Energy-saving control system and method for dust collectors based on visual recognition and operating condition linkage

Abstract

This invention discloses an energy-saving control system and method for dust collectors based on visual recognition and operational condition linkage, belonging to the field of control system technology. First, image sequences of each operational sub-area of ​​the target plant are collected and imported into a lightweight visual recognition model to obtain the operational recognition results for each sub-area. Temporal stability correction is performed on the operational recognition results of each sub-area. Real-time air network feedback data of each sub-area is correlated and compared through visual judgment. Whether each sub-area has entered an effective dust-generating state is determined through situational fusion criteria. Linkage matching analysis of the branch pipe dynamic pressure situation and the branch pipe second static pressure situation is performed to obtain the comprehensive demand intensity of each sub-area. Based on the comprehensive demand intensity of each sub-area, the opening degree of the air valves in each sub-area is obtained, realizing a multi-zone linkage and coordinated energy-saving control strategy.

Description

Technical Field

[0001] This invention relates to the field of control system technology, and in particular to an energy-saving control system and method for dust collectors based on visual recognition and operating condition linkage. Background Technology

[0002] Industrial plants continuously generate large amounts of dust during welding, cutting, grinding, crushing, and conveying operations. To meet occupational health, safety, and environmental emission requirements, plants are generally equipped with central dust collection systems. Dust is filtered and discharged through dust collectors. Dust collection conditions are highly dynamic and spatially discrete. The amount of dust generated in different work areas changes rapidly with factors such as personnel movements, equipment rhythm, and material type. It is difficult to perceive in real time whether each sub-area is truly dust-generating, and it is also impossible to flexibly adjust the opening of dampers and airflow distribution according to actual needs. Maintaining excessively high suction in non-working areas will lead to excessive energy consumption, while insufficient suction in high-dust-generating areas will cause dust overflow, pollution diffusion, and deterioration of the working environment.

[0003] Currently, dust removal systems commonly used in industrial production environments still primarily rely on fixed airflow and constant air pressure control. The vast majority of these systems provide stable negative pressure through a main fan, with the opening of branch pipe dampers fixed or manually set, remaining static and unchanging regardless of operational status or on-site dust generation. In recent years, however, advancements in industrial vision, sensor networks, and intelligent control technologies have enabled the deployment of lightweight visual models in various operational sub-areas to identify operational status. Combined with feedback data from the branch pipe dynamic and static pressure networks, dual-source cross-validation of dust generation conditions allows for millisecond-level determination of actual dust generation in a given area and accurate measurement of suction demand. Furthermore, based on the intensity of regional demand, the opening of dampers and airflow distribution are dynamically adjusted to achieve an energy-saving control strategy of strong suction in operational areas, weak suction in non-operational areas, and coordinated multi-area operation.

[0004] For example, CN115755776A discloses an energy-saving control method and early warning method for a dust collector in a plasma cutting machine, relating to the field of plasma cutting technology. The method includes the following steps: Before cutting begins, a program number is exported from the plasma cutting machine system; the plasma cutting machine system retrieves target data from the design department's database according to the program number; the control unit of the plasma cutting machine outputs a dust removal signal to the dust collector's control system according to the dust collector control command; the control unit of the dust collector receives the dust removal signal and controls the inverter output to control the motor speed to achieve the dust removal frequency specified by the dust collector control command; the motor drives a centrifugal fan to create an appropriate negative pressure at the point where the cutting generates dust, drawing the plasma cutting dust into the dust collector.

[0005] For example, announcement number CN112711187B discloses a multi-field coordinated control method for a dry electrostatic precipitator in a coal-fired power unit. The main steps of the method are as follows: S1. Static characteristic tests are conducted on the upstream electric field under various typical operating conditions to obtain unit load data and energy-saving operation data of the electrostatic precipitator; S2. Under environmental constraints, an open-loop control strategy based on a data-driven model is designed for the upstream electric field; S3. Dynamic characteristic tests are conducted on the final stage and the transfer function model of the final stage electric field is obtained; S4. Under the constraints of robustness and control performance, a closed-loop PID control strategy is designed for the final stage electric field, and an interference observer is added to improve the anti-interference capability of the closed-loop system.

[0006] The above-mentioned technology has at least the following technical problems:

[0007] The system lacks the ability to judge the real-time operating conditions of each work area. When a certain area is not in operation or the dust generation is very low, the system still maintains a high suction volume, resulting in a large amount of wasted electricity. When multiple areas are in operation at the same time or a certain area experiences a short period of high-intensity dust generation, the air volume distribution cannot be adjusted in time, which may lead to problems such as insufficient suction, dust escape, or even backflow. In addition, the dust removal fan operates at the power frequency or fixed frequency regardless of whether there is operation, and the branch pipe valves are always open, resulting in a large amount of wasted electricity when no load and serious wear and tear on valves and pipes due to long-term exposure to high-speed airflow. Summary of the Invention

[0008] To address the technical problems existing in the prior art, this invention provides an energy-saving control system for dust collectors based on visual recognition and operating condition linkage. The technical solution includes: a lightweight visual operation recognition and temporal stability correction module, used to collect image sequences of each operation sub-area of ​​the target plant and import them into a lightweight visual recognition model to obtain the operation recognition results of each operation sub-area of ​​the target plant, and to perform temporal stability correction on the operation recognition results of each operation sub-area of ​​the target plant.

[0009] The multi-source air network situation fusion and dust generation status determination module is used to compare and correlate real-time air network feedback data of each operating sub-area of ​​the target plant through visual judgment, determine whether each operating sub-area of ​​the target plant has entered an effective dust generation state through situation fusion criteria, and perform linkage matching analysis on the dynamic pressure status of the branch pipe and the second static pressure status of the branch pipe to obtain the comprehensive demand intensity of each operating sub-area of ​​the target plant.

[0010] The adaptive control and multi-zone collaborative allocation module for damper opening is used to determine the damper opening of each sub-area of ​​the target plant based on the comprehensive demand intensity of each sub-area of ​​the target plant, and to perform energy-saving collaborative allocation based on the air volume demand relationship of each sub-area of ​​the target plant.

[0011] The damper opening control adaptive tracking module is used to track and monitor the damper opening after triggering damper opening control to obtain the following relationship characteristics between the damper opening change and the corresponding branch pipe dynamic pressure response, and to trigger the flow limiting control of the corresponding branch pipe damper based on the following relationship characteristics.

[0012] A second aspect of the present invention also provides an energy-saving control method for dust collectors based on visual recognition and operating condition linkage, comprising: acquiring image sequences of each operating sub-area of ​​the target plant and importing them into a lightweight visual recognition model to obtain the operation recognition results of each operating sub-area of ​​the target plant, and performing time-series stability correction on the operation recognition results of each operating sub-area of ​​the target plant.

[0013] By visually judging and comparing the real-time air network feedback data of each sub-area of ​​the target plant, the situation fusion criteria are used to determine whether each sub-area of ​​the target plant has entered an effective dust generation state. The dynamic pressure situation of the branch pipe and the second static pressure situation of the branch pipe are linked and matched to obtain the comprehensive demand intensity of each sub-area of ​​the target plant.

[0014] The opening degree of the air valves in each sub-area of ​​the target plant is obtained based on the comprehensive demand intensity of each sub-area of ​​the target plant, and energy-saving coordinated allocation is performed on the air volume demand relationship of each sub-area of ​​the target plant.

[0015] The beneficial effects of the technical solutions provided by the embodiments of the present invention include at least the following:

[0016] (1) This invention proposes an energy-saving control system for dust collectors based on visual recognition and operational condition linkage. It constructs an intelligent energy-saving control process for dust-prone conditions by linking visual recognition with air network operational condition feedback across modalities. First, the system collects image sequences of each operating sub-area in the target plant. It uses a lightweight visual recognition model to obtain the operational status and eliminates short-term fluctuations, misidentifications, and pseudo-jumps through temporal stability correction, outputting a reliable operational status judgment. Then, it compares the corrected visual status with real-time air network feedback such as branch pipe dynamic pressure and static pressure. Through situational fusion criteria, it identifies whether each operating sub-area has entered an effective dust-generating state. Based on the dynamic pressure rise intensity, static pressure slight fall degree, and visual confidence level, it jointly constructs a comprehensive demand intensity characterization to achieve quantitative assessment of air volume demand. Finally, the controller calculates the corresponding valve opening based on the comprehensive demand intensity of each operating sub-area and performs collaborative optimization allocation of air volume demand across multiple areas. This ensures the suction capacity of high-load dust-generating areas while suppressing ineffective air volume consumption in low-load and non-operating areas, thereby significantly improving the response efficiency and overall energy-saving effect of the dust collection system. It achieves deep integration of visual information and wind network operating condition information, enabling the dust removal system to perceive changes in operations in real time, respond accurately to dust generation behavior, and allocate air volume resources on demand.

[0017] (2) This invention obtains the probability distribution of the operation status of each sub-region, and then introduces a temporal stability correction mechanism. Within a preset temporal stability window, the gradient change rate and local variance of each state probability are jointly statistically analyzed, and short-term fluctuations are filtered by the minimum duration constraint, thereby outputting a valid operation status that has been verified by stability. This visual recognition and temporal correction step ensures that the system can still obtain reliable operation status judgments in complex factory environments such as dust, light changes, and shading, providing reliable input for subsequent linkage analysis with wind network operation data. Through this mechanism, the dust removal system can accurately identify the actual operation trigger time without relying on high-frequency sensor prediction, reducing ineffective ventilation, reducing energy consumption, and making subsequent dynamic / static pressure situation fusion and demand allocation more targeted, thereby achieving efficient energy-saving control based on visual recognition and operation condition linkage.

[0018] (3) This invention deeply integrates the temporal stability correction of visual recognition with the real-time extraction of air network pressure status, achieving accurate determination of dust generation status in multiple sub-areas of the factory. Subsequently, differential and trend extraction are performed on the dynamic and static pressure sequences of the branch pipes within the dust generation monitoring window to obtain pressure status labels such as dynamic pressure rising, static pressure slightly falling, or static pressure stabilizing, which are used to describe the real-time air intake demand of each sub-area. This can effectively suppress misjudgments caused by visual false detection, optical interference, or pressure noise, making the dust generation status determination more robust. This provides an accurate basis for subsequent valve opening adjustment and air volume coordinated distribution, enabling the dust removal system to significantly reduce unnecessary air volume output while meeting dust collection efficiency, thereby achieving the goal of energy saving and consumption reduction.

[0019] (4) This invention further introduces pressure situation consistency verification for sub-regions that are visually identified as having an effective operating state. It strengthens the authenticity judgment of visual results by utilizing the matching relationship between the rising dynamic pressure of the branch pipe and the falling or stable static pressure. Subsequently, combined with the state probability distribution of the sub-region and the confirmed dust generation state, the comprehensive demand intensity of the sub-region is calculated according to the pre-set dynamic weighted allocation rules, realizing the adaptive fusion of visual state intensity and pressure situation intensity with real-time confidence. Finally, the comprehensive demand intensity of all sub-regions is input into the air volume allocation model of the central controller. The central controller sorts all sub-regions according to demand intensity and performs overall coordination on the air volume demand relationship of different sub-regions in combination with the current main static pressure, the adjustable margin of the main fan, and other global operating conditions, so as to achieve the optimal scheduling of the entire network air volume while ensuring the suction capacity of high dust generation areas. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the system modules provided in an embodiment of the present invention;

[0022] Figure 2 This is a flowchart of the operation status acquisition and timing stability correction based on visual recognition provided in the embodiments of the present invention;

[0023] Figure 3 This is a diagram of the lightweight visual recognition model architecture provided in an embodiment of the present invention;

[0024] Figure 4 The actual monitoring and intelligent triggering interface diagram of the target factory building provided in this embodiment of the invention;

[0025] Figure 5 The present invention provides a combined flowchart of visual operation status and branch pressure status;

[0026] Figure 6 This is a flowchart of a dust collector energy-saving control method based on visual recognition and operating condition linkage provided in an embodiment of the present invention. Detailed Implementation

[0027] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0028] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0029] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0030] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0031] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0032] like Figure 1 As shown, this embodiment of the invention provides an energy-saving control system for a dust collector based on visual recognition and operating condition linkage, including: a lightweight visual operation recognition and time-series stability correction module, used to collect image sequences of each operation sub-area of ​​the target plant and import them into a lightweight visual recognition model to obtain the operation recognition results of each operation sub-area of ​​the target plant, and to perform time-series stability correction on the operation recognition results of each operation sub-area of ​​the target plant.

[0033] The multi-source air network situation fusion and dust generation status determination module is used to compare and correlate real-time air network feedback data of each operating sub-area of ​​the target plant through visual judgment, determine whether each operating sub-area of ​​the target plant has entered an effective dust generation state through situation fusion criteria, and perform linkage matching analysis on the dynamic pressure status of the branch pipe and the second static pressure status of the branch pipe to obtain the comprehensive demand intensity of each operating sub-area of ​​the target plant.

[0034] The adaptive control and multi-zone collaborative allocation module for damper opening is used to determine the damper opening of each sub-area of ​​the target plant based on the comprehensive demand intensity of each sub-area of ​​the target plant, and to perform energy-saving collaborative allocation based on the air volume demand relationship of each sub-area of ​​the target plant.

[0035] The damper opening control adaptive tracking module is used to track and monitor the damper opening after triggering damper opening control to obtain the following relationship characteristics between the damper opening change and the corresponding branch pipe dynamic pressure response, and to trigger the flow limiting control of the corresponding branch pipe damper based on the following relationship characteristics.

[0036] Example 1: A 1.2 million cubic meter per second three-stage dust removal system is divided into N sub-areas (e.g., blast furnace area, desulfurization station, continuous casting rotary table, etc.). Industrial dust cameras are installed in each sub-area of ​​the plant. Each camera is fixed in a position that allows it to overlook the current work surface, and the lens is covered with a basic dust cover and regularly cleaned with air blowing to ensure a clear video stream even in a dusty environment. After the cameras are deployed, all real-time video is connected to the edge computing terminal on site, typically using an industrial embedded GPU module. The edge computing terminal is responsible for receiving video images in real time. The acquisition process is based on a continuous video stream. Non-continuous operations such as crane tipping, silo unloading, ladle tilting, manual unloading, and belt conveyor transfer in each sub-area of ​​the target plant are used as image acquisition triggers. The image sequence acquired per second is sent to a pre-trained lightweight visual recognition model at a fixed frame rate. The trained model is embedded in the edge computing device, which infers the images from each camera in real time at millisecond intervals, continuously outputting the probability and stability of the corresponding work state of the sub-area. Specifically, the process is as follows: First, a lightweight visual operation recognition and temporal stability correction module is used to identify the operation status of each sub-region of the target factory. The specific operation involves: acquiring image sequences of each sub-region, extracting local gradient increment features, including pixel brightness change amplitude, gradient increment density, average optical flow velocity, and texture block structure offset; and inputting the extracted feature vectors into a pre-trained lightweight visual recognition model to obtain the probability distribution of each state in each sub-region. Then, within a preset temporal stability window, the gradient change rate and local variance of each state probability are statistically analyzed to calculate the stable value of each state probability. The effective operation state is then determined by combining the minimum duration threshold of the state probability, thereby locking and outputting the operation status of each sub-region. Next, a multi-source wind network situation fusion and dust generation status determination module is used to analyze the dynamic operating conditions of the wind network in each sub-region. First, dynamic and static pressure signals from the branch pipes of each sub-area are collected. Differential and trend aggregation analyses are performed on the dynamic pressure within the dust generation monitoring window to obtain the average rate of change and net magnitude of change, thus marking an upward trend in dynamic pressure. Short-term trend extraction is performed on the static pressure signal to obtain the average rate of change and net magnitude of change, marking the static pressure state as either slightly declining or stable. Next, the visually confirmed operational status is compared with the branch pipe status. If the dynamic and static pressure status matches the visually recognized operational status, the sub-area is confirmed to be in an effective dust generation state. Based on the confirmed dust generation state and the probability distribution of each visually identified state, the comprehensive demand intensity of each operational sub-area in the target plant is calculated according to a dynamic weighted allocation rule, providing a basis for airflow control. Finally, the corresponding damper opening is determined based on the comprehensive demand intensity of each sub-area.The central distributor prioritizes the airflow to high-demand sub-areas based on overall demand intensity, while simultaneously coordinating the airflow demand across multiple areas by considering the main static pressure, main fan outlet airflow, and the adjustable margin of the main fan inverter. This achieves energy-efficient and coordinated airflow distribution among the branch pipes, ensuring that high-dust-generating areas receive sufficient suction in a timely manner, while preventing main fan overload and airflow crosstalk, thereby improving dust removal efficiency and optimizing system energy conservation.

[0037] like Figure 2 As shown, Figure 2 The flowchart for acquiring and correcting the temporal stability of a work status based on visual recognition provided in this embodiment of the invention first collects image sequences of each work sub-region of the target factory building and inputs them into a lightweight visual recognition model to obtain the probability distribution of each status. Then, within a preset temporal stability window, the gradient rate of change and local variance of each status probability are statistically analyzed to calculate the stable value and determine the valid work status by combining it with the minimum duration threshold. The specific process of collecting image sequences of each work sub-region of the target factory building and importing them into the lightweight visual recognition model to obtain the work recognition results of each work sub-region of the target factory building is as follows:

[0038] Extract local gradient increment data for each sub-region of the target factory, including pixel brightness change amplitude, gradient increment density, average velocity of optical flow field, and texture block structure offset. Establish a unified local gradient increment vector and input it into the pre-trained lightweight visual recognition model. Output the operation recognition results for each sub-region of the target factory, which include the probability distribution of each state of each sub-region of the target factory.

[0039] It should be noted that, as Figure 3 As shown, Figure 3This diagram illustrates the architecture of a lightweight visual recognition model. It extracts features from the input image and fuses information at different levels. Multi-scale feature extraction is achieved through Shuffle Blocks, CSP modules, upsampling, and feature concatenation. Finally, three prediction heads—Small, Medium, and Large—output the operation recognition results at different scales. Specifically, during the training of this lightweight visual recognition model, multiple batches of operational monitoring videos from the target factory are used as the data source. Each frame's sub-region is labeled with its corresponding status tag ("No Operation," "Operation Preparation," "Operating") using manual or semi-automatic annotation tools. When constructing the training set, all frames are enhanced with illumination perturbations, blurring, noise reduction, smoke and dust overlay, and occlusion simulation to ensure the model's robustness in real industrial environments. Subsequently, local gradient increment information is extracted from the video sequence, including the amplitude of brightness changes, gradient increment density, average velocity of the optical flow field, and the offset of the regional texture structure. After normalization, these features are concatenated into a unified local gradient increment vector, which is then input into the model along with the corresponding image region-cropped segments. The model itself employs a lightweight YOLO architecture, such as YOLOv5n, YOLOv8n, or YOLOv10-Nano, as its foundation. Real-time performance on low-computing-power devices is achieved by adjusting the backbone depth and channel width. During training, each sub-region of the factory is treated as a fixed-region classification task, utilizing the classification head (rather than the detection head) in YOLO to predict the probability distribution of the three types of work states. To integrate the gradient increment features you mentioned, the local gradient increment vector is mapped to the same dimension as the features in the middle layer of YOLO through a lightweight MLP or 1×1 convolutional branch. Feature fusion is then performed in the neck stage of YOLO (such as FPN / PAN or BiFPN), enabling the model to learn work features not only from image texture, shape, and optical flow, but also to explicitly utilize the dynamic change information brought by the gradient increment. The training process uses cross-entropy loss or BCE-with-logits as the three-class classification supervision signal, combined with YOLO's default data augmentation strategies (mosaic, mixup, HSV perturbation, multi-scale training). Simultaneously, EMA, mixed-precision training, and sparse pruning are enabled to improve training stability and reduce the number of model parameters. With each training iteration, the model gradually learns regional motion patterns, brightness variation rhythms, and structural perturbation patterns in different operational states, enabling the three categories—"no operation," "operation preparation," and "operation in progress"—to form clear and separable boundaries in the feature space. The resulting lightweight YOLO model can directly accept the image and its corresponding local gradient increment vector during the inference phase, and after fusion, outputs the probability distribution of the three operational states for each operational sub-region of the target factory, achieving stable and real-time operation recognition.

[0040] It should be noted that the probability distribution of each state in each sub-area of ​​the target plant includes the probability distribution of no operation, the probability distribution of operation preparation, and the probability distribution of operation in progress.

[0041] like Figure 4 As shown, Figure 4 This is a diagram of the actual monitoring and intelligent triggering interface of the target plant. The lightweight visual model identifies the crane reversing (in operation) action and automatically triggers dust removal accordingly. The interface includes the monitoring status of the sub-area, real-time identification results, regional operation trend changes, automatic dust removal records, operation status videos and statistical data. The operation status trend curve, YOLO operation identification results, operation condition interpretation and historical operation data are derived from the time-series stability confirmation of the visual identification results and the dust generation determination after the fusion of pressure situation. The number of dust removal device start-ups, operation trend curve and sub-area status interpretation reflect the comprehensive demand intensity calculation and the dynamic control logic of the central controller on the air volume. The automatic dust removal process includes performing temporal stability correction on the operation recognition results of each sub-area of ​​the target factory. The specific process is as follows: a temporal stability window is preset, and the probability of each state of the lightweight visual recognition model of each sub-area of ​​the target factory is aggregated and statistically analyzed in the temporal stability window to obtain the gradient change rate and local variance of each state probability of each sub-area of ​​the target factory. The gradient change rate and local variance of each state probability of each sub-area of ​​the target factory are added together to obtain the stable value of each state probability of each sub-area of ​​the target factory.

[0042] Extract the stable values ​​of the state probabilities of each sub-region of the target plant, and count the minimum duration of each state probability of each sub-region of the target plant in the time-series stability window. Combine this with the minimum duration threshold of the state probability stored in the database to obtain the effective operation status of each sub-region of the target plant.

[0043] It should be noted that when the probability of a certain state in a sub-region of a target plant's operation exceeds a relatively high entry threshold for the first time within the time-series stability window, a "countdown" will be initiated, and the system will continuously monitor the situation within the time-series stability processing window. Only when the probability of that state exceeds the entry threshold and the duration reaches the preset minimum duration threshold will the state be ultimately determined to be valid.

[0044] It should be noted that dust-resistant cameras are deployed in each sub-area of ​​the target factory, capturing images at a constant frame rate. Edge nodes feed each frame of the image, after brightness correction and smoke enhancement, into a lightweight visual recognition model to obtain probability sequences for three states. A time-stability window is preset on the controller side, covering the time scale of typical work actions—for example, under a sampling condition of twenty to thirty frames per second, the window length can be one to three seconds to balance response speed and noise suppression. Within each sliding window, the controller first calculates the change amplitude of the probability of adjacent frames for each state's probability sequence and takes their time average to characterize the short-term probability change rate. At the same time, it calculates the variance of the probability within the window to characterize the fluctuation amplitude. Then, the probability change rate and local variance are normalized to engineering dimensions and added together to obtain the stable value of the state. The duration accumulation is recorded internally by the controller in units of time slices. When the accumulated time of a state in a continuous sliding window reaches the minimum duration threshold of the corresponding state in the database, the system confirms the state as a valid work state and locks it as an output; if the state falls back during the accumulation period and does not reach the threshold, it is considered a transient state and the accumulated time is cleared. Example 2 also incorporates a short-term anomaly removal mechanism into the correction logic: if an isolated extreme probability peak appears within the window and its corresponding gradient change does not increase proportionally with the variance, then the frame is considered noise and is not included in the calculation of stable values, thereby avoiding misjudgments caused by sparks and instantaneous illumination changes.

[0045] The effective operational status of each operational sub-area in the target factory is obtained through the following process:

[0046] The stable values ​​of the state probabilities of each sub-area of ​​the target plant are compared with the state probability stability thresholds stored in the database. If the stable value of the state probability of a certain sub-area of ​​the target plant is higher than or equal to the stable value state probability stability threshold, and the minimum duration of the state probability of that sub-area of ​​the target plant is lower than or equal to the minimum duration threshold of the state probability, then the sub-area of ​​the target plant is confirmed as a valid operating state and the output is locked. Otherwise, if the minimum duration requirement is not met, the state is regarded as transient and not confirmed.

[0047] like Figure 5 As shown, Figure 5 The flowchart for the combined visual operation status and branch pipe pressure status provided in this embodiment of the invention describes the process of using the combined visual operation status and branch pipe pressure status to confirm the dust generation status and comprehensive demand intensity of each operation sub-area of ​​the target plant. The process of confirming the dust generation status and comprehensive demand intensity of each operation sub-area of ​​the target plant includes comparing and correlating the real-time air network feedback data of each operation sub-area of ​​the target plant through visual judgment. The specific process is as follows: a dust generation monitoring window is preset, and the pressure sequence of the branch pipe dynamic pressure signal and the pressure sequence of the branch pipe static pressure sensor of each operation sub-area of ​​the target plant are obtained.

[0048] In the dust generation monitoring window, differential and trend aggregation are performed on the pressure sequence of the branch pipe dynamic pressure signal of each sub-area of ​​the target plant to obtain the short-term average rate of change and net change amplitude of the dynamic pressure of the branch pipe of each sub-area of ​​the target plant.

[0049] Within the same dust generation monitoring window, the same short-term trend extraction was performed on the pressure sequence of the branch static pressure sensor to obtain the average rate of change and net magnitude of static pressure.

[0050] The branch pipe situation of each work sub-area of ​​the target plant is constituted by the average rate of change and net magnitude of change of static or dynamic pressure of the branch pipes.

[0051] It should be noted that since each operating sub-area is connected to the main dust collection network through corresponding branch pipes, the dynamic / static pressure characteristics of the branch pipes constitute the pressure-side operating condition expression of that operating sub-area. The "pressure situation" obtained at the branch pipe level will be directly mapped to the pressure evidence of the corresponding operating sub-area. Therefore, the branch pipe situation, as the pressure situation of the corresponding operating sub-area, is used as physical evidence for subsequent dust generation confirmation.

[0052] It should be noted that if the net change in static pressure is negative and its absolute value is between the main short-term fluctuation threshold and the upper limit of static pressure disturbance noise—specifically, the absolute value of the net change in static pressure is greater than the upper limit of static pressure disturbance noise (indicating that the decrease is not caused by sensor noise or random jitter) and less than the main short-term wind network pressure fluctuation threshold (indicating that the decrease has not reached the level of overall wind network load disturbance)—that is, when the net change in static pressure is negative and its absolute value simultaneously satisfies the condition of being higher than the upper limit of static pressure disturbance noise and lower than the main short-term wind network pressure fluctuation threshold, i.e., satisfying the condition of upper limit of static pressure disturbance noise <|net change in magnitude| < main short-term fluctuation threshold, it can be determined that the branch pipe is in a state of slight static pressure decline. At this time, the decrease has exceeded the range of sensor noise and random jitter (indicating that there is a clear pressure decline trend), but has not reached the level of large pressure disturbance caused by the overall wind network load fluctuation (indicating that the pressure decline is a local and mild pressure release process), so it can be identified as a slight negative pressure change caused by normal operation.

[0053] The branch pipe status of each work sub-area of ​​the target plant includes the first dynamic pressure rise of the branch pipes in each work sub-area of ​​the target plant and the second static pressure status of the branch pipes in each work sub-area of ​​the target plant.

[0054] The first dynamic pressure rise trend of the branch pipes in each sub-area of ​​the target plant is defined as the branch pipes in that sub-area of ​​the target plant being marked as having a first dynamic pressure rise trend when the short-term dynamic pressure change rate of the branch pipes in each sub-area of ​​the target plant within the dust generation monitoring window is positive and the net change amplitude exceeds the window noise threshold stored in the database.

[0055] It should be noted that marking the branch pipes in this sub-area of ​​the target plant as having a "first dynamic pressure increase" is to distinguish between "dynamic pressure increases caused by actual dust generation" and "meaningless changes such as noise, vibration, and fan fluctuations." Because the dynamic pressure of the branch pipes is affected by equipment vibration, background disturbances, and airflow fluctuations, even without dust generation, there may be sporadic, small increases. Therefore, dynamic pressure changes alone cannot be directly used as evidence of dust generation. By judging whether the "dynamic pressure change rate is positive" and "whether the net change amplitude exceeds the noise threshold" within the dust generation monitoring window, all transient, weak, and non-operational pressure disturbances can be filtered out, retaining only the obvious dynamic pressure increase trend truly caused by material falling, dumping, or spilling. This trend is then marked as "first dynamic pressure increase," providing reliable and robust pressure-side evidence for subsequent multi-source correlation analysis.

[0056] The second static pressure status of the branch pipes in each sub-area of ​​the target plant includes a slight decrease in static pressure and a stable static pressure status. If the net change in static pressure is negative and its absolute value is between the short-term fluctuation threshold of the main trunk and the upper limit of static pressure disturbance noise, the branch pipe in that sub-area of ​​the target plant is marked as having a slight decrease in static pressure. If the net change in static pressure is near zero and the rate of change falls into the stable static range, the branch pipe in that sub-area of ​​the target plant is marked as having a stable static pressure status.

[0057] It's important to note that dynamic pressure represents the strength of airflow velocity, reflecting the kinetic energy of fluid movement across the pipe cross-section. Therefore, when dust begins to be generated in a work area or when it enters a working phase, a significant increase in flow velocity typically occurs in the local suction branch pipe, which is reflected as an increase in dynamic pressure over time on the pressure sensor. The criteria in quotation marks require a positive rate of change in dynamic pressure, and the net change must exceed the window noise threshold. This separates the effective flow velocity increase at the branch pipe caused by the actual dust suction demand from environmental disturbances, enabling the system to accurately capture the true suction demand signal caused by "work initiation." Static pressure reflects the internal pressure of the fluid and is the steady-state expression of the system's delivery capacity in that branch pipe. When a work area enters a working phase and the suction volume increases, the static pressure in the branch pipe often decreases slightly because a more significant negative pressure area needs to be formed in the branch pipe to meet local suction demand; however, this decrease usually does not reach the level of disturbance to the entire air network, nor is it large enough to cause a systemic pressure collapse. Therefore, marking a static pressure drop whose absolute value is greater than the upper limit of noise but less than the short-term fluctuation threshold of the main line as a slight decline is to distinguish the slight suction negative pressure caused by operation from noise fluctuations, and also to avoid misjudging such slight drops as overall load disturbances in the ventilation network. As for a stable static pressure state, it means that the branch pipes are not significantly affected by suction disturbances, and the pressure fluctuates slowly within a small range, which is reflected as an unloaded or stable low-load state.

[0058] It should be noted that the initial dynamic pressure increase confirms that the actual suction velocity of the branch pipe is significantly increasing, serving as physical evidence of dust generation activity on the pressure side and as one of the sources of cross-verification for visual judgment. The slight decrease in static pressure indicates that the suction of the branch pipe is increasing but has not reached the level of network disturbance, i.e., light-level operational disturbance; the stable static pressure indicates that the branch pipe is in a undisturbed or weakly disturbed state.

[0059] Example 2: Based on Example 1, after completing the above two types of situation division, the branch pipe situation of each sub-area of ​​the target plant, that is, the pressure behavior of the branch pipe within the dust generation monitoring window, has actually covered the main operating condition characteristic range: one type is that the dynamic pressure shows a significant increase, which can reflect the high-speed airflow changes caused by material disturbance or falling material; the other type is that the static pressure level slightly drops or remains stable, which is used to characterize the subtle pressure recovery process after the pipeline is pumped and the load changes. Apart from these two categories, the remaining pressure changes will basically fall into two categories: either strong noise or invalid disturbances (such as transient spikes caused by fan startup or valve operation), because these fluctuations do not meet the trend conditions for rising dynamic pressure or the criteria for slight static pressure decline or static stability, so they will be automatically classified as "non-situation" and will not participate in subsequent operating condition inference; or extreme abnormal pressure jumps, such as sudden rises or falls in static pressure caused by branch pipe blockage, abnormal backflushing, or fan failure. The magnitude of these sudden changes will exceed the upper limit of slight static pressure decline or fall outside the noise zone, and they will also not be included in the normal situation, but will be handled by the system in an independent abnormality handling process.

[0060] It should be noted that when the typical pressure patterns corresponding to the effective operating state (such as the rising dynamic pressure of the branch pipe and the slight falling static pressure) occur simultaneously, the pressure situation is consistent with the visual situation, confirming the dust generation state of the area. If the pressure situation is inconsistent with the visual situation, for example, the visual system shows "operation preparation" but the branch pipe pressure shows no obvious disturbance, then dust generation is not confirmed for the time being. If a sub-area is in an uncertain state, the dust generation state is confirmed entirely by the combined occurrence of the rising dynamic pressure and the slight falling static pressure, without relying on visual results. If the pressure remains stable, it is determined that the sub-area has not entered the actual dust generation condition.

[0061] It should be noted that the specific process of confirming the dust generation status based on the combined occurrence of rising dynamic pressure and slightly declining static pressure is as follows: First, retrieve the rate of change and net magnitude of dynamic pressure change from the pressure sequence of the branch pipe to which the sub-region belongs, and verify its effectiveness based on the noise threshold; when the rate of change of dynamic pressure is continuously positive and the net magnitude of change exceeds the window noise threshold, the pressure behavior of the branch pipe is interpreted as "rising dynamic pressure", which means that the local air volume demand of the sub-region is rising sharply, usually in sync with typical dust generation behaviors such as particle throwing and dust start-up.

[0062] The process of determining whether each sub-area of ​​the target plant has entered an effective dust-generating state through situational fusion criteria is as follows:

[0063] If the visual recognition of a certain sub-area of ​​operation has confirmed that it is in an effective operation state, then the effective operation state of each sub-area of ​​operation in the target plant is compared with the current branch pipe status of that sub-area of ​​operation in the target plant. If the pressure status is consistent with the current branch pipe status of that sub-area of ​​operation in the target plant, then the dust generation status of each sub-area of ​​operation in the target plant is confirmed; otherwise, the dust generation status of each sub-area of ​​operation in the target plant is not confirmed for the time being.

[0064] It should be noted that, based on the valid operation status labels output by visual recognition, a typical pressure behavior template corresponding to this type of operation is extracted from the database. This template describes the direction of pressure disturbance that should occur in the branch pipe under this type of operation condition. For example, for material discharge operations, the typical pressure behavior template requires the branch pipe dynamic pressure to rise rapidly within the dust generation window, while the static pressure will slightly decrease; for light operation operations, the typical pressure template may require a smaller increase in dynamic pressure and stable static pressure. After obtaining the expected pressure pattern, the first dynamic pressure rise trend and the second static pressure trend of the branch pipe extracted in real time are compared item by item with the expected pressure template. If the first dynamic pressure rise of the branch pipe satisfies the dynamic pressure disturbance direction of the operation type corresponding to the vision, that is, the rate of change of dynamic pressure is positive and the net change amplitude exceeds the window noise threshold, and the second static pressure situation is consistent with the static pressure disturbance characteristics of the operation corresponding to the vision (for example, if the operation requires a slight drop in static pressure, the actual static pressure situation must be marked as a slight drop in static pressure), the system determines that the visual recognition and the pressure situation are consistent within the current window, thereby confirming that the target plant operation sub-area is in a real dust-generating state.

[0065] If the visual output indicates that the operation has entered a valid working state, but the dynamic and static pressure conditions exhibited by the branch pipe do not simultaneously meet the typical disturbance patterns of this operation category—for example, if the visual assessment indicates a high operation intensity but the dynamic pressure does not show an effective upward trend, or the visual representation shows a material dropping action but the static pressure remains completely stable—then it is considered that the visual results may be affected by plume obstruction, backlighting, fog, or misidentification. Furthermore, the pressure may not provide disturbance evidence consistent with the operation. Therefore, the dust generation status of this area is temporarily not confirmed, and the operation status of this area remains in a "pending confirmation" or "transient" state, awaiting multimodal joint verification in the next dust generation monitoring window. By cross-validating the probability distribution of the operation status obtained from visual recognition with the physical evidence of branch pipe pressure disturbances, the system avoids misjudging dust generation due to false high-probability outputs caused by plume obstruction, backlighting, fog, local blurring, or model misidentification. This ensures that the final operation confirmation must be supported by both visual and pressure consistency. Specifically, based on the state probability distribution given by the visual model, determine which state the current area is most likely to be in: "no work", "work preparation" or "work in progress"; then, read the dynamic and static pressure changes of the branch pipe in the same monitoring window, and compare whether the pressure changes show the typical disturbance patterns that should be present in "work preparation" or "work in progress". For example, the "work preparation" stage usually shows a slight upward arch of dynamic pressure or a slight drop in static pressure, while "work in progress" will show a more obvious short-term rise in dynamic pressure or a shift in static pressure. If the visual system perceives an area as "in operation," but no disturbance characteristics consistent with operation appear on the pressure side, such as stable dynamic pressure and no change in static pressure, it can be determined that the visual result may be biased due to occlusion, blurring, or misidentification. Furthermore, the pressure side fails to provide support, so the system will not confirm the "in operation" state but will keep the area in "pending confirmation / transient state." Conversely, only when the visual system perceives an area as "in operation" or "operation preparation," and the pressure disturbance pattern is consistent with the physical laws of this state, and this consistency is continuously satisfied within the monitoring window, will the system ultimately confirm the area as a valid "in operation" state, achieving mutual verification of visual and pressure evidence.

[0066] The comprehensive demand intensity of each operational sub-area of ​​the target plant is obtained through the following process:

[0067] Extract the probability distribution of each state of each sub-area of ​​the target plant and the dust generation state of each sub-area of ​​the target plant. Based on the probability distribution of each state of each sub-area of ​​the target plant and the dust generation state allocation rules of each sub-area of ​​the target plant, dynamically weight the data to obtain the comprehensive demand intensity of each sub-area of ​​the target plant.

[0068] It should be noted that, within the dust generation monitoring window, the branch pipe dynamic pressure signal sequence and branch pipe static pressure sequence of the sub-region are simultaneously analyzed. Short-term average dynamic pressure change rate, net dynamic pressure change amplitude, average static pressure change rate, and net static pressure change amplitude are extracted. This further marks the first dynamic pressure rise trend and the second static pressure trend of the branch pipe in the sub-region, forming a pressure trend vector. The consistency judgment between visual direction and pressure direction refers to the semantic and physical trend-based synergistic relationship between the "operational state direction" indicated by visual recognition and the "dust generation pressure behavior direction" reflected by the pressure trend. For example, when visual recognition confirms that a sub-region is in a state of intense operation or pre-dust generation action, the pressure side should simultaneously exhibit typical dust generation disturbance characteristics such as a continuous rise in dynamic pressure, an enhanced positive gradient in dynamic pressure, or a slight drop in static pressure. If visual recognition indicates that the sub-region is in a non-operational state, the pressure should show a non-dust-generating state with no significant positive increase in dynamic pressure and stable static pressure. By comparing the matching relationship between visual status labels and pressure status labels, when the visual output status label and pressure status label belong to the same dust generation logic cluster in the same direction, such as "visual effective operation + dynamic pressure increase", "visual preparation operation + static pressure slight decrease", "visual non-operation + static pressure stable", etc., it is determined that the visual and pressure directions are consistent. If contradictory combinations occur, such as "visual strong operation + static pressure stable" or "visual non-operation + dynamic pressure significant increase", it is determined that the two directions conflict. After the direction consistency is formed, the system performs dynamic fusion based on the matching strength of visual confidence and pressure confidence. If the directions are consistent, the visual confidence and pressure confidence are weighted and superimposed to generate an improved dust generation confidence. If the directions conflict, the confidence weights of the two directions are redistributed according to the current scene confidence rules. For example, the visual weight is reduced in the case of smoke plume obstruction or light interference, and the pressure weight is reduced when the pressure sensor noise increases. The system then determines whether the current operation sub-region has reached the dust generation confirmation threshold based on the weight adjustment result.

[0069] The opening degree of the air valves in each sub-area of ​​the target plant is obtained based on the comprehensive demand intensity of each sub-area. The specific process is as follows: the comprehensive demand intensity of each sub-area of ​​the target plant is input into the air volume distribution model of the central controller, and all sub-areas of the target plant are sorted according to the comprehensive demand intensity. Based on the sorting results, combined with the current main static pressure, the outlet air volume of the main fan, and the adjustable margin of the main fan frequency converter, the central controller coordinates the air volume demand relationship between different sub-areas of the target plant globally.

[0070] It should be noted that the allocable margin of the air network is obtained by real-time calculation of the coupling relationship between the main pipe pressure difference and the branch pipe pressure drop, and then the allocable margin is distributed proportionally to all sub-areas in effective operation. To prevent the main fan from being overloaded due to simultaneous operation of multiple areas, the central distributor does not simply superimpose the demand intensity, but normalizes the demand intensity of multiple areas and performs secondary convergence of the demand intensity based on the dynamic changes of the main pipe pressure difference, so that the target air volume allocated to each sub-area can be achieved under the air network stability constraint. After obtaining the target air volume of each operating sub-area, the system calls the branch pipe resistance allocation model to convert the target air volume into the equivalent suction cross section required by the corresponding branch pipe, and further converts it into the target opening degree of the branch pipe damper. During the conversion, the system uses the branch pipe characteristic curve (including valve flow coefficient, local resistance coefficient, flow-opening nonlinear function, etc.) to perform inverse calculation to determine the damper opening degree required to meet the target air volume under the current pressure conditions.

[0071] It should be noted that the real-time calculation of the coupling relationship between the main pipe pressure difference and the branch pipe pressure drop to obtain the distributable margin of the ventilation network is essentially an extension of the aerodynamic balancing technology used in industrial dust removal, HVAC, and mine ventilation. Its basic principle is based on known aerodynamic and pipeline balancing analysis methods in related fields, relying on the combined relationship between the real-time monitored main pipe pressure difference and the pressure drop of each branch pipe. The main pipe pressure difference reflects the total pressure that the fan can provide under the current load, while the static and dynamic pressure changes of each branch pipe reflect the instantaneous consumption of the total pressure by each branch. In practice, it is necessary to obtain the pressure difference between the inlet and end of the branch pipe main pipe as the real-time "total pressure supply capacity" of the ventilation network; subsequently, data for each branch pipe are collected. The static pressure, dynamic pressure, or differential pressure of the branch pipes are accumulated relative to the increments of the previous monitoring window to obtain the "total pressure consumption" of all branches within that window. Then, the real-time pressure supply capacity of the main trunk is subtracted from the instantaneous pressure consumption of all branches to obtain the pressure range that is currently remaining and not yet occupied by existing operations; this pressure range is defined as the "allocated margin." This system utilizes the natural physical laws of pressure conservation and load coupling in the wind network: when a branch pipe experiences a large suction demand, it consumes more differential pressure, and the available differential pressure of other branches decreases accordingly; conversely, when the disturbances in each branch are small, the difference between the differential pressure of the main trunk and the differential pressure of each branch pipe naturally increases, allowing the system to determine that the wind network has a significant amount of remaining dispatchable capacity.

[0072] It should be noted that when multiple sub-areas simultaneously enter effective operation, the system first extracts the overall demand intensity of all sub-areas of the target plant. The sum of these initial demand intensities reflects the upward trend of overall airflow demand. However, since the trunk pressure differential directly reflects the real-time load boundary of the fans, the system cannot rely solely on the initial demand intensity. Instead, it needs to perform secondary convergence of the demand intensity based on the dynamic changes in the trunk pressure differential. Specifically, the system continuously monitors the pressure differential curve between the trunk inlet and outlet and compares it with the normal operating bandwidth of the main fans. When the trunk pressure differential shows a continuous decrease or approaches the minimum pressure differential operating threshold, it indicates that the current air network load is nearing its limit. At this point, the initial demand intensity of each sub-area is automatically compressed proportionally, ensuring that the overall demand intensity remains within a stable range after convergence compared to the trunk pressure differential. Conversely, when the trunk pressure differential is ample, the demand intensity of some areas is allowed to recover to its original value, thereby fully utilizing the remaining air network capacity. Through this dynamic convergence mechanism that uses the trunk pressure difference as feedback, multi-region demand coordination of wind network load can be achieved without the need for complex models, so that the final allocated target air volume not only respects the regional operation intensity, but also does not damage the overall stability of the wind network.

[0073] It should be noted that the branch pipe resistance distribution model essentially utilizes the real-time pressure conditions of the branch pipe, valve structural characteristics, and local resistance characteristics to inversely calculate the target air volume into the corresponding equivalent suction cross-section, and then further converts it into the valve opening. This process relies on the real-time dynamic and static pressure provided by sensors, as well as the actual flow rate corresponding to the current valve position. Combined with the valve's characteristic curves (such as the nonlinear relationship between opening and flow rate, and the influence of local resistance on pressure loss), the valve opening required to meet the target air volume can be calculated by inversely looking up the characteristic curve or performing simple table lookup interpolation. In other words, it does not directly predict the flow rate, but rather uses real-time pressure feedback to inversely deduce the equivalent flow area that the valve needs to provide, so that the pressure loss of the branch pipe under the target air volume is balanced with the current pressure conditions of the air network. This enables real-time, closed-loop, and physically consistent valve opening adjustment, ensuring stable and controllable air volume distribution without relying on complex air network simulation models.

[0074] A global consistency check was performed on the opening of all branch pipe dampers. The check included: (1) whether the set of damper openings could be satisfied under the current maximum air volume of the main fan. (2) whether the superposition of the openings of all sub-area dampers would cause the main static pressure to drop below the safe lower limit of the main pressure. (3) whether there was a risk of backflow or flow in the bypass area due to excessive opening of a branch pipe.

[0075] If the verification passes, the target opening degree is output as the final damper opening degree. If the verification fails, the central controller automatically triggers the opening degree convergence process, which reduces the target opening degree in low-demand areas and limits the opening degree increase in some medium-demand areas, so that the overall air volume is redistributed within the allowable range of the main pipeline pressure. After the convergence is completed, the air network differential pressure is checked again. If the air network stability condition is met, the final damper opening degree is output. If it is still not met, the global opening degree distribution is further compressed until it enters the safe operating range of the air network.

[0076] To prevent dust accumulation in long-term closed branch pipes, the system is set to "inspection mode". During non-operational periods (such as maintenance breaks), each branch pipe is turned on in turn and the fan is run at full speed for 5-10 minutes to perform "cleaning and blowing".

[0077] It should be noted that after triggering the damper opening adjustment, the damper opening is tracked and monitored to obtain the following relationship characteristics between the damper opening change and the corresponding branch pipe dynamic pressure response. Based on the following relationship characteristics, the flow limiting control of the corresponding branch pipe damper is triggered. This is achieved by extracting the correspondence characteristics between the damper opening change rate and the branch pipe dynamic pressure change trend. This allows the system to determine whether the current damper adjustment is truly and effectively converted into an increase in ventilation capacity, and to distinguish between "ineffective opening" or "excessive air supply" caused by pipeline blockage, sudden changes in local wind resistance, or multi-area linkage interference. Specifically, after triggering the damper opening adjustment for a specific sub-area of ​​the target plant, the system synchronously collects the actual opening change sequence of the damper and the real-time dynamic pressure response sequence of the corresponding branch pipe within a preset time window. The cumulative change in damper opening and the cumulative change in corresponding branch pipe dynamic pressure within the preset time window are calculated respectively. Based on the cumulative change in damper opening and the cumulative change in corresponding branch pipe dynamic pressure, each is divided by the preset time window length to obtain the cumulative change rate of damper opening and the corresponding dynamic pressure response. The cumulative rate of change of branch pipe dynamic pressure is monitored. If either the cumulative rate of change of the damper opening or the cumulative rate of change of the corresponding branch pipe dynamic pressure is less than a set threshold, an early warning is sent to the management terminal. The set thresholds include the cumulative rate of change thresholds for both the damper opening and the branch pipe dynamic pressure. Otherwise, the cumulative rates of change of the damper opening and the corresponding branch pipe dynamic pressure are normalized. If the rate deviation between the cumulative rate of change of the damper opening and the corresponding branch pipe dynamic pressure is within a preset allowable deviation range, the current target opening control state of the damper is maintained. Conversely, if the cumulative rate of change of the damper opening is less than the cumulative rate of change of the corresponding branch pipe dynamic pressure, a target opening increase request signal is transmitted to the management terminal. If the cumulative rate of change of the damper opening is greater than the cumulative rate of change of the corresponding branch pipe dynamic pressure, flow restriction control of the corresponding branch pipe damper is triggered. Before executing flow restriction control, the target opening is reduced by a preset opening reduction value to obtain the target opening after flow restriction and then controlled.

[0078] Based on this, the flow restriction control setting can determine the actual air intake demand state according to the dynamic matching characteristics between the damper adjustment behavior and the dynamic pressure response of the air network. This avoids ineffective air intake caused by excessive air volume under low dust generation or low load conditions, while ensuring sufficient air volume supply in a timely manner under high dust generation or high resistance conditions. The judgment is based on the relative relationship between the damper adjustment rate and the dynamic pressure response rate, making the air volume regulation process more in line with the air network operation mechanism and improving the targeting and stability of the damper adjustment. It should be noted that under normal operating conditions, there is a unidirectional response relationship between the change in damper opening and the corresponding change in branch pipe dynamic pressure. That is, when the damper opening increases, the branch pipe dynamic pressure increases accordingly, and when the damper opening decreases, the branch pipe dynamic pressure decreases accordingly. Therefore, in this embodiment, when calculating the cumulative change rate of damper opening and the cumulative change rate of branch pipe dynamic pressure, the data is normalized and statistically analyzed based on the absolute changes of the two, thereby ensuring that the rate comparison reflects the matching relationship between the magnitudes of the changes.

[0079] like Figure 6 As shown, the second aspect of the present invention also provides a flowchart of a dust collector energy-saving control method based on visual recognition and working condition linkage. The method includes: collecting image sequences of each working sub-area of ​​the target plant and importing them into a lightweight visual recognition model to obtain the working recognition results of each working sub-area of ​​the target plant, and performing time sequence stability correction on the working recognition results of each working sub-area of ​​the target plant.

[0080] By visually judging and comparing the real-time air network feedback data of each sub-area of ​​the target plant, the situation fusion criteria are used to determine whether each sub-area of ​​the target plant has entered an effective dust generation state. The dynamic pressure situation of the branch pipe and the second static pressure situation of the branch pipe are linked and matched to obtain the comprehensive demand intensity of each sub-area of ​​the target plant.

[0081] The opening degree of the air valves in each sub-area of ​​the target plant is obtained based on the comprehensive demand intensity of each sub-area of ​​the target plant, and energy-saving coordinated allocation is performed on the air volume demand relationship of each sub-area of ​​the target plant.

[0082] After triggering the damper opening regulation, the damper opening is tracked and monitored to obtain the following relationship characteristics between the damper opening change and the corresponding branch pipe dynamic pressure response, and the flow limiting control of the corresponding branch pipe damper is triggered based on the following relationship characteristics.

[0083] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0084] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0085] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0086] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0087] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0088] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0089] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0090] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0091] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0092] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A dust collector energy-saving control system based on visual recognition and operating condition linkage, characterized in that, The system includes: The lightweight visual operation recognition and temporal stability correction module is used to collect image sequences of each operation sub-area of ​​the target factory and import them into the lightweight visual recognition model to obtain the operation recognition results of each operation sub-area of ​​the target factory, and to perform temporal stability correction on the operation recognition results of each operation sub-area of ​​the target factory. The process of collecting image sequences of each operational sub-area of ​​the target factory building and importing them into a lightweight visual recognition model to obtain the operational recognition results of each operational sub-area of ​​the target factory building is as follows: Local gradient increment data of each sub-region of the target factory are extracted, including the brightness change amplitude of pixels, gradient increment density, average velocity of optical flow field, and texture block structure offset. A unified local gradient increment vector is established and input into the trained lightweight visual recognition model. The output is the operation recognition result of each sub-region of the target factory. The operation recognition result of each sub-region of the target factory includes the probability distribution of each state of each sub-region of the target factory. The specific process for performing time-series stability correction on the operation identification results of each sub-area of ​​the target factory is as follows: A time-series stabilization window is preset. Within the time-series stabilization window, the state probabilities of the lightweight visual recognition model for each sub-region of the target factory are aggregated and statistically analyzed to obtain the gradient change rate and local variance of the state probabilities for each sub-region of the target factory. The gradient change rate and local variance of the state probabilities for each sub-region of the target factory are added together to obtain the stable value of the state probabilities for each sub-region of the target factory. Extract the stable values ​​of the state probabilities of each sub-area of ​​the target plant, and count the minimum duration of the state probabilities of each sub-area of ​​the target plant in the time-series stability window. Combine this with the minimum duration threshold of the state probabilities stored in the database to obtain the effective operation status of each sub-area of ​​the target plant. The multi-source wind network situation fusion and dust generation status determination module is used to compare and correlate real-time wind network feedback data of each working sub-area of ​​the target plant through visual judgment, determine whether each working sub-area of ​​the target plant has entered an effective dust generation state through situation fusion criteria, and perform linkage matching analysis on the dynamic pressure situation of the branch pipe and the second static pressure situation of the branch pipe to obtain the comprehensive demand intensity of each working sub-area of ​​the target plant. The damper opening adaptive control and multi-area collaborative allocation module is used to obtain the damper opening of each work sub-area of ​​the target plant based on the comprehensive demand intensity of each work sub-area of ​​the target plant, and at the same time perform energy-saving collaborative allocation on the air volume demand relationship of each work sub-area of ​​the target plant. The control adaptation tracking module is used to track and monitor the valve opening after triggering the valve opening control to obtain the following relationship characteristics between the valve opening change and the corresponding branch pipe dynamic pressure response, and to trigger the flow limiting control of the corresponding branch pipe valve based on the following relationship characteristics.

2. The dust collector energy-saving control system based on visual recognition and operating condition linkage according to claim 1, characterized in that, The specific process for obtaining the effective operational status of each operational sub-area of ​​the target factory is as follows: The stable values ​​of the state probabilities of each sub-area of ​​the target plant are compared with the state probability stability threshold stored in the database. If the stable value of the state probability of a certain sub-area of ​​the target plant is lower than or equal to the stable value state probability stability threshold, and the minimum duration of the state probability of that sub-area of ​​the target plant is higher than or equal to the minimum duration threshold of the state probability, then the sub-area of ​​the target plant is confirmed as a valid operation state and the output is locked. Otherwise, if the minimum duration requirement is not met, the state is regarded as transient and not confirmed.

3. The dust collector energy-saving control system based on visual recognition and operating condition linkage according to claim 1, characterized in that, The process of correlating and comparing real-time ventilation network feedback data from each operational sub-area of ​​the target factory through visual judgment is as follows: Set a preset dust monitoring window to acquire the pressure sequence of the branch pipe dynamic pressure signal and the pressure sequence of the branch pipe static pressure sensor in each operating sub-area of ​​the target plant. In the dust generation monitoring window, differential and trend aggregation are performed on the pressure sequence of the branch pipe dynamic pressure signal of each sub-area of ​​the target plant to obtain the short-term average rate of change and net change amplitude of the dynamic pressure of the branch pipe of each sub-area of ​​the target plant. Within the same dust generation monitoring window, the same short-term trend extraction was performed on the pressure sequence of the branch static pressure sensor to obtain the average rate of change and net magnitude of static pressure. The branch pipe situation of each work sub-area of ​​the target plant is constituted by the average rate of change and net magnitude of change of static or dynamic pressure of the branch pipes.

4. The dust collector energy-saving control system based on visual recognition and operating condition linkage according to claim 3, characterized in that, The branch pipe status of each working sub-area of ​​the target plant includes the first dynamic pressure rising status of the branch pipes in each working sub-area of ​​the target plant and the second static pressure status of the branch pipes in each working sub-area of ​​the target plant. The first dynamic pressure rising trend of the branch pipes in each sub-area of ​​the target plant is defined as follows: when the rate of change of dynamic pressure in the branch pipes of each sub-area of ​​the target plant within the dust generation monitoring window is positive and the net change exceeds the window noise threshold stored in the database, the branch pipes of that sub-area of ​​the target plant are marked as having a first dynamic pressure rising trend. The second static pressure status of the branch pipes in each sub-area of ​​the target plant includes a slightly declining static pressure status and a stable static pressure status. If the net change in static pressure is negative and its absolute value is between the short-term fluctuation threshold of the main trunk and the upper limit of static pressure disturbance noise, the branch pipe of that sub-area of ​​the target plant is marked as having a slightly declining static pressure status. If the net change in static pressure is near zero and the rate of change falls into the stable static range, the branch pipe of that sub-area of ​​the target plant is marked as having a stable static pressure status.

5. The dust collector energy-saving control system based on visual recognition and operating condition linkage according to claim 2, characterized in that, The process of determining whether each sub-area of ​​the target plant has entered an effective dust-generating state through situational fusion criteria is as follows: If the visual recognition of a certain sub-area of ​​operation has confirmed that it is in an effective operation state, then the effective operation state of each sub-area of ​​operation in the target plant is compared with the current branch pipe status of that sub-area of ​​operation in the target plant. If the pressure status is consistent with the current branch pipe status of that sub-area of ​​operation in the target plant, then the dust generation status of each sub-area of ​​operation in the target plant is confirmed; otherwise, the dust generation status of each sub-area of ​​operation in the target plant is not confirmed for the time being.

6. The dust collector energy-saving control system based on visual recognition and operating condition linkage according to claim 1, characterized in that, The specific process for obtaining the comprehensive demand intensity of each operational sub-area of ​​the target plant is as follows: Extract the probability distribution of each state of each sub-area of ​​the target plant and the dust generation state of each sub-area of ​​the target plant. Based on the probability distribution of each state of each sub-area of ​​the target plant and the dust generation state allocation rules of each sub-area of ​​the target plant, dynamically weight the data to obtain the comprehensive demand intensity of each sub-area of ​​the target plant.

7. The dust collector energy-saving control system based on visual recognition and operating condition linkage according to claim 1, characterized in that, The process of determining the valve opening degree of each operational sub-area of ​​the target plant based on the comprehensive demand intensity of each sub-area is as follows: The comprehensive demand intensity of each sub-area of ​​the target plant is input into the air volume distribution model of the central distributor. All sub-areas of the target plant are sorted according to their comprehensive demand intensity. Based on the sorting results, the central distributor, combined with the current main static pressure, the main fan outlet air volume, and the adjustable margin of the main fan inverter, coordinates the air volume demand relationship between different sub-areas of the target plant globally.

8. A method applied to the dust collector energy-saving control system based on visual recognition and operating condition linkage as described in any one of claims 1-7, comprising: Image sequences of each sub-area of ​​the target factory are collected and imported into a lightweight visual recognition model to obtain the operation recognition results of each sub-area of ​​the target factory. Temporal stability correction is then performed on the operation recognition results of each sub-area of ​​the target factory. By visual judgment, the real-time wind network feedback data of each sub-area of ​​the target plant are correlated and compared. The situation fusion criteria are used to determine whether each sub-area of ​​the target plant has entered an effective dust generation state. The dynamic pressure situation of the branch pipe and the second static pressure situation of the branch pipe are linked and matched to obtain the comprehensive demand intensity of each sub-area of ​​the target plant. The opening degree of the air valves in each sub-area of ​​the target plant is obtained based on the comprehensive demand intensity of each sub-area of ​​the target plant, and energy-saving coordinated allocation is performed on the air volume demand relationship of each sub-area of ​​the target plant.

Images