data to generate corresponding environmental data on this basis. The simulated environmental data is evaluated according to the observed environmental data to meet the correlation of the airflow state distribution data with the impact on the environmental data. Through the above two correlation constraints, airflow state distribution data that conforms to the real scene is generated. The airflow state distribution data contains the airflow size and direction of different positions to solve the above-mentioned ending problem.

[0084] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A compost material airflow detection method, characterized in that: include: Randomly generate different airflow state distribution data in the compost material area according to the airflow law; The airflow state distribution data is simulated by a finite element simulation model, wherein the simulation results include simulated gas concentration data and temperature data; Obtain observation values, including gas concentration and temperature data in the composting area, and compare and evaluate simulation results based on the observation values. The gas concentration and temperature data are for different locations in the composting area and are collected by corresponding sensors. According to the evaluation results, the airflow state distribution data is iteratively adjusted through the optimization algorithm, and simulation and evaluation are performed based on the iteratively adjusted airflow state distribution data until the error between the observation value and the simulation result in the iterative adjustment is less than the threshold or the maximum number of iterations is reached, thereby obtaining the optimal airflow state distribution data.

2. The method according to claim 1, characterized in that The airflow state distribution data is generated by a generator in a generative adversarial network, wherein the generator is used to simulate the law of the airflow state distribution data, wherein the generative adversarial network is trained by training samples, wherein the training samples include airflow data measured in experiments and airflow data generated by simulation.

3. The method according to claim 2, characterized in that The generative adversarial network includes a generator and a discriminator, wherein the generator is used to generate airflow state distribution data based on latent variables and auxiliary variables, and different airflow state distribution data are obtained by adjusting the latent variables. The discriminator is used to score the airflow state distribution data based on training samples and auxiliary variables. The generator is optimized according to the score of the discriminator. During the optimization process, the training samples are used to optimize the generator.

4. The method according to claim 1, wherein Auxiliary variables include gas concentration data and temperature data.

5. The method according to claim 1, wherein In a finite element simulation model, a simulation model is constructed for the compost material area and the compost material, and the compost material area is used as a boundary constraint and the compost material is used as an emission source, and simulation calculations are performed through a fluid mechanics calculation model and a thermodynamic calculation model; The fluid mechanics calculation model and the thermodynamics calculation model are modified according to the collected airflow state distribution data and temperature and gas concentration data.

6. The method according to claim 1, characterized in that The airflow state distribution data is iteratively adjusted according to the evaluation results through an optimization algorithm, wherein, in a single adjustment of the airflow state distribution data, the latent variable is taken as an individual, and several individuals are taken as a population, wherein the airflow state distribution data is generated according to the latent variable; the individuals are adjusted according to the evaluation results, wherein the latent variables corresponding to the minimum errors between several observation values and simulation results in the evaluation results are taken as local optimal solutions, the population is updated according to the local optimal solution in combination with the inertia weight, and after the update, the updated population is updated again based on a random step size, and the corresponding airflow state distribution data is regenerated according to the individuals in the updated population.

7. The method according to claim 1, characterized in that The goal of the optimization algorithm is to minimize the weighted sum of the differences between the gas concentration data and temperature data in the composting material area and the simulation results.

8. A compost material airflow monitoring system, characterized in that: Used to perform the method according to any one of claims 1 to 7.

Images