Kitchen waste pretreatment spraying deodorization regulation and control method based on multi-mode sensing fusion

Abstract

The invention relates to the technical field of program control systems, in particular to a kitchen waste pretreatment spraying deodorization regulation and control method based on multi-mode sensing fusion, which comprises the following steps: acquiring multi-sensing data containing odor concentration in a plurality of sub-control areas; calling a working condition identification program to determine a current preprocessing state; and the control system calls matched spraying control parameters from a control strategy library according to the current state, executes an overall spraying control program and obtains control feedback data. When the feedback data meet preset local control conditions, local or linkage spraying control is executed according to program logic; and in combination with multiple control feedback results, updating a control parameter combination. According to the invention, on the basis of'integral-local-integral 'multi-region cooperative control, comprehensive inhibition of peculiar smell spatial diffusion in the whole kitchen waste pretreatment process and optimal configuration of deodorant and energy consumption are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of program control systems, and in particular to a method for controlling the pretreatment spraying and deodorization of kitchen waste based on multimodal sensor fusion. Background Art

[0002] The pretreatment of food waste easily produces a large amount of foul-smelling gases, requiring efficient control through a spray deodorization system. Current systems often rely on timed program control methods based on programmable logic controllers (PLCs) and other controllers, which periodically trigger the spraying process. These systems lack the ability to control the actual odor level, leading to over- and under-spraying, which impacts resource efficiency and deodorization effectiveness.

[0003] In recent years, some systems have introduced sensors such as gas concentration, temperature and humidity, and attempted to build feedback control logic based on threshold judgment. However, the control program structure is still relatively simple, lacking the fusion of multi-source environmental information and intelligent adjustment mechanism, making it difficult to adapt to the complex and changeable working conditions during food waste treatment.

[0004] Furthermore, the pretreatment process often consists of multiple functional areas, each with significantly different odor characteristics and control requirements. Traditional spray programs often use a unified control model, without incorporating regionally differentiated responses and linkage mechanisms into the control logic, making it impossible to achieve intelligent, coordinated "local-global" control.

[0005] With the improvement of data collection capabilities, a large amount of odor data, process status and equipment parameters can be collected during the preprocessing process. However, the existing control system fails to use industrial data processing technology to effectively process and utilize this big data. The control program is still mainly rule-triggered and lacks dynamic optimization capabilities.

[0006] Therefore, how to invent a spray deodorization control method that is suitable for various food waste pretreatment process conditions, which can not only realize accurate response control of odor at local workstations based on the fusion perception results of multi-modal sensors, but also dynamically adjust the overall spray parameters through multi-region collaborative strategies to achieve comprehensive suppression of odor spatial diffusion and optimal configuration of deodorants and energy consumption, thereby achieving the "whole-local-whole" collaborative intelligent deodorization control goal in the entire pretreatment process.

[0007] To this end, a method for spraying and deodorizing food waste pretreatment based on multimodal sensor fusion was proposed. Summary of the Invention

[0008] The purpose of the present invention is to provide a method for controlling the spray deodorization of food waste pretreatment based on multimodal sensor fusion. Based on the multi-region collaborative control of "whole-local-whole", it realizes the comprehensive suppression of the spatial diffusion of odor in the whole process of food waste pretreatment and the optimal configuration of deodorant and energy consumption.

[0009] To achieve the above object, the present invention provides the following technical solutions: The method for controlling the pretreatment spraying and deodorization of kitchen waste based on multimodal sensor fusion includes: The pretreatment area is divided into N sub-control areas, and a multimodal sensor is set in each sub-control area to collect multimodal sensing data, including odor concentration data; The current pre-treatment working condition is identified based on the multimodal sensing data using a working condition recognition model; according to the pre-treatment working condition, the control system automatically matches a combination of spray control parameters from a control strategy library, executes an overall control program, and obtains an overall control result, including an overall odor index and an overall disturbance index; If the overall odor index satisfies the first local control condition, a local encryption control program is executed according to the odor concentration data of each sub-control area to obtain a first local control result; If the overall disturbance index satisfies the second local control condition, the local linkage control program is executed through the regional linkage control model to obtain the second local control result; The control parameters are regulated by using the overall optimization model in combination with the overall control result, the first local control result and the second local control result.

[0010] Preferably, the multimodal sensor includes: an odor concentration sensor, a temperature and humidity sensor, a vibration sensor, a motor current sensor and an airflow velocity sensor; the multimodal sensing data also includes: airflow velocity, equipment vibration data, temperature, humidity and motor operating current data.

[0011] Preferably, the operating state recognition model includes: an operating state feature extraction unit, an operating state determination unit and an operating state conversion detection unit; The operating condition feature extraction unit includes a basic physical feature extraction subunit, an operating condition event feature construction subunit and an operating condition feature generation subunit, which are used to extract operating condition features based on the multimodal sensing data; the operating condition features are double-layer features based on "physics-event", including basic physical features and operating condition event features; the basic physical feature extraction subunit extracts the basic physical features that characterize the basic operating status of the equipment from the multimodal sensing data, including time domain statistical features, frequency domain features and wavelet energy features; the operating condition event feature construction subunit maps each basic physical feature to the operating condition event feature based on the expert knowledge base; the operating condition feature generation subunit performs weighted fusion of the basic physical features and the operating condition event features to obtain the operating condition features; the operating condition state discrimination unit uses the operating condition features to identify the current pre-processing operating condition state; the operating condition conversion detection unit identifies the operating condition switching moment based on the pre-processing operating condition state, generates a transition period control parameter combination, and stores it in the control strategy library.

[0012] Preferably, the overall control result also includes: overall spray coverage, overall deodorant consumption and overall system energy consumption index; the control strategy library contains M preset control parameter combinations and transition period control parameter combinations; the preset control parameter combinations include nozzle type, nozzle opening and closing matrix, spray pressure level, spray duration and spray interval period.

[0013] Preferably, the overall odor index satisfies the first local control condition, and the specific judgment process is: when the overall odor index exceeds a preset overall odor threshold and / or the odor concentration data of any sub-control area exceeds a local odor threshold, triggering local encryption control; A local encrypted control program is executed according to the odor concentration data of each sub-control area to obtain a first local control result. The specific process includes: calculating the odor concentration gradient of each sub-control area according to the odor concentration data of each sub-control area, marking the sub-control area where the odor concentration gradient is higher than the odor concentration gradient threshold as a key control area, and performing local encrypted control on the key control area; obtaining the first local odor concentration, the first local spray coverage, the first local deodorant consumption and the first local energy consumption of each control sub-area after local encrypted control, and fusing them respectively to obtain the first local control result.

[0014] Preferably, the overall disturbance index satisfies the second local control condition, and the specific judgment process is: based on the odor concentration data of each sub-control area and the overall disturbance index, the odor concentration difference and the odor diffusion rate between adjacent sub-control areas are calculated, and when the odor concentration difference exceeds the odor concentration difference threshold and the odor diffusion rate exceeds the odor diffusion rate threshold, the local linkage control is triggered; The regional linkage control model includes: an odor diffusion trend prediction unit, a linkage area selection unit, a linkage parameter dynamic adjustment unit, and a second local control result output unit; The odor diffusion trend prediction unit predicts the potential odor diffusion path based on the odor diffusion rate and diffusion direction; the linkage area selection unit automatically identifies and determines the linkage area according to the odor diffusion path, and generates a linkage response area list; the linkage parameter dynamic adjustment unit dynamically adjusts the spray linkage parameters for the sub-control areas in the linkage response area list, based on the original spray parameters, according to the distance between each sub-control area in the linkage response area list and the odor source area, the predicted odor arrival time and the odor concentration data, to perform local linkage control; the second local control result output unit obtains the second local odor concentration, the second local spray coverage, the second local deodorant consumption and the second local energy consumption of each control sub-area after the local linkage control, and integrates them respectively to obtain the second local control result.

[0015] Preferably, the overall optimization model includes: a multi-objective function construction unit, a constraint setting unit and a control parameter optimization unit; the multi-objective function construction unit establishes a multi-objective function of maximizing deodorization effect, minimizing deodorant consumption, minimizing energy consumption and minimizing equipment load; the constraint setting unit sets the lower limit of spray coverage, the upper limit of deodorant consumption and the system response time constraint; the control parameter optimization unit solves the optimal control parameter combination through the particle swarm optimization algorithm.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention integrates multimodal sensor data such as odor, temperature and humidity, equipment vibration, motor current and airflow, and constructs a working condition identification model based on the "physical-event" dual-layer characteristics, which provides a deep insight into the real-time status of food waste pretreatment. It can map the basic physical characteristics of equipment operation to specific working condition event characteristics, thereby predicting the source, intensity and trend of odor. This control logic based on process understanding can automatically match the spray parameters that are most suitable for the current working conditions from the control strategy library, realizing the transition from extensive management to adaptive refined control, greatly improving the foresight and accuracy of deodorization regulation, and avoiding the problems of response lag or excessive spraying.

[0017] 2. The present invention designs a hierarchical collaborative control architecture of "overall spraying-local encryption-local linkage", which significantly improves the flexibility and deodorization capacity of spray deodorization control. On the basis of overall control, when the local odor concentration or its gradient is abnormal, local encryption control is started, and only the odor source or the key control area with heavy pollution is strengthened by spraying, thus avoiding waste of resources. At the same time, when it is monitored that there is a risk of odor diffusion between regions, the diffusion path is predicted through the local linkage control model, and the linkage response area that may be affected downstream is forward-lookingly pre-processed. This strategy of combining "precision strike" with "forward-looking interception" effectively curbs the local outbreak and regional spread of odor, and improves the deodorization efficiency and effect.

[0018] 3. The present invention not only focuses on the deodorization effect, but also pursues the maximization of the comprehensive benefits of the overall operation of the system. By establishing a multi-objective optimization model that includes maximizing the deodorization effect, minimizing the deodorant consumption, minimizing the energy consumption and minimizing the equipment load, and combining it with the actual working condition constraints to perform global optimization. This multi-objective optimization model can find the best control parameter combination that takes into account the effect, cost and energy consumption among multiple conflicting objectives. It ensures that the system significantly reduces the consumption of deodorant and electricity while achieving excellent deodorization performance, reduces equipment loss, and realizes the unity of environmental benefits, economic benefits and sustainable equipment operation, and has extremely high practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic flow chart of a method for controlling kitchen waste pretreatment, spraying, and deodorization based on multimodal sensing fusion according to an embodiment of the present invention; Figure 2 A schematic diagram of the process of spraying and deodorizing kitchen waste pretreatment according to an embodiment of the present invention; Figure 3 A schematic diagram of the structure of the operating state recognition model provided by an embodiment of the present invention; Figure 4 A schematic diagram of the structure of the overall optimization model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] This invention proposes a method for controlling spray deodorization during food waste pretreatment based on multimodal sensor fusion. This method, through "whole-local-whole" multi-region collaborative control, achieves comprehensive suppression of odor diffusion throughout the food waste pretreatment process, as well as optimal allocation of deodorant and energy consumption. To demonstrate the effectiveness of this method in achieving the goal of "whole-local-whole" collaborative intelligent deodorization control throughout the pretreatment process, the following two examples illustrate its effectiveness.

[0022] Example 1: This embodiment is applied to a food waste pretreatment workshop at a waste treatment plant. This workshop processes 200 tons of food waste daily. The pretreatment area, covering approximately 3,000 square meters, includes multiple functional areas, including receiving, sorting, crushing, and dehydration. Due to the complex composition of food waste, the pretreatment process produces a large amount of foul-smelling gases, seriously impacting the work environment and the lives of surrounding residents.

[0023] In the examples of the present application, the method proposed in the present invention is used to describe in detail the process of achieving the "whole-part-whole" collaborative intelligent deodorization control goal in the entire pretreatment process. Figure 1This is a specific flow chart of the method of the present invention, including: dividing the pretreatment area into N sub-control areas, setting a multimodal sensor in each sub-control area, collecting multimodal sensing data, including odor concentration data; using an operating condition recognition model, identifying the current pretreatment operating condition based on the multimodal sensing data; according to the pretreatment operating condition, the control system automatically matches the spray control parameter combination from the control strategy library, executes the overall control program, and obtains an overall control result, including an overall odor index and an overall disturbance index; if the overall odor index meets the first local control condition, then the local encryption control program is executed according to the odor concentration data of each sub-control area to obtain a first local control result; if the overall disturbance index meets the second local control condition, then the local linkage control program is executed through the regional linkage control model to obtain a second local control result; and the control parameters are regulated by the overall optimization model in combination with the overall control result, the first local control result, and the second local control result. Figure 2 This is a flow chart of the spray deodorization control process for pre-treatment of kitchen waste. Figure 1 and Figure 2 The following content is described: The pretreatment area is divided into N sub-control areas, and a multimodal sensor is set in each sub-control area to collect multimodal sensing data, including odor concentration data; The multimodal sensor includes: an odor concentration sensor, a temperature and humidity sensor, a vibration sensor, a motor current sensor and an airflow velocity sensor; the multimodal sensing data also includes: airflow velocity, equipment vibration data, temperature, humidity and motor operating current data.

[0024] Specifically, according to the actual layout and functional zoning of the waste treatment center, the 3,000-square-meter pretreatment area is divided into 12 sub-control areas (N=12), including: 2 sub-control areas (A1 and A2) in the receiving area, 3 sub-control areas (B1, B2, and B3) in the sorting area, 3 sub-control areas (C1, C2, and C3) in the crushing area, 2 sub-control areas (D1 and D2) in the dehydration area, and 2 sub-control areas (E1 and E2) in the conveying channel.

[0025] Multimodal sensors are deployed within each sub-control area. For example, in the crushing area C1, four odor concentration sensors are installed, located at the crusher feed and discharge ports, the operating platform, and the center of the area; two temperature and humidity sensors; one vibration sensor, located on the crusher body; one motor current sensor, which monitors the crusher motor's operating status; and two air velocity sensors, one at the air inlet and one at the exhaust.

[0026] This embodiment uses multimodal sensors to collect multimodal sensor data, achieving comprehensive and in-depth awareness of the production environment and equipment status. This provides a rich, high-dimensional data foundation for subsequent precise identification of operating conditions and prediction of diffusion trends. This system not only focuses on the odor itself but also incorporates the behavior of the equipment that generates the odor (vibration, current) and the environmental factors that influence odor diffusion (temperature, humidity, and airflow). This allows the system to move beyond simply reacting to odors and instead gain a deeper understanding of the specific sources that cause them, significantly improving the accuracy and foresight of decision-making.

[0027] Further, refer to Figure 3 , identifying the current pre-processing operating state based on the multimodal sensing data through an operating state recognition model; The working condition recognition model includes: a working condition feature extraction unit, a working condition determination unit and a working condition conversion detection unit; The operating condition feature extraction unit includes a basic physical feature extraction subunit, an operating condition event feature construction subunit and an operating condition feature generation subunit, which are used to extract operating condition features based on the multimodal sensing data; the operating condition features are double-layer features based on "physics-event", including basic physical features and operating condition event features; the basic physical feature extraction subunit extracts the basic physical features that characterize the basic operating status of the equipment from the multimodal sensing data, including time domain statistical features, frequency domain features and wavelet energy features; the operating condition event feature construction subunit maps each basic physical feature to the operating condition event feature based on the expert knowledge base; the operating condition feature generation subunit performs weighted fusion of the basic physical features and the operating condition event features to obtain the operating condition features; the operating condition state discrimination unit uses the operating condition features to identify the current pre-processing operating condition state; the operating condition conversion detection unit identifies the operating condition switching moment based on the pre-processing operating condition state, generates a transition period control parameter combination, and stores it in the control strategy library.

[0028] Specifically, the time domain statistical features include the mean, variance, peak value and kurtosis of the odor concentration, which reflect the overall level and fluctuation of pollutants; the frequency domain features extract the main frequency and spectrum energy distribution by performing fast Fourier transform on the vibration signal, and identify the equipment operation mode; the wavelet energy features use the db4 wavelet basis to perform a 5-layer decomposition of the motor current signal through wavelet transform, extract the energy proportion of each frequency band, and judge the equipment load status.

[0029] For example, the simultaneous occurrence of two physical characteristics, "motor current surge" and "high-frequency vibration energy concentration", is constructed as a "hard material crushing event" feature; and "motor current continues to run steadily at a high level" is constructed as a "stable pressing event" feature. The working condition feature generation subunit uses an attention mechanism to perform weighted fusion of the basic physical features and the working condition event features; The working condition identification unit uses a machine learning regression classifier pre-trained based on historical data to accurately identify the overall pretreatment working condition of the current workshop, such as the unloading peak period, the crusher fully loaded operation, the press stable operation or the equipment idle period.

[0030] This embodiment converts the raw sensor data into working condition intelligence with clear business logic through the working condition recognition model. By constructing the "physical-event" dual-layer features, it is possible to deeply understand the current detailed working condition, such as crushing hard materials or equipment idleness, rather than simply judging whether it is turned on or off. This deep insight into the working condition can predict the intensity and type of odor generation based on events, thereby performing predictive and precise regulation. At the same time, the setting of the working condition conversion detection unit ensures the smooth transition of the control strategy when switching working conditions, and improves the stability of subsequent overall-local deodorization regulation.

[0031] Furthermore, according to the pretreatment working condition, the control system automatically matches the spray control parameter combination from the control strategy library, executes the overall control program, and obtains the overall control result, including the overall odor index and the overall disturbance index; The overall control result also includes: overall spray coverage, overall deodorant consumption and overall system energy consumption index; the control strategy library contains M preset control parameter combinations and transition period control parameter combinations; the preset control parameter combinations include nozzle type, nozzle opening and closing matrix, spray pressure level, spray duration and spray interval period.

[0032] This implementation specifies the composition of overall control results and the content of the control strategy library, providing clear, quantifiable execution objectives and action menus for the entire control system. By transforming deodorization performance into monitoring of specific indicators such as spray coverage, deodorant consumption, and system energy consumption, the control effect becomes quantifiable. Furthermore, detailed control parameter combinations, including nozzles, pressure, and time, are defined, providing clear control targets and adjustment dimensions for subsequent optimization algorithms.

[0033] Furthermore, if the overall odor index satisfies the first local control condition, a local encryption control program is executed according to the odor concentration data of each sub-control area to obtain a first local control result; The overall odor index satisfies the first local control condition, and the specific judgment process is as follows: when the overall odor index exceeds a preset overall odor threshold and / or the odor concentration data of any sub-control area exceeds a local odor threshold, triggering local encryption control; A local encrypted control program is executed according to the odor concentration data of each sub-control area to obtain a first local control result. The specific process includes: calculating the odor concentration gradient of each sub-control area according to the odor concentration data of each sub-control area, marking the sub-control area where the odor concentration gradient is higher than the odor concentration gradient threshold as a key control area, and performing local encrypted control on the key control area; obtaining the first local odor concentration, the first local spray coverage, the first local deodorant consumption and the first local energy consumption of each control sub-area after local encrypted control, and fusing them respectively to obtain the first local control result.

[0034] Specifically, local encryption control includes: increasing the spraying frequency, extending the spraying duration and increasing the number of open nozzles.

[0035] This example describes the triggering and execution mechanism for localized intensive control, enhancing the ability to quickly respond to and precisely control localized, sudden odor sources. By incorporating odor concentration gradients as a basis for judgment, it can more sensitively detect sudden changes in odor, thereby identifying the most severely polluted areas as key areas for intensive treatment at the earliest stages of a problem. This avoids the waste of resources caused by initiating large-scale, global spraying due to localized issues, achieving efficient and economical targeted removal.

[0036] Furthermore, if the overall disturbance index satisfies the second local control condition, the local linkage control program is executed through the regional linkage control model to obtain the second local control result; The overall disturbance index satisfies the second local control condition. The specific judgment process is as follows: based on the odor concentration data of each sub-control area and the overall disturbance index, the odor concentration difference and the odor diffusion rate between adjacent sub-control areas are calculated. When the odor concentration difference exceeds the odor concentration difference threshold and the odor diffusion rate exceeds the odor diffusion rate threshold, the local linkage control is triggered; The regional linkage control model includes: an odor diffusion trend prediction unit, a linkage area selection unit, a linkage parameter dynamic adjustment unit, and a second local control result output unit; The odor diffusion trend prediction unit predicts a potential odor diffusion path based on the odor diffusion rate and diffusion direction; the linkage area selection unit automatically identifies and determines a linkage area according to the odor diffusion path, and generates a linkage response area list; the linkage parameter dynamic adjustment unit, for a sub-control area in the linkage response area list, dynamically adjusts a spray linkage parameter on the basis of original spray parameters, according to the distance between each sub-control area in the linkage response area list and the odor source area, the predicted odor arrival time and the odor concentration data, and performs local linkage control; and the second local control result output unit obtains the second local odor concentration, the second local spray coverage, the second local deodorant consumption and the second local energy consumption of each control sub-area after local linkage control, and respectively fuses them to obtain the second local control result.

[0037] Specifically, the odor diffusion trend prediction unit predicts an odor diffusion path based on a fluid mechanics model in combination with the measured air flow velocity and diffusion direction; The predicted odor arrival time is calculated by the odor diffusion rate and the distance between the sub-control area and the odor source area; The dynamic adjustment of the spray linkage parameter performs local linkage control. For example, for the D1 area, preventive adjustment is performed on the basis of the original spray parameters: the spray is started 10 s in advance (the odor arrival time is 30 s), the spray pressure is increased, the spray angle is adjusted to 45 degrees towards the diffusion direction to form an odor interception barrier. For the E1 area, a relatively mild preventive measure is adopted, and the spray interval is only shortened from 180 s to 120 s.

[0038] The embodiment realizes the prospective interception and preventive control of cross-regional odor diffusion through local linkage control. Compared with the traditional method of passively waiting for odor diffusion to the downstream area for processing, the local linkage control of the present application identifies potential pollution paths and affected areas in advance through an odor diffusion trend prediction model, and starts linkage spraying in these areas. This preventive strategy can effectively prevent local pollution from evolving into widespread spread, greatly improving the overall deodorization effect and control ability in the spatial dimension.

[0039] Further, the control parameters are regulated by using a global optimization model in combination with the global control result, the first local control result and the second local control result.

[0040] Reference Figure 4The overall optimization model includes: a multi-objective function construction unit, a constraint setting unit and a control parameter optimization unit; the multi-objective function construction unit establishes a multi-objective function of maximizing deodorization effect, minimizing deodorant consumption, minimizing energy consumption and minimizing equipment load; the constraint setting unit sets the lower limit of spray coverage, the upper limit of deodorant consumption and the system response time constraint; the control parameter optimization unit solves the optimal control parameter combination through the particle swarm optimization algorithm.

[0041] Specifically, the particle swarm optimization algorithm uses historical deodorization result data for pre-training and optimization, and ultimately obtains the optimal control parameter combination.

[0042] This embodiment utilizes a holistic optimization model to perform global optimization and adaptive evolution for the entire system. By establishing a multi-objective function that considers deodorization effectiveness, chemical consumption, energy consumption, and equipment load, and solving it under constraints, it can find the optimal balance between multiple conflicting objectives, ensuring the system operates at the highest overall efficiency. By continuously adjusting control parameters using a particle swarm optimization algorithm, the system continuously learns and adapts to environmental changes, achieving long-term, dynamic optimization of operating costs and efficiency.

[0043] This embodiment provides multi-zone coordinated control based on a "whole-local-whole" approach, achieving comprehensive suppression of odor diffusion and optimal configuration of deodorizers and energy consumption throughout the entire food waste pretreatment process. By integrating data from multiple sensors to identify operating conditions, the system goes beyond simply reacting to odors and instead delve deeper into the source, understanding whether crushing, pressing, or other processes are causing the odor. This makes subsequent "whole-zone control" proactive and targeted, enabling the implementation of the most appropriate strategy at the outset of a problem. Next, a three-dimensional control architecture for "whole-local" coordinated operations is constructed. By designing a layered response mechanism, it incorporates both a global control program, localized encrypted control to handle local emergencies, and regional linkage control to prevent the spread of problems. This multi-layered coordination enables the system to both manage daily operations with ease and flexibly and efficiently handle various emergencies and diffusion events. Finally, a closed-loop self-evolutionary mechanism based on a holistic "execution-feedback-optimization" process is established. After all control operations are executed, a comprehensive optimization model is used to comprehensively evaluate all results based on multiple factors, such as deodorization effectiveness and resource consumption, to automatically identify the optimal control parameter combination. This enables the system to continuously improve efficiency in the long run and automatically find the optimal balance between the best results and the lowest costs.

[0044] Example 2: In embodiment one, the method of the present application successfully realizes the overall inhibition of odor space diffusion in the whole process of kitchen waste pretreatment, and optimizes the configuration of deodorant and energy consumption. In order to further verify the effectiveness of the present application, the spraying deodorization control of another kitchen waste pretreatment workshop in the waste treatment plant is also carried out in the embodiments of the present application.

[0045] The pretreatment area is divided into N sub-control areas, and a multi-modal sensor is arranged in each sub-control area to collect multi-modal sensing data, including odor concentration data. The multi-modal sensor includes an odor concentration sensor, a temperature and humidity sensor, a vibration sensor, a motor current sensor and an air flow speed sensor; the multi-modal sensing data further includes air flow speed, equipment vibration data, temperature, humidity and motor operating current data.

[0046] Further, by a working condition state recognition model, the current pretreatment working condition state is recognized based on the multi-modal sensing data; according to the pretreatment working condition state, a spraying control parameter combination is automatically matched from a control strategy library by a control system, an overall control program is executed, and an overall control result is obtained, including an overall odor index and an overall disturbance index; The working condition state recognition model includes a working condition feature extraction unit, a working condition state judgment unit and a working condition conversion detection unit. The working condition feature extraction unit includes a basic physical feature extraction subunit, a working condition event feature construction subunit and a working condition feature generation subunit, and is used for extracting working condition features based on the multi-modal sensing data; the working condition features are double-layer features based on "physics-event", including basic physical features and working condition event features; the basic physical feature extraction subunit extracts the basic physical features representing the basic running state of the equipment from the multi-modal sensing data, including time domain statistical features, frequency domain features and wavelet energy features; the working condition event feature construction subunit maps each basic physical feature to the working condition event feature based on an expert knowledge base; the working condition feature generation subunit performs weighted fusion on the basic physical features and the working condition event features to obtain the working condition features; the working condition state judgment unit identifies the current pretreatment working condition state using the working condition features; and the working condition conversion detection unit identifies the working condition switching time based on the pretreatment working condition state, generates a transition period control parameter combination and stores it in the control strategy library.

[0047] Since the pretreatment workshop is a dynamic system, the operation of different stages (such as crushing vs. squeezing) will significantly change the odor generation rule; if the overall operating condition is not identified first, the local response will lack context information, and the control will easily fail or be excessive. Therefore, the present application first carries out "overall identification driving", based on the multiple process operating states existing in the pretreatment link, such as crushing, squeezing, temporary storage, etc., it is proposed to obtain the multi-source data of the workshop environment and equipment operation through the layout of multi-modal sensors (including odor concentration, motor current, temperature and humidity, etc.), and the operating condition recognition program is called by the control system to identify the current pretreatment stage. The system then automatically matches the corresponding spray parameter combination (such as spray time, pressure, spray head matrix, etc.) from the preset control strategy library according to the identification result, and executes the overall spray control program. This stage enables the system to complete the automatic identification of complex processing states and the calling of response strategies through program control, realizing the operating condition driving and automatic starting logic of the spray control.

[0048] Further, the overall control result further includes: overall spray coverage, overall deodorant consumption, and overall system energy consumption index; the control strategy library contains M preset control parameter combinations and transition period control parameter combinations; the preset control parameter combinations include spray head type, spray head opening and closing matrix, spray pressure level, spray duration, and spray interval period.

[0049] Further, if the overall odor index meets the first local control condition, a local encryption control program is executed according to the odor concentration data of each sub-control region to obtain a first local control result. The overall odor index meets the first local control condition, and the specific judgment process is: when the overall odor index exceeds the preset overall odor threshold and / or the odor concentration data of any sub-control region exceeds the local odor threshold, local encryption control is triggered; According to the odor concentration data of each sub-control region, a local encryption control program is executed to obtain a first local control result, and the specific process includes: calculating the odor concentration gradient of each sub-control region according to the odor concentration data of each sub-control region, marking the sub-control region with an odor concentration gradient higher than the odor concentration gradient threshold as a key control region, and performing local encryption control on the key control region; obtaining the first local odor concentration, the first local spray coverage, the first local deodorant consumption, and the first local energy consumption of each control sub-region after local encryption control, and fusing them respectively to obtain the first local control result.

[0050] Further, if the overall disturbance index meets the second local control condition, a local linkage control program is executed through a regional linkage control model to obtain a second local control result. The overall disturbance index satisfies a second local control condition, and the specific judgment process is: based on the odor concentration data of each sub-control region and the overall disturbance index, the odor concentration difference and the odor diffusion rate between adjacent sub-control regions are calculated, and when the odor concentration difference exceeds the odor concentration difference threshold and the odor diffusion rate exceeds the odor diffusion rate threshold, the local linkage control is triggered. The regional linkage control model includes: an odor diffusion trend prediction unit, a linkage region selection unit, a linkage parameter dynamic adjustment unit and a second local control result output unit. The odor diffusion trend prediction unit predicts the potential odor diffusion path based on the odor diffusion rate and the diffusion direction; the linkage region selection unit automatically identifies and determines the linkage region according to the odor diffusion path, and generates a linkage response area list; the linkage parameter dynamic adjustment unit adjusts the spraying linkage parameters for the sub-control regions in the linkage response area list based on the original spraying parameters, the distance between each sub-control region in the linkage response area list and the odor source region, the predicted odor arrival time and the odor concentration data, and performs local linkage control; The second local control result output unit obtains the second local odor concentration, the second local spraying coverage, the second local deodorant consumption and the second local energy consumption of each control sub-region after local linkage control, and fuses them respectively to obtain the second local control result.

[0051] After the overall control, it is considered that there may still be some residual odor or diffusion signs in some sub-regions; if the overall control is continued, it will cause waste of resources; therefore, a "conditionally triggered" local control program is needed. Therefore, "local dynamic response" is adopted, after the overall control is executed, control feedback data is further collected and logical judgment is performed. When the overall odor index exceeds the first threshold, the system performs a local encryption control program based on the odor concentration of each sub-control region, realizing rapid reinforcement response to high odor areas; when the overall disturbance index exceeds the second threshold, the system calls the regional linkage control logic and executes the local linkage control program to carry out collaborative spraying regulation in related sub-regions. This "local dynamic response" link adopts a hierarchical condition triggering mechanism and distributed control logic, which can not only accurately respond to local odor mutations, but also realize regional collaborative control under odor diffusion trend, significantly improving the response sensitivity and resource allocation efficiency of the system.

[0052] Further, the control parameters are regulated by using the overall optimization model in combination with the overall control result, the first local control result and the second local control result.

[0053] The overall optimization model comprises a multi-objective function construction unit, a constraint condition setting unit and a control parameter optimization unit; the multi-objective function construction unit establishes a multi-objective function of maximizing deodorization effect, minimizing deodorant consumption, minimizing energy consumption and minimizing equipment load; the constraint condition setting unit sets a lower limit of spray coverage, an upper limit of deodorant consumption and a system response time constraint; and the control parameter optimization unit solves the optimal control parameter combination by a particle swarm optimization algorithm.

[0054] After multiple rounds of control execution, considering that the environmental protection and cost of the entire system need to be ensured after the above-mentioned "overall-local" deodorization control is completed, the feedback optimization process based on the overall is returned again, the control effect is judged through feedback data, the deodorization control results of the local and the overall are uniformly included into the overall strategy optimization logic, and continuous optimization of program parameters is realized, which is the "overall-local-overall" closed-loop logic. The overall control result, the first local control result and the second local control result are taken as feedback inputs, and the spray control parameters are logically calculated and strategy updated by the optimization control module. The optimization process follows the preset control rules, and takes the odor suppression effect, the spray agent usage and the energy consumption data as basic variables to form the parameter combination required for the next round of control, so that the programmatic update of the control strategy is realized. The parameter correction mode driven by the logical rules is adopted, which does not depend on manual intervention or complex learning algorithm, so that the adaptive iteration of the control strategy can be completed, and the stability and engineering deployability of the control system are enhanced.

[0055] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for controlling the pretreatment and spraying of kitchen waste based on multimodal sensor fusion, characterized in that: include: The pretreatment area is divided into N sub-control areas, and a multimodal sensor is set in each sub-control area to collect multimodal sensing data, including odor concentration data; The current pre-treatment working condition is identified based on the multimodal sensing data using a working condition recognition model; according to the pre-treatment working condition, the control system automatically matches a combination of spray control parameters from a control strategy library, executes an overall control program, and obtains an overall control result, including an overall odor index and an overall disturbance index; If the overall odor index satisfies the first local control condition, a local encryption control program is executed according to the odor concentration data of each sub-control area to obtain a first local control result; If the overall disturbance index satisfies the second local control condition, the local linkage control program is executed through the regional linkage control model to obtain the second local control result; The control parameters are regulated by using the overall optimization model in combination with the overall control result, the first local control result and the second local control result.

2. The method for controlling kitchen waste pretreatment spray deodorization based on multimodal sensor fusion according to claim 1, characterized in that: The multimodal sensor includes: an odor concentration sensor, a temperature and humidity sensor, a vibration sensor, a motor current sensor and an airflow velocity sensor; the multimodal sensing data also includes: airflow velocity, equipment vibration data, temperature, humidity and motor operating current data.

3. The method for controlling kitchen waste pretreatment spray deodorization based on multimodal sensing fusion according to claim 1, characterized in that: The working condition recognition model includes: a working condition feature extraction unit, a working condition determination unit and a working condition conversion detection unit; The operating condition feature extraction unit includes a basic physical feature extraction subunit, an operating condition event feature construction subunit and an operating condition feature generation subunit, which are used to extract operating condition features based on the multimodal sensing data; the operating condition features are two-layer features based on "physics-event", including basic physical features and operating condition event features; the basic physical feature extraction subunit extracts the basic physical features that characterize the basic operating status of the equipment from the multimodal sensing data, including time domain statistical features, frequency domain features and wavelet energy features; the operating condition event feature construction subunit maps each basic physical feature into the operating condition event feature based on the expert knowledge base; the operating condition feature generation subunit performs weighted fusion of the basic physical features and the operating condition event features to obtain the operating condition features; the operating condition state discrimination unit uses the operating condition features to identify the current pre-processing operating condition state; the operating condition conversion detection unit identifies the operating condition switching moment based on the pre-processing operating condition state, generates a transition period control parameter combination, and stores it in the control strategy library.

4. The method for controlling kitchen waste pretreatment spray deodorization based on multimodal sensing fusion according to claim 1, characterized in that: The overall control result also includes: overall spray coverage, overall deodorant consumption and overall system energy consumption index; the control strategy library contains M preset control parameter combinations and transition period control parameter combinations; the preset control parameter combinations include nozzle type, nozzle opening and closing matrix, spray pressure level, spray duration and spray interval period.

5. The method for controlling kitchen waste pretreatment spray deodorization based on multimodal sensing fusion according to claim 1, characterized in that: The overall odor index satisfies the first local control condition, and the specific judgment process is as follows: when the overall odor index exceeds a preset overall odor threshold and / or the odor concentration data of any sub-control area exceeds a local odor threshold, triggering local encryption control; A local encrypted control program is executed according to the odor concentration data of each sub-control area to obtain a first local control result. The specific process includes: calculating the odor concentration gradient of each sub-control area according to the odor concentration data of each sub-control area, marking the sub-control area where the odor concentration gradient is higher than the odor concentration gradient threshold as a key control area, and performing local encrypted control on the key control area; obtaining the first local odor concentration, the first local spray coverage, the first local deodorant consumption and the first local energy consumption of each control sub-area after local encrypted control, and fusing them respectively to obtain the first local control result.

6. The method for controlling kitchen waste pretreatment spray deodorization based on multimodal sensor fusion according to claim 1, characterized in that: The overall disturbance index satisfies the second local control condition. The specific judgment process is as follows: based on the odor concentration data of each sub-control area and the overall disturbance index, the odor concentration difference and the odor diffusion rate between adjacent sub-control areas are calculated. When the odor concentration difference exceeds the odor concentration difference threshold and the odor diffusion rate exceeds the odor diffusion rate threshold, the local linkage control is triggered; The regional linkage control model includes: an odor diffusion trend prediction unit, a linkage area selection unit, a linkage parameter dynamic adjustment unit, and a second local control result output unit; The odor diffusion trend prediction unit predicts the potential odor diffusion path based on the odor diffusion rate and diffusion direction; the linkage area selection unit automatically identifies and determines the linkage area according to the odor diffusion path, and generates a linkage response area list; the linkage parameter dynamic adjustment unit dynamically adjusts the spray linkage parameters for the sub-control areas in the linkage response area list, based on the original spray parameters, according to the distance between each sub-control area in the linkage response area list and the odor source area, the predicted odor arrival time and the odor concentration data, to perform local linkage control; the second local control result output unit obtains the second local odor concentration, the second local spray coverage, the second local deodorant consumption and the second local energy consumption of each control sub-area after the local linkage control, and integrates them respectively to obtain the second local control result.

7. The method for controlling kitchen waste pretreatment spray deodorization based on multimodal sensor fusion according to claim 1, characterized in that: The overall optimization model includes: a multi-objective function construction unit, a constraint setting unit and a control parameter optimization unit; the multi-objective function construction unit establishes a multi-objective function of maximizing deodorization results, minimizing deodorant consumption, minimizing energy consumption and minimizing equipment load; the constraint setting unit sets the lower limit of spray coverage, the upper limit of deodorant consumption and the system response time constraint; the control parameter optimization unit solves the optimal control parameter combination through a particle swarm optimization algorithm.

Images