Deep learning driven poultry farm air shower system fault early warning method and system
By collecting multi-dimensional operating parameters and constructing a deep learning-driven disinfection evaluation model, the problem of inaccurate fault warning in existing air shower systems has been solved. This enables timely fault identification and effective control of air shower systems, ensuring the disinfection effect and environmental safety of poultry farms.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGSU INST OF POULTRY SCI
- Filing Date
- 2026-03-13
- Publication Date
- 2026-06-12
AI Technical Summary
Existing methods for early warning of air shower system malfunctions rely on monitoring a single parameter, which makes it difficult to comprehensively and accurately reflect the actual operating conditions and to detect in a timely manner any issues such as poor UV disinfection or potential malfunctions, thus affecting the disinfection effect.
By collecting multi-dimensional operating parameters in real time, a deep learning-driven disinfection assessment model is constructed, which includes assessment channels for ultraviolet dose, fogging effect, and chemical residue. Combined with graded early warning rules, a fusion analysis is performed to generate graded early warning signals.
It enables timely detection and accurate early warning of air shower system malfunctions, ensures stable system operation, improves disinfection effectiveness and management efficiency, and promotes the intelligent and scientific development of poultry farming.
Smart Images
Figure CN122196744A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of disinfection technology in poultry farms, specifically to a deep learning-driven method and system for early warning of faults in air shower systems for poultry farms. Background Technology
[0002] With the continuous development of the poultry farming industry, the scale and number of farms are increasing day by day. As a key piece of equipment for disinfection in farms, the stable operation of air shower systems is crucial to ensuring the health of poultry and the safety of the farming environment. However, existing methods for early warning of air shower system malfunctions often rely on monitoring a single parameter, which is difficult to comprehensively and accurately reflect the actual operating status of the air shower system, resulting in the inability to detect problems such as poor ultraviolet disinfection effects in a timely manner.
[0003] Moreover, existing early warning methods lack comprehensive analysis of multi-dimensional operating parameters and cannot effectively identify potential faults in the air shower system. This may result in situations such as the disinfectant not being fully effective or the airflow not being timely, affecting the disinfection effect. Summary of the Invention
[0004] This application provides a deep learning-driven method and system for early warning of faults in poultry farm air shower systems, which solves the technical problems of existing technologies being unable to effectively identify potential faults and unable to detect inadequate ultraviolet disinfection effects in a timely manner.
[0005] The technical solution to the above-mentioned technical problems in this application is as follows:
[0006] In a first aspect, this application provides a deep learning-driven method for early warning of faults in air shower systems in poultry farms, the method comprising:
[0007] Real-time acquisition of multi-dimensional operating parameters of the air shower system, wherein the multi-dimensional operating parameters include at least ultraviolet lamp operating parameters, air-mist shower timing parameters, and environmental state parameters;
[0008] Construct a deep learning-driven disinfection assessment model, wherein the disinfection assessment model includes at least an ultraviolet dose assessment channel, a fogging effect assessment channel, and a chemical residue assessment channel;
[0009] The multi-dimensional operating parameters are input into the disinfection assessment model to obtain multi-dimensional disinfection assessment results, and then combined with preset hierarchical early warning rules for fusion analysis to obtain hierarchical early warning signals, wherein the hierarchical early warning signals include fault type identifiers.
[0010] The feedback of the graded early warning signal will trigger a fault warning.
[0011] Secondly, this application provides a deep learning-driven fault early warning system for air shower systems in poultry farms, including:
[0012] The parameter acquisition module is used to acquire multi-dimensional operating parameters of the air shower system in real time. The multi-dimensional operating parameters include at least the ultraviolet lamp operating parameters, the air-mist shower timing parameters, and the environmental state parameters.
[0013] The model building module is used to build a deep learning-driven disinfection assessment model, wherein the disinfection assessment model includes at least an ultraviolet dose assessment channel, a fogging effect assessment channel, and a chemical residue assessment channel.
[0014] The result evaluation module is used to input the multi-dimensional operating parameters into the disinfection evaluation model, obtain multi-dimensional disinfection evaluation results, and perform fusion analysis in combination with preset hierarchical early warning rules to obtain hierarchical early warning signals, wherein the hierarchical early warning signals include fault type identifiers.
[0015] The signal feedback module is used to provide feedback on the graded early warning signal to perform fault early warning.
[0016] This application provides one or more technical solutions, which have at least the following technical effects or advantages:
[0017] This application provides a deep learning-driven method and system for early warning of faults in air shower systems for poultry farms. First,
[0018] Real-time acquisition of multi-dimensional operating parameters of the air shower system can comprehensively and accurately reflect its actual operating status, overcoming the shortcomings of existing methods that rely on single-parameter monitoring. Secondly, a deep learning-driven disinfection assessment model is constructed, including channels for ultraviolet dose assessment, misting effect assessment, and chemical residue assessment. This model can evaluate the disinfection effect of the air shower system from different perspectives. Then, the multi-dimensional operating parameters are input into the disinfection assessment model to obtain multi-dimensional disinfection assessment results, which are then fused and analyzed with preset graded early warning rules to obtain graded early warning signals. This comprehensive analysis effectively identifies potential faults in the air shower system and generates corresponding graded early warning signals based on different assessment results, enabling farm managers to understand the system's operating status in a timely manner. Finally, the graded early warning signals are fed back to execute fault warnings, and the parameter adjustment direction is obtained based on the analysis of the graded early warning signals. The air-mist timing parameters are iteratively optimized, and a fault control plan is output.
[0019] Through the above technical solutions, this application achieves timely detection, accurate early warning, and effective control of air shower system malfunctions, ensuring the stable operation of the air shower system and thus providing strong protection for poultry health and the safety of the breeding environment. Deep learning improves the management efficiency and disinfection effect of the air shower system in poultry farms, promoting the intelligent and scientific development of the poultry farming industry. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0021] Figure 1 This is a flowchart illustrating the deep learning-driven fault early warning method for air shower systems in poultry farms provided in this application embodiment;
[0022] Figure 2 This is a schematic diagram of the structure of the deep learning-driven fault early warning system for air shower systems in poultry farms provided in the embodiments of this application.
[0023] The components represented by each number in the attached diagram are explained below:
[0024] Parameter acquisition module 11, model building module 12, result evaluation module 13, signal feedback module 14. Detailed Implementation
[0025] This application provides a deep learning-driven method and system for early warning of faults in poultry farm air shower systems, which addresses the technical problems of existing technologies being unable to effectively identify potential faults and failing to detect inadequate ultraviolet disinfection effects in a timely manner.
[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0027] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.
[0028] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid unnecessarily obscuring the description of this application. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.
[0029] Example 1, as Figure 1 As shown in the embodiment of this application, a deep learning-driven fault early warning method for air shower systems in poultry farms is provided, including:
[0030] S10: Real-time acquisition of multi-dimensional operating parameters of the air shower system, wherein the multi-dimensional operating parameters include at least ultraviolet lamp operating parameters, air-mist shower timing parameters, and environmental state parameters;
[0031] Specifically, the operating parameters of the ultraviolet lamp include at least the ultraviolet radiation intensity, the cumulative irradiation time of the ultraviolet lamp tube, and the rated irradiation time;
[0032] The wind-mist timing parameters include at least disinfectant spray timing control parameters and airflow blowing timing control parameters;
[0033] The environmental state parameters include at least ambient temperature and ambient humidity.
[0034] In this embodiment of the application, firstly, sensor devices are used to collect multi-dimensional operating parameters of the air shower system, including ultraviolet lamp operating parameters, air-mist shower timing parameters, and environmental state parameters.
[0035] The operating parameters of the ultraviolet (UV) lamp include at least the UV radiation intensity, the cumulative irradiation time of the UV lamp, and the rated irradiation time. A UV radiation intensity sensor is used to monitor the UV radiation intensity in real time; this sensor converts the UV intensity into an electrical signal. Simultaneously, by recording the on and off times of the UV lamp and combining this with the rated irradiation time, the cumulative irradiation time of the UV lamp is calculated.
[0036] Secondly, the programmable logic controller (PLC) collects the air-mist timing parameters to record the control parameters for both the disinfectant spraying and airflow blowing. The air-mist timing parameters include the disinfectant spraying timing control parameters and the airflow blowing timing control parameters. According to a preset program, the spraying time and interval of the disinfectant, as well as the start and end times of the airflow blowing, are controlled, and the data is fed back in real time.
[0037] Next, temperature and humidity sensors are used to collect environmental parameters. These sensors are adaptable to the complex environment of the farm, measuring ambient temperature and humidity. The collected multi-dimensional operating parameters are transmitted to the data processing center for preliminary filtering and calibration to remove noise interference and ensure data accuracy and reliability. The high-quality, multi-dimensional operating parameters collected lay the foundation for subsequently building a deep learning-driven disinfection assessment model and accurately evaluating the operation of the air shower system.
[0038] S20: Construct a deep learning-driven disinfection assessment model, wherein the disinfection assessment model includes at least an ultraviolet dose assessment channel, a fogging effect assessment channel, and a chemical residue assessment channel;
[0039] In this embodiment, a deep learning-driven disinfection evaluation model is constructed using a deep neural network. First, historical data is collected, covering multi-dimensional operating parameters of the air shower system under different operating conditions and the corresponding disinfection effect evaluation results. The collected data is then used to train the model to ensure that it can learn the complex relationships between different parameters and disinfection effectiveness.
[0040] First, for the ultraviolet dose assessment channel, parameters such as ultraviolet radiation intensity, cumulative irradiation time of ultraviolet lamps and rated irradiation time are input into the multilayer sensor network. The actual ultraviolet dose is calculated based on the input data and compared with the preset standard dose to evaluate the effectiveness of ultraviolet disinfection.
[0041] Secondly, the fogging effect evaluation channel focuses on the spraying timing control parameters of the disinfectant in the wind-fog timing parameters. A neural network is first constructed, using parameters such as the spraying time and interval of the disinfectant as input. By analyzing the impact of the input parameters on factors such as droplet distribution and coverage, the effectiveness of the fogging is then evaluated.
[0042] Furthermore, the chemical residue assessment pathway considers environmental parameters and the use of pesticides. A recurrent neural network is used to capture time-series information from the data. By analyzing data related to ambient temperature, humidity, and chemical residues, the risk level of chemical residues can be assessed.
[0043] Furthermore, during model training, stochastic gradient descent is employed to optimize model parameters, thereby improving model accuracy and generalization ability. Simultaneously, regularization techniques such as L1 and L2 regularization are used to prevent overfitting. After multiple iterations of training, when the model's performance on the validation set stabilizes and meets the requirements, a deep learning-driven elimination evaluation model is obtained.
[0044] The process of constructing a deep learning-driven disinfection assessment model includes:
[0045] Acquire intrinsic ultraviolet (UV) sample data of the target scene, and extract behavioral features based on the intrinsic UV sample data to obtain UV usage behavior features;
[0046] Using the aforementioned ultraviolet (UV) usage behavior characteristics as an invariant constraint, and combining the tolerance constraint defined based on the device stability parameters of the target UV device, homogeneous sample collection is performed to obtain UV homogeneous sample data.
[0047] Based on the intrinsic ultraviolet sample data and the homologous ultraviolet sample data, a training method is used to obtain an ultraviolet intensity baseline, wherein the ultraviolet intensity baseline takes the cumulative irradiation time as input and outputs the real-time ultraviolet intensity.
[0048] The product of the ultraviolet intensity baseline and the rated irradiation time is defined as the real-time ultraviolet dose, and the ratio of the real-time ultraviolet dose to the rated ultraviolet dose of the target scene is defined as the ultraviolet dose assessment result, thus constructing the ultraviolet dose assessment channel.
[0049] In this embodiment, firstly, intrinsic ultraviolet (UV) sample data of the target scene is collected through a data acquisition system, including UV operation information under different conditions. Then, a feature extraction algorithm is used to extract features reflecting UV usage behavior from the intrinsic sample data, such as the timing pattern of UV activation and intensity variation patterns.
[0050] Secondly, the extracted ultraviolet (UV) usage behavior characteristics are used as invariant constraints to ensure consistency between the data and the UV usage behavior in the target scene during subsequent sample collection. Simultaneously, tolerance constraints defined based on the device stability parameters of the target UV device are used to reasonably limit the scope of sample collection. Under these constraints, UV homologous sample data from the same source as the intrinsic sample data are collected.
[0051] Then, based on the collected intrinsic and homologous ultraviolet (UV) sample data, a UV intensity baseline is constructed using statistical analysis and machine learning algorithms. This baseline takes cumulative exposure time as input and outputs real-time UV intensity through learning and analysis of a large amount of data.
[0052] Finally, based on the relationship between the baseline UV intensity and the rated exposure time, the real-time UV dose is defined as the product of the baseline UV intensity and the rated exposure time. The real-time UV dose is compared with the rated UV dose of the target scene, and the ratio between the two is calculated as the UV dose assessment result.
[0053] Based on the above methods, an ultraviolet dose assessment channel is constructed, which can accurately assess the effectiveness of ultraviolet disinfection and provide a basis for early warning of air shower system malfunctions.
[0054] Furthermore, a deep learning-driven disinfection evaluation model is constructed, wherein the construction process of the fogging effect evaluation channel includes:
[0055] Based on the historical disinfection records of the target scene, determine the most unfavorable fogging conditions, including the most unfavorable object conditions and the most unfavorable environmental conditions.
[0056] Construct a fog simulation analysis model, execute the simulation analysis, and obtain the simulation analysis results;
[0057] Based on the simulation analysis results, fog evaluation indicators are calculated and feature layer fusion is performed accordingly to obtain the fog effect evaluation value of the sample. The fog evaluation indicators include at least coverage, the most unfavorable effective wetting time and the deposition amount per unit area.
[0058] Randomly extract from the simulation analysis results, and calculate the sample entropy based on the sample fog effect evaluation value based on the extracted results;
[0059] The simulation analysis and extraction are performed iteratively until the sample entropy and sample quantity both meet the preset binary sample constraints. The extraction results are used as training data to construct and supervise the training of the disinfection evaluation model based on deep learning.
[0060] In this embodiment, the historical disinfection records of the target scene are first analyzed to identify the most unfavorable fogging conditions. The most unfavorable object conditions may involve factors such as the material, shape, and surface area of the object being disinfected, affecting the adhesion and coverage of fog droplets. The most unfavorable environmental conditions include ambient temperature, humidity, and wind speed; different environmental conditions will affect the evaporation rate and diffusion range of fog droplets.
[0061] Secondly, after determining the most unfavorable fogging conditions, a fogging simulation analysis model is constructed. This model considers multiple factors such as droplet generation, trajectory, and interaction with the target organism. Computer simulation technology is used to analyze the fogging process, simulating droplet distribution and coverage under different conditions, thereby obtaining simulation analysis results.
[0062] Next, based on the simulation analysis results, fogging evaluation indicators were calculated. Coverage reflects the degree of droplet coverage on the surface of the object being disinfected and is one of the indicators for evaluating fogging effectiveness. The most unfavorable effective wetting time refers to the time under the most unfavorable conditions for droplets to remain moist on the surface of the object being disinfected. Deposition amount per unit area reflects the number of droplets deposited per unit area, affecting the effectiveness of the disinfectant. The fogging evaluation indicators were fused using a feature layer, comprehensively considering the influence of each indicator, to obtain the sample fogging effectiveness evaluation value.
[0063] Then, random samples are extracted from the simulation analysis results. After each extraction, the sample entropy is calculated based on the sample fog effect evaluation value corresponding to the extraction result. Sample entropy can measure the complexity and uncertainty of the data, and is used to determine whether the extracted data is sufficiently representative.
[0064] Furthermore, simulation analysis and extraction operations are continuously iterated until both sample entropy and sample quantity meet the preset binary sample constraints. These preset binary sample constraints are set based on actual needs and experience to ensure that the training data has sufficient diversity while also guaranteeing data quality and representativeness. Once the constraints are met, the extracted results are used as training data to construct a deep learning-based disinfection evaluation model. During model training, supervised learning is employed, using known sample fogging effect evaluation values as labels. This allows the model to learn the relationship between input data and fogging effects, continuously adjusting model parameters to improve accuracy and generalization ability. This enables accurate evaluation of fogging effects and provides strong support for fault early warning of air shower systems.
[0065] For example, a disinfection assessment model is built and trained based on a deep learning model. The specific steps are as follows:
[0066] First, data preparation involves collecting and organizing multi-dimensional operating parameters of the air shower system, including previously collected ultraviolet lamp operating parameters, air-mist shower timing parameters, and environmental condition parameters. Corresponding disinfection effect evaluation data are also collected. The collected data is then cleaned and preprocessed to remove outliers and missing values.
[0067] Secondly, the model is built by taking multi-dimensional operating parameter data as input and evaluating the fog shower effect as output. The number of nodes in the input layer is equal to the dimension of the input features. For example, if the multi-dimensional operating parameters of the air shower system have a total of 3 features, then the input layer contains 3 nodes. 1-3 hidden layers are set, and the number of nodes in each layer is adjusted through experiments, such as 64, 32, etc. The activation function is ReLU. The number of nodes in the output layer is equal to the number of fog shower effects evaluated. For example, if the evaluation time requires 1 node, the output layer generally does not use an activation function and directly outputs continuous values.
[0068] Finally, the model is trained, and its parameters are continuously adjusted to minimize the error between predicted and actual values. Cross-validation is used to evaluate the trained model, verifying its generalization ability and accuracy. In each training iteration, known sample fogging effect evaluation values are used as supervision labels. The Adam optimizer and mean squared error (MSE) loss function are used to construct the training framework, and the model parameters are adjusted using the gradient descent algorithm. The batch size is set to 32, the total number of training epochs is 50, and an early stopping mechanism with a patience of 5 is introduced. When the validation set loss does not decrease for 5 consecutive epochs, the training process is automatically terminated, resulting in the trained fogging evaluation model.
[0069] Furthermore, a deep learning-driven disinfection assessment model is constructed, wherein the construction process of the chemical residue assessment channel includes:
[0070] Based on prior knowledge, obtain the safety residue constraints corresponding to the disinfectant preparations;
[0071] Based on the safety residue constraints, an orthogonal experimental scheme is planned in conjunction with the air shower parameter space of the target air shower system, and the orthogonal experimental scheme is executed to obtain chemical residue training data, wherein the chemical residue training data is marked with safety Boolean values;
[0072] Based on the chemical residue training data, a chemical residue evaluation channel based on an LSTM network is trained under supervision.
[0073] In this embodiment, the first step is to determine the safety residue constraints for the disinfectant based on prior knowledge. Prior knowledge is derived from relevant industry standards, professional research, and past practical experience. For example, by referring to safety standards for chemical residues after the use of disinfectants in poultry farms, or by drawing on successful experiences in farms of similar size and environment, the maximum permissible residue levels of disinfectants on environmental and equipment surfaces under different conditions are determined, thereby setting safety residue constraints.
[0074] Secondly, based on the determined safety residue constraints, an orthogonal experimental design was planned using the air shower parameter space of the target air shower system. The air shower parameter space includes the concentration of the disinfectant, spraying time, airflow velocity, ambient temperature, and humidity. Through orthogonal experimental design, the influence of each air shower parameter on chemical residues was investigated with as few experiments as possible.
[0075] For example, different concentration gradients of disinfectant agents, different spraying time intervals, and different combinations of airflow speeds are set. During the execution of the orthogonal experimental design, the chemical residue situation after each experiment is measured and recorded to obtain chemical residue training data. Furthermore, a safety Boolean value is assigned to the training data, that is, whether the chemical residue in this experiment meets the previously set safety residue constraints. If it meets the constraints, it is marked as "safe" with a Boolean value of 1; otherwise, it is marked as "unsafe" with a Boolean value of 0.
[0076] Furthermore, Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network that can solve the gradient vanishing or gradient exploding problems that occur when traditional recurrent neural networks process long sequences of data.
[0077] Finally, based on the labeled chemical residue training data, a method similar to that used in training the disinfection assessment model was employed to supervise the training of the chemical residue assessment channel using a Long Short-Term Memory (LSTM) network. The LSTM network can process time-series data and is suitable for analyzing the temporal correlation in chemical residue data. During training, the chemical residue training data was input into the LSTM network, with labeled safety Boolean values used as supervision labels. By continuously adjusting parameters such as network weights and biases, the network learned the complex relationship between chemical residue data and safety status. As training progressed, the network gradually became able to accurately predict whether a chemical residue was safe based on the input data, thus achieving effective assessment of chemical residues. This provides a basis for early warning of air shower system malfunctions, ensuring the safe and reliable operation of air shower systems in poultry farms.
[0078] S30: Input the multi-dimensional operating parameters into the disinfection evaluation model to obtain the multi-dimensional disinfection evaluation results, and perform fusion analysis in combination with the preset hierarchical early warning rules to obtain hierarchical early warning signals, wherein the hierarchical early warning signals include fault type identifiers;
[0079] In this embodiment, after inputting multi-dimensional operational parameters into the disinfection evaluation model, the model analyzes and processes the input multi-dimensional operational parameters based on the relationship between different parameters and disinfection effects learned during previous training, thereby outputting multi-dimensional disinfection evaluation results. These multi-dimensional disinfection evaluation results include information such as ultraviolet dose assessment, fogging effect assessment, and chemical residue assessment.
[0080] Secondly, the multi-dimensional disinfection assessment results are integrated and analyzed by combining preset graded early warning rules. These rules are pre-set based on the normal operating standards of the air shower system and potential malfunctions. For example, in ultraviolet (UV) dose assessment, if the real-time UV dose is lower than a certain percentage of the rated UV dose, the UV disinfection effect may be deemed poor; in fog shower effect assessment, if the coverage is below a certain threshold, the most unfavorable effective wetting time is too short, or the deposition per unit area is insufficient, it may mean that the fog shower effect is substandard; and in chemical residue assessment, if the predicted chemical residue is in an unsafe state, then there is a risk related to chemical residue.
[0081] Furthermore, based on the matching results of different assessments with the tiered early warning rules, corresponding tiered early warning signals are generated. These signals include fault type identifiers. The fault type identifier can be specific to a problem occurring in a particular assessment channel, such as "insufficient UV dose fault," "poor misting effect fault," or "excessive chemical residue fault." Through this method, potential faults in the air shower system can be identified, allowing for timely repair and adjustment to ensure its normal operation and provide reliable disinfection protection for poultry farms.
[0082] Specifically, step S30 in the method includes:
[0083] The multidimensional disinfection assessment results include at least the ultraviolet dose assessment results, the fogging effect score, and the chemical residue confidence level;
[0084] The ultraviolet dose assessment result is compared with a preset dose safety factor to calculate and obtain the first assessment factor;
[0085] The fog effect score is compared with a preset fog effect threshold, and a second evaluation coefficient is calculated.
[0086] The chemical residue confidence level is compared with a preset residue confidence threshold, and a third evaluation Boolean value is defined based on the comparison result.
[0087] In this embodiment, the multidimensional disinfection assessment results first include ultraviolet (UV) dose assessment results, fogging effect score, and chemical residue confidence level. The UV dose assessment results are compared with a preset dose safety factor. If the UV dose assessment results are lower than the dose safety factor, it indicates that the UV disinfection effect may be insufficient. The first assessment coefficient calculated in this case reflects the degree of difference.
[0088] For example, if the dose safety factor is set to 1 and the ultraviolet dose assessment result is 0.8, then the first assessment factor may be 0.8. The lower the dose safety factor, the further the ultraviolet disinfection effect deviates from the safety standard.
[0089] Secondly, the misting effect score is compared with a preset misting effect threshold. The misting effect threshold is set based on the misting effect the air shower system should achieve during normal operation. If the misting effect score is lower than this threshold, it indicates that the misting effect is substandard. By comparing and calculating a second evaluation coefficient, the difference between the misting effect and the standard can be quantified. For example, if the preset misting effect threshold is 80 points, but the actual misting effect score is 60 points, the second evaluation coefficient might be 0.75, reflecting the quality of the misting effect.
[0090] Then, the chemical residue confidence level is compared with a preset residue confidence threshold. The chemical residue confidence level reflects the degree of confidence that the chemical residue is in a safe state. If the chemical residue confidence level is lower than the residue confidence threshold, it indicates that there is a high probability that the chemical residue is unsafe, and the defined third assessment Boolean value is 0, indicating unsafety; if it is higher than the threshold, it is defined as 1, indicating safety.
[0091] By comparing and calculating the above three aspects, the disinfection effect of the air shower system can be comprehensively evaluated from different dimensions, providing a strong basis for generating accurate graded early warning signals in combination with graded early warning rules, thereby timely detecting possible faults in the air shower system and ensuring the stable operation and disinfection effect of the air shower system in poultry farms.
[0092] The multi-dimensional operating parameters are input into the disinfection assessment model to obtain multi-dimensional disinfection assessment results, and then fused and analyzed in conjunction with preset graded early warning rules to obtain graded early warning signals. The method also includes:
[0093] Based on the aforementioned tiered early warning rules, a fusion analysis is performed to obtain tiered early warning signals, including:
[0094] If either the first evaluation coefficient or the second evaluation coefficient is less than 1, a first-level graded early warning signal is generated to warn of insufficient disinfection.
[0095] If both the first evaluation coefficient and the second evaluation coefficient are greater than or equal to 1, and the third evaluation Boolean value is 1, then a secondary-level early warning signal is generated to warn of residual risks.
[0096] In this embodiment, firstly, a level-one warning signal is generated based on the magnitude of the first evaluation coefficient and the second evaluation coefficient. When either the first evaluation coefficient or the second evaluation coefficient is less than 1, it indicates that the ultraviolet disinfection effect or the fogging effect has not met the safety standard, and there is insufficient disinfection.
[0097] For example, if the first evaluation coefficient is 0.9 and the second evaluation coefficient is 0.8, it indicates that the ultraviolet dose or fogging effect deviates from the safety standard. At this time, a first-level warning signal is generated. This signal will promptly remind the staff that there is a problem with the air shower system in terms of disinfection, and that the ultraviolet lamps or fogging equipment need to be inspected and maintained to improve the disinfection effect and ensure the safety of the poultry farm environment.
[0098] Secondly, if both the first and second evaluation coefficients are greater than or equal to 1, it indicates that the effects of ultraviolet disinfection and fogging are at a normal level. Then, a third evaluation Boolean value is used for judgment. If the third evaluation Boolean value is 1, it means that chemical residues are in a safe state, but there is still a certain risk of residue. In this case, a level-two warning signal will be generated, mainly reminding staff that although the current disinfection effect meets the standards and chemical residues are safe, attention still needs to be paid to the chemical residue situation. It may be necessary to fine-tune the operating parameters of the air shower system or optimize the use of disinfectant to further reduce the risk of residue and ensure the long-term stable operation of the air shower system and the biosecurity of the poultry farm.
[0099] Through the aforementioned tiered early warning mechanism, corresponding early warning signals are issued in a timely and accurate manner based on different assessment results, effectively ensuring the normal operation and disinfection effect of the air shower system in poultry farms, and creating a favorable environment for the healthy growth of poultry.
[0100] Furthermore, after obtaining the tiered early warning signal, it also includes:
[0101] Analyze the graded early warning signals to obtain the direction of parameter adjustment;
[0102] Based on the adjustment direction of the parameters, the spraying sequence control parameters of the disinfectant in the wind-mist spraying sequence parameters are iteratively optimized.
[0103] Based on the optimized spraying timing control parameters of the disinfectant, the airflow spraying timing control parameters in the wind-mist spraying timing parameters are iteratively optimized.
[0104] Simultaneously, the optimized spraying timing control parameters of the disinfectant and the airflow blowing timing control parameters are output as a fault control plan.
[0105] In this embodiment, the graded warning signals are first analyzed, and the direction of parameter adjustment is determined based on the fault type identifier. For example, if the graded warning signal shows "insufficient ultraviolet dose fault," the direction of parameter adjustment may be to increase the irradiation time of the ultraviolet lamp or increase its power; if it is "poor misting effect fault," it may be necessary to adjust parameters such as the spray concentration of the disinfectant, the spraying time, or the airflow speed; if it is "excessive chemical residue fault," it may be necessary to consider reducing the amount of disinfectant used or adjusting the spraying sequence.
[0106] Secondly, based on the determined parameter adjustment direction, the spraying timing control parameters of the disinfectant in the air-mist spray sequence parameters were iteratively optimized. During the optimization process, parameters such as the spray start time, spray duration, and spray interval of the disinfectant were gradually adjusted, taking into account the actual operation of the air shower system and previous experimental data. Simulation experiments or small-scale actual tests were used to observe the disinfection effect and chemical residue under different parameter combinations. The parameters were continuously adjusted based on the feedback results until a more ideal effect was achieved.
[0107] Then, after optimizing the spraying sequence control parameters of the disinfectant, the airflow spraying sequence control parameters in the air-mist spraying sequence parameters were iteratively optimized based on the optimized parameters. Parameters such as the start time, duration, and intensity of airflow spraying affect the diffusion and coverage effect of the mist spraying, as well as the removal of chemical residues. By adjusting the airflow spraying sequence control parameters to match the optimized disinfectant spraying sequence, the disinfection effect and safety of the air shower system are further improved.
[0108] Finally, the optimized spraying timing control parameters and airflow blowing timing control parameters of the disinfectant are output as a fault control plan.
[0109] This contingency plan provides operational guidance for operators of the air shower system. When a fault warning occurs in the air shower system, the parameters can be adjusted according to the plan to solve the problem in a timely manner, ensuring the stable operation of the air shower system and the biosecurity of the poultry farm.
[0110] S40: Feedback the graded early warning signal to execute fault early warning.
[0111] In this embodiment, after obtaining the graded early warning signal, the signal is fed back to execute a fault warning. The graded early warning signal is fed back to relevant personnel, such as by setting up an audible and visual alarm. When a first-level graded early warning signal is generated, a rapid alarm sound and flashing lights are emitted to remind on-site staff that there is a serious problem of insufficient disinfection in the air shower system, which needs to be dealt with immediately. When a second-level graded early warning signal is generated, a relatively soothing prompt sound and soft flashing lights are emitted to inform staff that there is a residual risk and that subsequent situations need to be monitored.
[0112] In summary, compared to existing technologies, this application utilizes orthogonal experiments to obtain training data for multi-dimensional operating parameters, and conducts supervised training on the disinfection assessment model and chemical residue assessment channel, respectively, enabling accurate analysis of the disinfection effect and chemical residue status of the air shower system. Furthermore, by combining preset graded early warning rules, different levels of graded early warning signals are generated to clarify the fault type, providing direction for subsequent parameter adjustments and fault handling.
[0113] In summary, the embodiments of this application have at least the following technical effects:
[0114] This application provides a deep learning-driven method for early warning of faults in poultry farm air shower systems. First, it collects multi-dimensional operating parameters of the air shower system in real time, comprehensively and accurately reflecting its actual operating status, overcoming the shortcomings of existing methods that rely on single-parameter monitoring. Second, it constructs a deep learning-driven disinfection assessment model, which includes ultraviolet dose assessment channels, misting effect assessment channels, and chemical residue assessment channels. This allows for evaluation of the disinfection effect of the air shower system from different perspectives. Then, it inputs the multi-dimensional operating parameters into the disinfection assessment model to obtain multi-dimensional disinfection assessment results, which are then fused and analyzed using preset graded early warning rules to obtain graded early warning signals. This comprehensive analysis effectively identifies potential faults in the air shower system and generates corresponding graded early warning signals based on different assessment results, enabling farm managers to understand the system's operating status in a timely manner. Finally, it feeds back the graded early warning signals to execute fault warnings, analyzes the graded early warning signals to obtain parameter adjustment directions, iteratively optimizes the air-mist timing parameters, and outputs a fault control plan. Through the above technical solutions, this application achieves timely detection, accurate early warning, and effective control of air shower system malfunctions, ensuring the stable operation of the air shower system and thus providing strong protection for poultry health and the safety of the breeding environment. Deep learning improves the management efficiency and disinfection effect of the air shower system in poultry farms, promoting the intelligent and scientific development of the poultry farming industry.
[0115] Example 2, as Figure 2 As shown, based on the same inventive concept as the deep learning-driven fault early warning method for poultry farm air shower systems provided in Embodiment 1, this application also provides a deep learning-driven fault early warning system for poultry farm air shower systems, including:
[0116] The parameter acquisition module 11 is used to acquire multi-dimensional operating parameters of the air shower system in real time. The multi-dimensional operating parameters include at least the ultraviolet lamp operating parameters, the air-mist shower timing parameters, and the environmental state parameters.
[0117] The model building module 12 is used to build a deep learning-driven disinfection assessment model, wherein the disinfection assessment model includes at least an ultraviolet dose assessment channel, a fogging effect assessment channel, and a chemical residue assessment channel.
[0118] The result evaluation module 13 is used to input the multi-dimensional operating parameters into the disinfection evaluation model, obtain the multi-dimensional disinfection evaluation results, and perform fusion analysis in combination with the preset graded early warning rules to obtain graded early warning signals, wherein the graded early warning signals include fault type identifiers.
[0119] The signal feedback module 14 is used to provide feedback on the graded early warning signal to perform fault early warning.
[0120] Furthermore, in one embodiment of the application, the operating parameters of the ultraviolet lamp include at least the ultraviolet radiation intensity, the cumulative irradiation time of the ultraviolet lamp tube, and the rated irradiation time;
[0121] The wind-mist timing parameters include at least disinfectant spray timing control parameters and airflow blowing timing control parameters;
[0122] The environmental state parameters include at least ambient temperature and ambient humidity.
[0123] Furthermore, in one embodiment of the application, a deep learning-driven disinfection assessment model is constructed, wherein the construction process of the ultraviolet dose assessment channel includes:
[0124] Acquire intrinsic ultraviolet (UV) sample data of the target scene, and extract behavioral features based on the intrinsic UV sample data to obtain UV usage behavior features;
[0125] Using the aforementioned ultraviolet (UV) usage behavior characteristics as an invariant constraint, and combining the tolerance constraint defined based on the device stability parameters of the target UV device, homogeneous sample collection is performed to obtain UV homogeneous sample data.
[0126] Based on the intrinsic ultraviolet sample data and the homologous ultraviolet sample data, a training method is used to obtain an ultraviolet intensity baseline, wherein the ultraviolet intensity baseline takes the cumulative irradiation time as input and outputs the real-time ultraviolet intensity.
[0127] The product of the ultraviolet intensity baseline and the rated irradiation time is defined as the real-time ultraviolet dose, and the ratio of the real-time ultraviolet dose to the rated ultraviolet dose of the target scene is defined as the ultraviolet dose assessment result, thus constructing the ultraviolet dose assessment channel.
[0128] Furthermore, in one embodiment of the application, a deep learning-driven disinfection evaluation model is constructed, wherein the construction process of the fogging effect evaluation channel includes:
[0129] Based on the historical disinfection records of the target scene, determine the most unfavorable fogging conditions, including the most unfavorable object conditions and the most unfavorable environmental conditions.
[0130] Construct a fog simulation analysis model, execute the simulation analysis, and obtain the simulation analysis results;
[0131] Based on the simulation analysis results, fog evaluation indicators are calculated and feature layer fusion is performed accordingly to obtain the fog effect evaluation value of the sample. The fog evaluation indicators include at least coverage, the most unfavorable effective wetting time and the deposition amount per unit area.
[0132] Randomly extract from the simulation analysis results, and calculate the sample entropy based on the sample fog effect evaluation value based on the extracted results;
[0133] The simulation analysis and extraction are performed iteratively until the sample entropy and sample quantity both meet the preset binary sample constraints. The extraction results are used as training data to construct and supervise the training of the disinfection evaluation model based on deep learning.
[0134] Furthermore, a deep learning-driven disinfection assessment model is constructed, wherein the construction process of the chemical residue assessment channel includes:
[0135] Based on prior knowledge, obtain the safety residue constraints corresponding to the disinfectant preparations;
[0136] Based on the safety residue constraints, an orthogonal experimental scheme is planned in conjunction with the air shower parameter space of the target air shower system, and the orthogonal experimental scheme is executed to obtain chemical residue training data, wherein the chemical residue training data is marked with safety Boolean values;
[0137] Based on the chemical residue training data, a chemical residue evaluation channel based on an LSTM network is trained under supervision.
[0138] In one embodiment, the result evaluation module 13 is specifically used for:
[0139] The multidimensional disinfection assessment results include at least the ultraviolet dose assessment results, the fogging effect score, and the chemical residue confidence level;
[0140] The ultraviolet dose assessment result is compared with a preset dose safety factor to calculate and obtain the first assessment factor;
[0141] The fog effect score is compared with a preset fog effect threshold, and a second evaluation coefficient is calculated.
[0142] The chemical residue confidence level is compared with a preset residue confidence threshold, and a third evaluation Boolean value is defined based on the comparison result.
[0143] Furthermore, in one embodiment, the multi-dimensional operating parameters are input into the disinfection assessment model to obtain multi-dimensional disinfection assessment results, and then fused and analyzed in conjunction with preset graded early warning rules to obtain graded early warning signals, further comprising:
[0144] Based on the aforementioned tiered early warning rules, a fusion analysis is performed to obtain tiered early warning signals, including:
[0145] If either the first evaluation coefficient or the second evaluation coefficient is less than 1, a first-level graded early warning signal is generated to warn of insufficient disinfection.
[0146] If both the first evaluation coefficient and the second evaluation coefficient are greater than or equal to 1, and the third evaluation Boolean value is 1, then a secondary-level early warning signal is generated to warn of residual risks.
[0147] Furthermore, after obtaining the tiered early warning signal, it also includes:
[0148] Analyze the graded early warning signals to obtain the direction of parameter adjustment;
[0149] Based on the adjustment direction of the parameters, the spraying sequence control parameters of the disinfectant in the wind-mist spraying sequence parameters are iteratively optimized.
[0150] Based on the optimized spraying timing control parameters of the disinfectant, the airflow spraying timing control parameters in the wind-mist spraying timing parameters are iteratively optimized.
[0151] Simultaneously, the optimized spraying timing control parameters of the disinfectant and the airflow blowing timing control parameters are output as a fault control plan.
[0152] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.
[0153] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
[0154] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.
Claims
1. A deep learning-driven method for early warning of faults in air shower systems of poultry farms, characterized in that, include: Real-time acquisition of multi-dimensional operating parameters of the air shower system, wherein the multi-dimensional operating parameters include at least ultraviolet lamp operating parameters, air-mist shower timing parameters, and environmental state parameters; Construct a deep learning-driven disinfection assessment model, wherein the disinfection assessment model includes at least an ultraviolet dose assessment channel, a fogging effect assessment channel, and a chemical residue assessment channel; The multi-dimensional operating parameters are input into the disinfection assessment model to obtain multi-dimensional disinfection assessment results, and then combined with preset hierarchical early warning rules for fusion analysis to obtain hierarchical early warning signals, wherein the hierarchical early warning signals include fault type identifiers. The feedback of the graded early warning signal will trigger a fault warning.
2. The deep learning-driven fault early warning method for air shower systems in poultry farms as described in claim 1, characterized in that, The operating parameters of the ultraviolet lamp include at least the ultraviolet radiation intensity, the cumulative irradiation time of the ultraviolet lamp tube, and the rated irradiation time. The wind-mist timing parameters include at least disinfectant spray timing control parameters and airflow blowing timing control parameters; The environmental state parameters include at least ambient temperature and ambient humidity.
3. The deep learning-driven fault early warning method for air shower systems in poultry farms as described in claim 1, characterized in that, Constructing a deep learning-driven disinfection assessment model, wherein the construction process of the ultraviolet dose assessment channel includes: Acquire intrinsic ultraviolet (UV) sample data of the target scene, and extract behavioral features based on the intrinsic UV sample data to obtain UV usage behavior features; Using the aforementioned ultraviolet (UV) usage behavior characteristics as an invariant constraint, and combining the tolerance constraint defined based on the device stability parameters of the target UV device, homogeneous sample collection is performed to obtain UV homogeneous sample data. Based on the intrinsic ultraviolet sample data and the homologous ultraviolet sample data, a training method is used to obtain an ultraviolet intensity baseline, wherein the ultraviolet intensity baseline takes the cumulative irradiation time as input and outputs the real-time ultraviolet intensity. The product of the ultraviolet intensity baseline and the rated irradiation time is defined as the real-time ultraviolet dose, and the ratio of the real-time ultraviolet dose to the rated ultraviolet dose of the target scene is defined as the ultraviolet dose assessment result, thus constructing the ultraviolet dose assessment channel.
4. The deep learning-driven fault early warning method for air shower systems in poultry farms as described in claim 1, characterized in that, Constructing a deep learning-driven disinfection evaluation model, wherein the construction process of the fogging effect evaluation channel includes: Based on the historical disinfection records of the target scene, determine the most unfavorable fogging conditions, including the most unfavorable object conditions and the most unfavorable environmental conditions. Construct a fog simulation analysis model, execute the simulation analysis, and obtain the simulation analysis results; Based on the simulation analysis results, fog evaluation indicators are calculated and feature layer fusion is performed accordingly to obtain the fog effect evaluation value of the sample. The fog evaluation indicators include at least coverage, the most unfavorable effective wetting time and the deposition amount per unit area. Randomly extract from the simulation analysis results, and calculate the sample entropy based on the sample fog effect evaluation value based on the extracted results; The simulation analysis and extraction are performed iteratively until the sample entropy and sample quantity both meet the preset binary sample constraints. The extraction results are used as training data to construct and supervise the training of the disinfection evaluation model based on deep learning.
5. The deep learning-driven fault early warning method for air shower systems in poultry farms as described in claim 1, characterized in that, Constructing a deep learning-driven pest control assessment model, wherein the construction process of the chemical residue assessment channel includes: Based on prior knowledge, obtain the safety residue constraints corresponding to the disinfectant preparations; Based on the safety residue constraints, an orthogonal experimental scheme is planned in conjunction with the air shower parameter space of the target air shower system, and the orthogonal experimental scheme is executed to obtain chemical residue training data, wherein the chemical residue training data is marked with safety Boolean values; Based on the chemical residue training data, a chemical residue evaluation channel based on an LSTM network is trained under supervision.
6. The deep learning-driven fault early warning method for air shower systems in poultry farms as described in claim 1, characterized in that, The multi-dimensional operating parameters are input into the disinfection assessment model to obtain multi-dimensional disinfection assessment results. These results are then fused and analyzed in conjunction with preset graded early warning rules to obtain graded early warning signals, including: The multidimensional disinfection assessment results include at least the ultraviolet dose assessment results, the fogging effect score, and the chemical residue confidence level; The ultraviolet dose assessment result is compared with a preset dose safety factor to calculate and obtain the first assessment factor; The fog effect score is compared with a preset fog effect threshold, and a second evaluation coefficient is calculated. The chemical residue confidence level is compared with a preset residue confidence threshold, and a third evaluation Boolean value is defined based on the comparison result.
7. The deep learning-driven fault early warning method for air shower systems in poultry farms as described in claim 6, characterized in that, The multi-dimensional operating parameters are input into the disinfection assessment model to obtain multi-dimensional disinfection assessment results, and then fused and analyzed in conjunction with preset graded early warning rules to obtain graded early warning signals. The method also includes: Based on the aforementioned tiered early warning rules, a fusion analysis is performed to obtain tiered early warning signals, including: If either the first evaluation coefficient or the second evaluation coefficient is less than 1, a first-level graded early warning signal is generated to warn of insufficient disinfection. If both the first evaluation coefficient and the second evaluation coefficient are greater than or equal to 1, and the third evaluation Boolean value is 1, then a secondary-level early warning signal is generated to warn of residual risks.
8. The deep learning-driven fault early warning method for air shower systems in poultry farms as described in claim 1, characterized in that, After obtaining the graded early warning signal, it also includes: Analyze the graded early warning signals to obtain the direction of parameter adjustment; Based on the adjustment direction of the parameters, the spraying sequence control parameters of the disinfectant in the wind-mist spraying sequence parameters are iteratively optimized. Based on the optimized spraying timing control parameters of the disinfectant, the airflow spraying timing control parameters in the wind-mist spraying timing parameters are iteratively optimized. Simultaneously, the optimized spraying timing control parameters of the disinfectant and the airflow blowing timing control parameters are output as a fault control plan.
9. A deep learning-driven fault early warning system for air shower systems in poultry farms, characterized in that: A method for performing a deep learning-driven fault early warning system for a poultry farm air shower system as described in any one of claims 1-8 includes: The parameter acquisition module is used to acquire multi-dimensional operating parameters of the air shower system in real time. The multi-dimensional operating parameters include at least the ultraviolet lamp operating parameters, the air-mist shower timing parameters, and the environmental state parameters. The model building module is used to build a deep learning-driven disinfection assessment model, wherein the disinfection assessment model includes at least an ultraviolet dose assessment channel, a fogging effect assessment channel, and a chemical residue assessment channel. The result evaluation module is used to input the multi-dimensional operating parameters into the disinfection evaluation model, obtain multi-dimensional disinfection evaluation results, and perform fusion analysis in combination with preset hierarchical early warning rules to obtain hierarchical early warning signals, wherein the hierarchical early warning signals include fault type identifiers. The signal feedback module is used to provide feedback on the graded early warning signal to perform fault early warning.