Pesticide product quality analysis method

Abstract

The invention discloses a pesticide product quality analysis method, and relates to the technical field of quality analys.The method includes the steps that the flora metabolism state is analyzed through flow cytometry and single-cell RNA sequencing, an initial activity database is established, the posterior probability of multiple growth models is analyzed and calculated through Bayesian factors, the model weight is dynamically adjusted, and the growth rate of the multiple growth models is calculated; an activity prediction model is constructed, so that self-adaptive adjustment can be performed according to growth characteristics of different flora, and the prediction precision is improved; in combination with storage environment data, establishing environment influence activity change mapping by adopting a long-short term memory (LSTM) network, and optimizing a prediction model by utilizing multi-task learning, so that the prediction model is suitable for different storage and transportation conditions; in the storage stage, data such as temperature, humidity, illumination and oxygen concentration are obtained through an environment sensor, the flora activity is monitored in combination with an ATP bioluminescence detection method, actual detection values are input into an activity prediction model for correction, a flora activity change curve in the storage stage is generated, and the real-time performance and accuracy of model prediction are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of quality analysis, and particularly to a method for analyzing the quality of pesticide products. Background Art

[0002] In the quality control and analysis of biological pesticides, common stability assessment methods mainly follow the high-temperature accelerated degradation model of chemical pesticides, such as the Arrhenius equation, and predict the shelf life through exponential decay calculations under laboratory conditions; however, the active ingredients of biological pesticides are not simple chemical substances, but are composed of live bacterial populations or metabolites, and their stability is affected by environmental factors. During storage, the bacterial population undergoes a short period of adaptive growth before entering the decay stage, and under different temperature and humidity conditions, their metabolic activity may also change significantly, resulting in a large deviation between the shelf life determined in the laboratory and the actual use effect.

[0003] In response to the above problems, most manufacturers adopt low-temperature refrigeration in the storage link to slow down the decay rate of the bacterial population, and use biochemical detection methods such as ELISA to quantitatively analyze the residual situation of protein metabolites to determine and measure the activity level of pesticides. However, in such methods, the measured viable bacteria count is not completely equivalent to biological activity. Even if the bacteria survive, their metabolic level may have decreased significantly, resulting in an actual control effect lower than expected. Although the refrigeration strategy can extend the product storage period, once it enters the normal-temperature transportation or field application stage, decay is still inevitable.

[0004] In summary, it can be seen that the essence of the problem lies in that the activity change of biological pesticides is continuous and dynamic. The laboratory can measure the biological activity at a certain moment, but it cannot predict the activity trend during the entire storage, transportation, and even application stages. If the activity decay process of biological pesticides cannot be predicted, even if the factory inspection is qualified, the efficacy may still decline due to environmental changes during the circulation process, affecting market recognition and farmers' trust; therefore, there is an urgent need for a pesticide product quality analysis solution to solve such problems. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a method for analyzing the quality of pesticide products to solve the problem that traditional quality detection of biological pesticides lacks dynamic prediction and only relies on static viable bacteria count or component content, making it difficult to accurately reflect the activity changes after storage and application.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] An embodiment of the present invention provides a method for analyzing the quality of pesticide products, which includes,

[0009] Step S1: Analyze the metabolic state of the biological pesticide flora by flow cytometry combined with single-cell RNA sequencing, measure the viable cell count and metabolic activity, and establish an initial activity database.

[0010] Step S2: Based on the initial activity database in Step S1, construct an activity prediction model.

[0011] Step S3: Based on the activity prediction model in Step S2, combined with the storage environment data, use a long short-term memory network to establish an environmental impact-activity change mapping.

[0012] Step S4: Based on the environmental impact-activity change mapping in Step S3, combined with the actual storage environment data, calculate the decay rate of the flora to obtain the change curve of the flora activity during the storage period.

[0013] Step S5: Based on the change curve of the flora activity during the storage period, combined with the influence of different transportation conditions, use a long short-term memory network to simulate the activity change of the flora in the transportation environment, and dynamically adjust the prediction combined with the actual storage environment data to obtain the transportation stability parameters.

[0014] Step S6: According to the transportation stability parameters in Step S5, use photonic crystal sensing technology for on-site rapid detection, and combined with the activity prediction model to correct the change trend of the flora in the field environment, generate the change curve of the flora activity under the application conditions as the final quality evaluation standard.

[0015] As a preferred solution of the method for analyzing the quality of a pesticide product described in the present invention, wherein: in Step S2, use Bayesian factor analysis to calculate the posterior probability of multiple growth models and dynamically adjust the model weights to construct an activity prediction model.

[0016] As a preferred solution of the method for analyzing the quality of a pesticide product described in the present invention, wherein: the step of constructing an activity prediction model based on the initial activity database in Step S1 is as follows:

[0017] Define multiple candidate growth models M i , where i = 1, 2,..., n, representing different growth kinetic models, including the Logistic growth model, the Gompertz growth model, and the Baranyi Roberts growth model.

[0018] The parameters of each model M i are estimated according to the initial activity database in Step S1, and the growth model form is defined as:

[0019] G i (t) = f i (θ i , t),

[0020] where Gi (t) represents the model M i The predicted microbial community activity at time t, f i (θ i , t) is the growth model M i 's specific calculation function, θ i is the parameter set of model M i including the growth rate and the saturation activity value,

[0021] Let the observed data set D be the sequence of microbial community activities measured in the experiment, and calculate the posterior probability of each model M i The calculation formula is:

[0022]

[0023] where, P(M i |D) is the posterior probability of model M i under the condition of given data D, and P(D|M i ) is the marginal likelihood of model M i , indicating the goodness of fit of the model to the data. The calculation formula is:

[0024] P(D|M i ) = ∫P(D|θ i , M i )P(θ i |M i )dθ i ,

[0025] P(M i ) is the prior probability of model M i , indicating the credibility of the model before data observation,

[0026] is the normalization factor,

[0027] Calculate the Bayes factor B ij Compare model M i with the reference model M j . The calculation formula is:

[0028]

[0029] where, B ij is the Bayes factor of model M i relative to model M j , and P(D|M i ) and P(D|M j ) are the marginal likelihoods of model M i and M j ,

[0030] Dynamically adjust the model weights through posterior probability to construct a weighted prediction model

[0031]

[0032] Among them, is the final activity prediction model, and w i is the weight of model M i , defined as w i =P(M i |D).

[0033] As a preferred solution of the method for analyzing the quality of a pesticide product according to the present invention, wherein: in step S3, multi-task learning is used to optimize the applicability of the activity prediction model so that it is applicable to different storage and transportation conditions and serves as a reference model for activity monitoring during the storage stage;

[0034] The storage environment data includes temperature, humidity, light, and oxygen concentration data.

[0035] As a preferred solution of the method for analyzing the quality of a pesticide product according to the present invention, wherein: for the activity prediction model based on step S2, in combination with the storage environment data, the steps of establishing an environmental impact-activity change mapping using a long short-term memory network are as follows

[0036] Define the storage environment variable X t as the input of the LSTM model, expressed as:

[0037] X t =(T t ,H t ,L t ,O t ),

[0038] Among them, X t is the storage environment feature vector at time t, T t is the temperature, H t is the humidity, L t is the light intensity, O t is the oxygen concentration,

[0039] Use the long short-term memory LSTM network to learn the impact of environmental data on the activity of the flora. The calculation steps of the LSTM network include:

[0040] Calculate the forget gate:

[0041] f t =σ(W f X t +U f h t-1 +b f ),

[0042] Calculation of the input gate:

[0043] i t = σ(W i X t + U i h t-1 + b i ),

[0044] Calculation of the cell state update:

[0045]

[0046] Calculation of the output gate:

[0047] o t = σ(W o X t + U o h t-1 + b o ),

[0048] Calculation of the hidden state:

[0049] h t = o t ⊙ tanh(c t ),

[0050] where f t is the forget gate, i t is the input gate, o t is the output gate, is the candidate cell state, c t is the cell state, c t-1 represents the cell state at the previous time step, h t is the hidden state, h t-1 represents the hidden state at the previous time step, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, W f , W i , W o , W c are the input weight matrices, U f , U i , U o , U c are the hidden state weight matrices, b f , b i , b o , b c are the bias terms, ⊙ represents element-wise multiplication,

[0051] The hidden state h t calculated based on the LSTM and the activity prediction model in step S2 Perform activity calculation, and the calculation formula is:

[0052]

[0053] Among them, A t is the predicted value of the microbial community activity at time t, represents calculating the microbial community activity based on the LSTM state h t and the activity prediction model Calculate the microbial community activity.

[0054] As a preferred solution of a method for analyzing the quality of a pesticide product according to the present invention, wherein: in step S4, verify the microbial community activity level by the ATP bioluminescence detection method, input the microbial community activity level into the activity prediction model in step S2 for correction, and assist in generating the change curve of the microbial community activity during the storage period.

[0055] As a preferred solution of a method for analyzing the quality of a pesticide product according to the present invention, wherein: the step of calculating the microbial community decay rate based on the environmental impact-activity change mapping in step S3 and combining the actual storage environment data to obtain the change curve of the microbial community activity during the storage period is,

[0056] Let the observed value of the microbial community activity at the storage time t be The LSTM predicted value is Define the microbial community decay rate R t as:

[0057]

[0058] Among them, R t is the microbial community decay rate at time t, is the actually observed microbial community activity level, is the microbial community activity level predicted by the LSTM in step S3, and Δt is the time interval,

[0059] Use LSTM recursive prediction to calculate the change curve of the microbial community activity during the storage period. The calculation formula is:

[0060]

[0061] Among them, is the microbial community activity predicted by LSTM, h t is the hidden state of the LSTM model, is the function for LSTM to calculate the microbial community activity,

[0062] Obtain the actual microbial community activity level by the ATP bioluminescence detection method Correct the prediction model, and the correction formula is:

[0063]

[0064] Among them, is the corrected microbial community activity, is the measured value of the microbial community activity detected by ATP bioluminescence, and α is the correction coefficient.

[0065] As a preferred embodiment of the method for analyzing the quality of a pesticide product according to the present invention, wherein: the different transportation conditions include temperature control, normal temperature, and cold chain.

[0066] As a preferred embodiment of the method for analyzing the quality of a pesticide product according to the present invention, wherein: the step of combining the influence of different transportation conditions, using a long short-term memory network to simulate the activity change of the microbial community in the transportation environment, and dynamically adjusting the prediction in combination with the actual storage environment data to obtain the transportation stability parameter is as follows:

[0067] Define the transportation environment variable as

[0068]

[0069] Among them, is the transportation environment feature vector at time t, is the temperature during transportation, is the humidity during transportation, is the light during transportation, is the oxygen concentration during transportation,

[0070] Use the long short-term memory network LSTM to establish the mapping relationship between the transportation environment and the microbial community activity. The LSTM calculation formula is:

[0071]

[0072] Among them, is the hidden state of the transportation environment LSTM model, represents the mapping calculation of LSTM on the environmental variables, is the microbial community activity under the transportation environment at time t,

[0073] Define the transportation stability parameter S trans Evaluate the activity stability of the microbial community in different transportation environments. The formula is:

[0074]

[0075] Among them, S trans is the transportation stability parameter. The smaller the value, the smaller the fluctuation of the microbial community activity and the more stable it is. T is the number of time steps of the transportation cycle.

[0076] As a preferred solution of the method for analyzing the quality of a pesticide product according to the present invention, wherein: the step of generating the curve of the change in the activity of the microbial community under the application conditions is as follows:

[0077] The photon crystal sensing technology is used to monitor the activity of the microbial community in the field environment in real time to obtain detection data

[0078]

[0079] Wherein, is the value of the activity of the microbial community detected by photon crystal sensing, λ t is the wavelength response of the photon crystal, I t is the change in the light intensity of the photon crystal, g(λ t , I t ) is the photon crystal sensing activity measurement function,

[0080] The activity prediction model of step S2 is used to correct the activity of the microbial community in the field to obtain the final activity curve The correction formula is:

[0081]

[0082] Wherein, is the activity curve of the microbial community under field conditions, and β is the correction coefficient to control the correction amplitude of the model.

[0083] The beneficial effects of the present invention are as follows: In the present invention, the metabolic state of the microbial community is analyzed by flow cytometry and single-cell RNA sequencing, an initial activity database is established, the posterior probabilities of multiple growth models are calculated by Bayesian factor analysis, and the model weights are dynamically adjusted to construct an activity prediction model, enabling it to adaptively adjust according to the growth characteristics of different microbial communities, thereby improving the prediction accuracy.

[0084] In the present invention, combined with the stored environmental data, the long short-term memory network LSTM is used to establish a mapping of the environmental impact on the activity change, and multi-task learning is used to optimize the prediction model, making it applicable to different storage and transportation conditions, and making up for the deficiency that the traditional method cannot dynamically evaluate the impact of environmental factors on the pesticide activity; during the storage stage, data such as temperature, humidity, light, and oxygen concentration are obtained through environmental sensors, the activity of the microbial community is monitored in combination with the ATP bioluminescence detection method, and the actual detected values are input into the activity prediction model for correction to generate the curve of the change in the activity of the microbial community during the storage period, improving the real-time performance and accuracy of the model prediction.

[0085] In the transportation stage of the present invention, considering the influence of different transportation conditions on the activity of the bacterial community, an LSTM network is used to simulate the change of the bacterial community activity under the transportation environment, and combined with the real-time storage of environmental data for dynamic adjustment of the prediction, calculating the transportation stability parameter, so as to measure the activity fluctuation degree of the bacterial community under different transportation modes, providing support for the quality control of pesticide products in the supply chain; based on the transportation stability parameter, a photonic crystal sensing technology is adopted for on-site rapid detection, and combined with the activity prediction model to correct the change trend of the bacterial community in the field environment, generating the change curve of the bacterial community activity under the application conditions as the final basis for quality evaluation.

[0086] In summary, the present invention significantly improves the accuracy of predicting the activity of biological pesticides, and can adaptively adjust according to the actual environmental changes, ensuring the effectiveness of pesticides in each stage of storage, transportation and application, and reducing the risk of pesticide failure caused by the attenuation of the bacterial community activity. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0088] Figure 1 It is a schematic flow chart of the method for analyzing the quality of pesticide products of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0089] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention with reference to the drawings of the specification.

[0090] In the following description, many specific details are set forth to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar promotions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0091] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" appearing in different places in this specification does not all refer to the same embodiment, nor is it a separate or selectively exclusive embodiment from other embodiments.

[0092] Example 1, referring to Figure 1 , this embodiment provides a method for analyzing the quality of pesticide products, including the following steps:

[0093] Step S1, analyze the metabolic state of the biopesticide flora by flow cytometry combined with single-cell RNA sequencing, measure the viable cell count and metabolic activity, and establish an initial activity database;

[0094] Step S2, construct an activity prediction model based on the initial activity database in Step S1;

[0095] In Step S2, calculate the posterior probabilities of multiple growth models using Bayesian factor analysis and dynamically adjust the model weights to construct an activity prediction model;

[0096] The steps for constructing an activity prediction model based on the initial activity database in Step S1 are as follows:

[0097] Define multiple candidate growth models M i , where i = 1, 2, …, n, representing different growth kinetic models, including the Logistic growth model, Gompertz growth model, and Baranyi Roberts growth model;

[0098] The parameters of each model M i are estimated according to the initial activity database in Step S1, and the growth model form is defined as:

[0099] G i (t) = f i (θ i , t),

[0100] where G i (t) represents the predicted flora activity of model M i at time t, f i (θ i , t) is the specific calculation function of growth model M i , and θ i is the parameter set of model M i , including the growth rate and saturation activity value;

[0101] Let the observed data set D be the experimentally measured flora activity sequence, and calculate the posterior probability of each model M i , and the calculation formula is:

[0102]

[0103] where P(M i |D) is the posterior probability of model M i under the condition of given data D, and P(D|M i ) is the marginal likelihood of model M i , representing the goodness of fit of the model to the data, and the calculation formula is:

[0104] P(D|Mi ) = ∫P(D|θ i , M i )P(θ i |M i )dθ i ,

[0105] P(M i ) is the prior probability of model M i , representing the credibility of the model before data observation,

[0106] is the normalization factor,

[0107] Calculate the Bayes factor B ij Compare model M i with the reference model M j , and the calculation formula is:

[0108]

[0109] where B ij is the Bayes factor of model M i relative to model M j , P(D|M i ) and P(D|M j ) are the marginal likelihoods of models M i and M j ,

[0110] Dynamically adjust the model weights through the posterior probability to construct a weighted prediction model

[0111]

[0112] where is the final activity prediction model, w i is the weight of model M i , defined as w i = P(M i |D);

[0113] Specifically, based on the initial activity database constructed in step S1, use the Bayes factor analysis method to calculate the posterior probabilities of multiple growth models and dynamically adjust the model weights to improve the accuracy of the flora activity prediction model. The Bayes factor can effectively measure the fitting degree of different models to the observed data, enabling the activity prediction model to adaptively adjust, thereby improving the prediction accuracy;

[0114] Step S3, based on the activity prediction model in step S2, combined with the storage environment data, use the long short-term memory network to establish an environmental impact-activity change mapping;

[0115] In step S3, the applicability of the activity prediction model is optimized using multi-task learning to make it applicable to different storage and transportation conditions, serving as a reference model for activity monitoring during the storage stage;

[0116] The storage environment data includes temperature, humidity, light, and oxygen concentration data;

[0117] Based on the activity prediction model in step S2, combined with the storage environment data, the steps to establish the mapping of environmental impact - activity change using a long short-term memory network are as follows:

[0118] Define the storage environment variable X t As the input to the LSTM model, expressed as:

[0119] X t =(T t , H t , L t , O t ),

[0120] where X t is the storage environment feature vector at time t, T t is the temperature, H t is the humidity, L t is the light intensity, Q t is the oxygen concentration,

[0121] Use the long short-term memory (LSTM) network to learn the impact of environmental data on the activity of the microbial community. The calculation steps of the LSTM network include:

[0122] Calculate the forget gate:

[0123] f t =σ(W f X t +U f h t-1 +b f ),

[0124] Calculate the input gate:

[0125] i t =σ(W i X t +U i h t-1 +b i ),

[0126] Calculate the cell state update:

[0127]

[0128] Calculate the output gate:

[0129] o t =σ(Wo X t +U o h t-1 +b o ),

[0130] Calculate the hidden state:

[0131] h t = o t ⊙ tanh(c t ),

[0132] where f t is the forget gate, i t is the input gate, o t is the output gate, is the candidate cell state, c t is the cell state, c t-1 represents the cell state at the previous time step, h t is the hidden state, h t-1 represents the hidden state at the previous time step, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, W f , W i , W o , W c is the input weight matrix, U f , U i , U o , U c is the hidden state weight matrix, b f , b i , b o , b c is the bias term, ⊙ represents element-wise multiplication,

[0133] Calculate the activity based on the hidden state h t calculated by LSTM and the activity prediction model in step S2 The calculation formula is:

[0134]

[0135] where A t is the predicted value of the microbial community activity at time t, represents calculating the microbial community activity based on the LSTM state h t and the activity prediction model ;

[0136] Specifically, here, based on the activity prediction model in step S2 and combined with the storage environment data, a long short-term memory (LSTM) network is used to establish a mapping of the environmental impact on activity changes, and the trend of microbial community activity changes is predicted according to different storage conditions to achieve dynamic adjustment;

[0137] Step S4. Based on the environmental impact - activity change mapping in step S3, calculate the decay rate of the microbial community by combining the actual storage environment data, and obtain the change curve of the microbial community activity during the storage period;

[0138] In step S4, verify the activity level of the microbial community by the ATP bioluminescence detection method, input the activity level of the microbial community into the activity prediction model in step S2 for correction, and assist in generating the change curve of the microbial community activity during the storage period;

[0139] The step of calculating the decay rate of the microbial community by combining the actual storage environment data based on the environmental impact - activity change mapping in step S3 and obtaining the change curve of the microbial community activity during the storage period is as follows.

[0140] Let the observed value of the microbial community activity at the storage time t be The LSTM predicted value be Define the decay rate R of the microbial community t as:

[0141]

[0142] where R t is the decay rate of the microbial community at time t, is the actually observed activity level of the microbial community, is the activity level of the microbial community predicted by the LSTM in step S3, and Δt is the time interval.

[0143] Use LSTM recursive prediction to calculate the change curve of the microbial community activity during the storage period. The calculation formula is:

[0144]

[0145] where is the activity of the microbial community predicted by LSTM, h t is the hidden state of the LSTM model, is the function for LSTM to calculate the activity of the microbial community,

[0146] Obtain the actual activity level of the microbial community by the ATP bioluminescence detection method Correct the prediction model. The correction formula is:

[0147]

[0148] where is the corrected activity of the microbial community, is the measured value of the activity of the microbial community detected by ATP bioluminescence, and α is the correction coefficient;

[0149] Specifically, in combination with the actual storage environment data here, the decay rate of the microbial community is calculated to obtain the change curve of the microbial community activity during the storage period. The numerical differentiation method is used to ensure the real-time evaluation of the downward trend of the microbial community activity. At the same time, the ATP bioluminescence detection method is used to experimentally verify the microbial community activity, and the detected value is fed back to the prediction model for correction to improve the prediction accuracy of the model.

[0150] Step S5: Based on the change curve of the microbial community activity during the storage period, in combination with the influence of different transportation conditions, use the long short-term memory network to simulate the change of the microbial community activity in the transportation environment, and dynamically adjust the prediction in combination with the actual storage environment data to obtain the transportation stability parameters.

[0151] The different transportation conditions include temperature control, normal temperature, and cold chain.

[0152] The steps of using the long short-term memory network to simulate the change of the microbial community activity in the transportation environment in combination with the influence of different transportation conditions and dynamically adjusting the prediction in combination with the actual storage environment data to obtain the transportation stability parameters are as follows.

[0153] Define the transportation environment variable as

[0154]

[0155] where is the transportation environment feature vector at time t, is the temperature during transportation, is the humidity during transportation, is the light during transportation, is the oxygen concentration during transportation,

[0156] Use the long short-term memory network LSTM to establish the mapping relationship between the transportation environment and the microbial community activity. The LSTM calculation formula is:

[0157]

[0158] where is the hidden state of the transportation environment LSTM model, represents the mapping calculation of LSTM for the environmental variables, is the microbial community activity under the transportation environment at time t,

[0159] Define the transportation stability parameter S trans Evaluate the activity stability of the microbial community under different transportation environments. The formula is:

[0160]

[0161] where S transis the transportation stability parameter. The smaller the value, the smaller the fluctuation of the microbial community activity and the more stable it is. T is the number of time steps of the transportation cycle;

[0162] Specifically, based on the change curve of the microbial community activity during the storage period established in step S4, combined with the influence of different transportation conditions, the LSTM model is used to simulate the change of the microbial community activity in the transportation environment, and the prediction is dynamically adjusted by combining the actual storage environment data to obtain the transportation stability parameter. The LSTM network learns the dynamic change mode of the microbial community under different transportation environments, and calculates the transportation stability parameter through the prediction error to measure the fluctuation degree of the microbial community activity during transportation;

[0163] Step S6: According to the transportation stability parameter in step S5, use the photonic crystal sensing technology for on-site rapid detection, and combine the activity prediction model to correct the change trend of the microbial community in the field environment, and generate the change curve of the microbial community activity under the application conditions as the final quality evaluation standard;

[0164] The steps to generate the change curve of the microbial community activity under the application conditions are as follows:

[0165] Use the photonic crystal sensing technology to monitor the activity of the microbial community in the field environment in real time to obtain the detection data

[0166]

[0167] Among them, is the activity value of the microbial community detected by the photonic crystal sensing, λ t is the wavelength response of the photonic crystal, I t is the change in the light intensity of the photonic crystal, g(λ t ,I t ) is the photonic crystal sensing activity measurement function,

[0168] Use the activity prediction model in step S2 to correct the activity of the microbial community in the field to obtain the final activity curve The correction formula is:

[0169]

[0170] Among them, is the activity curve of the microbial community under field conditions, and β is the correction coefficient to control the model correction amplitude;

[0171] Specifically, based on the transportation stability parameter in step S5, use the photonic crystal sensing technology for on-site rapid detection, and combine the activity prediction model to correct the change trend of the microbial community in the field environment, and finally generate the change curve of the microbial community activity under the application conditions as the quality evaluation standard;

[0172] The photonic crystal sensing technology can achieve highly sensitive detection of the activity of the bacterial community, determine the biological activity level of the bacterial community through the changes in light intensity and wavelength, and in combination with the activity prediction model constructed in step S2, further correct the actual change trend of the activity of the bacterial community in the field environment and improve the accuracy of the curve.

[0173] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for analyzing the quality of pesticide products, characterized in that: include, Step S1, using flow cytometry combined with single-cell RNA sequencing to analyze the metabolic state of the biopesticide bacterial community, determine the number of live bacteria and metabolic activity, and establish an initial activity database; Step S2, constructing an activity prediction model based on the initial activity database of step S1; Step S3, based on the activity prediction model of step S2 and in combination with the stored environmental data, a long short-term memory network is used to establish an environmental impact-activity change mapping; Step S4, based on the environmental impact-activity change mapping of step S3, the bacterial flora decay rate is calculated in combination with the actual storage environment data to obtain a bacterial flora activity change curve during storage; Step S5, based on the bacterial flora activity change curve during storage period and combined with the influence of different transportation conditions, the long short-term memory network is used to simulate the activity change of the bacterial flora under the transportation environment, and the prediction is dynamically adjusted in combination with the actual storage environment data to obtain the transportation stability parameter; Step S6, based on the transport stability parameters of step S5, use photonic crystal sensing technology to perform rapid on-site detection, and combine the activity prediction model to correct the changing trend of the flora in the field environment, and generate a flora activity change curve under the application conditions as the final quality evaluation standard.

2. A method for analyzing the quality of pesticide products according to claim 1, characterized in that: In step S2, the posterior probabilities of multiple growth models are calculated using Bayesian factor analysis, and the model weights are dynamically adjusted to construct an activity prediction model.

3. A method for analyzing the quality of pesticide products as claimed in claim 2, characterized in that: The step of constructing the activity prediction model based on the initial activity database in step S1 is: Define multiple candidate growth models M i , where i = 1, 2, ..., n, represents different growth kinetic models, including the Logistic growth model, the Gompertz growth model and the Baranyi Roberts growth model. Each model M i The parameters of are estimated based on the initial activity database in step S1, and the growth model is defined as: G i (t)=f i (θ i ,t), Among them, G i (t) represents the model M i Predicted bacterial colony activity at time t, f i (θ i ,t) is the growth model M i The specific calculation function of θ i For model M i The parameter set includes growth rate and saturation activity value, Assume that the observed data set D is the bacterial activity sequence measured experimentally, and calculate each model M i The posterior probability is calculated as: Among them, P(M i |D) is the model M under the given data D condition. i The posterior probability, P(D|M i ) is the model M i The marginal likelihood of , which indicates the goodness of fit of the model to the data, is calculated as: P(D|M i )=∫P(D|θ i ,M i )P(θ i |M i )dθ i , P(M i ) is the model M i The prior probability of , which indicates the credibility of the model before data observation, is the normalization factor, Calculate Bayes factor B ij Compare Model M i and the reference model M j , the calculation formula is: Among them, B ij For model M i Relative to model M j The Bayes factor of P(D|M i ) and P(D|M j ) is the model M i and M j The marginal likelihood of Dynamically adjust the model weights through posterior probability to build a weighted prediction model in, is the final activity prediction model, w i For model M i The weight is defined as w i =P(M i |D).

4. A method for analyzing the quality of pesticide products as claimed in claim 3, characterized in that: In step S3, multi-task learning is used to optimize the applicability of the activity prediction model so that it is applicable to different storage and transportation conditions and serves as a reference model for activity monitoring during the storage stage; The storage environment data includes temperature, humidity, light and oxygen concentration data.

5. A method for analyzing the quality of pesticide products as claimed in claim 4, characterized in that: The activity prediction model based on step S2, combined with the stored environmental data, uses a long short-term memory network to establish an environmental impact-activity change mapping step, Define storage environment variable X t is the input of the LSTM model, expressed as: X t =(T t ,H t ,L t ,O t ), Among them, X t is the storage environment feature vector at time t, T t is temperature, H t is humidity, L t is the light intensity, O t is the oxygen concentration, The effect of environmental data on bacterial activity is studied using a long short-term memory (LSTM) network. The LSTM network calculation steps include: Calculate the forget gate: f t =σ(W f X t +U f h t-1 +b f ), Calculate the input gate: i t =σ(W i X t +U i h t-1 +b i ), Compute cell state updates: Compute the output gate: the t =σ(W o X t +U o h t-1 +b o ), Compute the hidden state: h t =o t ⊙tanh(c t ), Among them, f t For the forget gate, i t is the input gate, o t is the output gate, is the candidate cell state, c t is the cell state, c t-1 represents the cell state at the previous time step, h t is the hidden state, h t-1 represents the hidden state of the previous time step, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, W f ,W i ,W o ,W c is the input weight matrix, U f ,U i ,U o ,U c is the hidden state weight matrix, b f ,b i ,b o ,b c is the bias term, ⊙ represents element-by-element multiplication, The hidden state h calculated based on LSTM t and the activity prediction model of step S2 The activity calculation is performed using the following formula: Among them, A t is the predicted value of bacterial flora activity at time t, Represents the LSTM state h t and activity prediction models Calculate bacterial colony activity.

6. A method for analyzing the quality of pesticide products as claimed in claim 5, characterized in that: In step S4, the bacterial activity level is verified by ATP bioluminescence detection method, and the bacterial activity level is input into the activity prediction model of step S2 for correction, so as to assist in generating the bacterial activity change curve during storage period.

7. A method for analyzing the quality of pesticide products as claimed in claim 6, characterized in that: The step of calculating the bacterial flora decay rate based on the environmental impact-activity change mapping in step S3 and combining the actual storage environment data to obtain the bacterial flora activity change curve during the storage period is as follows: Assume that the observed value of bacterial colony activity at storage time t is The LSTM prediction value is Define the bacterial population decay rate R t for: Among them, R t is the bacterial colony decay rate at time t, is the actual observed bacterial activity level, is the bacterial community activity level predicted by step S3LSTM, Δt is the time interval, LSTM was used to recursively predict the change of bacterial flora activity and calculate the change curve of bacterial flora activity during storage. The calculation formula is: in, is the bacterial community activity predicted by LSTM, h t is the hidden state of the LSTM model, The function for LSTM to calculate bacterial colony activity, ATP bioluminescence detection method is used to obtain the actual bacterial activity level The prediction model is corrected, and the correction formula is: in, is the corrected bacterial activity, is the bacterial colony activity measurement value detected by ATP bioluminescence, and α is the correction factor.

8. A method for analyzing the quality of pesticide products according to claim 7, characterized in that: The different transportation conditions include temperature control, normal temperature and cold chain.

9. A method for analyzing the quality of pesticide products as claimed in claim 8, characterized in that: The steps of combining the influence of different transportation conditions, using the long short-term memory network to simulate the activity changes of the flora under the transportation environment, and dynamically adjusting the prediction in combination with the actual storage environment data to obtain the transportation stability parameter are as follows: Define the transport environment variables as in, is the characteristic vector of the transportation environment at time t, is the temperature during transportation, is the humidity during transportation, For light during transportation, is the oxygen concentration during transportation, The long short-term memory network LSTM is used to establish the mapping relationship between the transportation environment and the bacterial flora activity. The LSTM calculation formula is: in, is the hidden state of the transportation environment LSTM model, Represents the LSTM's mapping calculation of environmental variables. is the bacterial activity in the transport environment at time t, Define the transport stability parameter S trans To evaluate the activity stability of the bacterial flora under different transportation environments, the formula is: Among them, S trans is the transport stability parameter. The smaller the value, the smaller the fluctuation of bacterial activity and the more stable it is. T is the number of time steps in the transport cycle.

10. A method for analyzing the quality of pesticide products according to claim 9, characterized in that: The step of generating a curve of bacterial flora activity change under application conditions is: Photonic crystal sensing technology is used to monitor the activity of bacterial flora in the field environment in real time and obtain detection data in, is the bacterial activity value detected by photonic crystal sensing, λ t is the wavelength response of the photonic crystal, I t is the light intensity change of the photonic crystal, g(λ t ,I t ) is the photonic crystal sensing activity measurement function, Using the activity prediction model of step S2 Correct the activity of field flora to obtain the final activity curve The correction formula is: in, is the bacterial community activity curve under field conditions, and β is the correction coefficient, which controls the correction amplitude of the model.

Images