Dust concentration monitoring method, device and equipment of tobacco dust removal room and storage medium

Abstract

The application discloses a dust concentration monitoring method, device and equipment of a tobacco dust removal room and a storage medium, relates to the technical field of tobacco, and acquires original dust concentration data through pump suction type active sampling; carries out time series filtering and outlier rejection on the original data, fills in missing data in the forward direction, extracts concentration time series features after normalization processing; a concentration trend prediction model is constructed based on the preprocessed concentration data, and a dust concentration change trend is predicted; the prediction index is compared with the three-level early warning threshold item by item, the early warning level is judged, and the corresponding graded early warning instruction is generated, so that graded response and accurate early warning are realized; the fan frequency and the air valve opening degree control amount are calculated through the PID control algorithm, the operation state of the dust production equipment and the workshop ventilation parameter are linked and adjusted, and the linkage control execution result is obtained; the control deviation statistical index is calculated based on the dust concentration feedback data, the PID control parameter and the three-level early warning threshold are updated, and the adaptive optimization and long-term stable operation of the system are realized.

Description

Technical Field

[0001] This application relates to the field of tobacco technology, and in particular to a method, apparatus, equipment and storage medium for monitoring dust concentration in a tobacco dust removal room. Background Technology

[0002] Tobacco production is a core part of cigarette manufacturing, encompassing multiple processes such as tobacco leaf rehydration, shredding, drying, screening, and conveying. During production, processing, and conveying, tobacco leaves generate a large amount of fine dust particles. This dust is lightweight, easily dispersed, and highly accumulative; if not treated promptly, it will spread and accumulate in and around the dust collection room. As the core location for centralized dust collection and treatment, the dust collection room is a crucial hub for the "environmental protection + safety" of the cigarette production line. Controlling the dust concentration inside and at its inlets and outlets directly affects the stability of the entire production process.

[0003] Currently, dust control in some tobacco companies' tobacco dust removal rooms still relies on traditional methods, which has obvious shortcomings: 1. Outdated monitoring methods: The main methods used are manual inspection and sampling or fixed-point test strip testing, which have problems such as low detection frequency, data lag, and limited coverage. They cannot capture the dynamic changes in dust concentration in real time and it is difficult to detect the risk of instantaneous exceedance.

[0004] 2. Significant safety risks: Tobacco dust is a combustible dust. When the dust concentration in the dust removal room reaches or approaches the lower explosive limit, it is very easy to cause dust combustion or even explosion once it encounters ignition sources such as equipment friction sparks, electrostatic discharge, or high-temperature heat sources, resulting in personal injury and equipment damage.

[0005] 3. Occupational health and environmental compliance pressures: Long-term inhalation of tobacco dust can damage the respiratory tract and lungs of operators; in addition, if the dust removal room is not properly sealed or the dust removal efficiency decreases, dust will be emitted into the workshop or outdoors in an unorganized manner. If relevant regulations are violated, the company will face risks such as environmental penalties and production shutdowns.

[0006] 4. Blind operation of equipment and process control: The lack of real-time concentration data makes it impossible to accurately judge problems such as dust collector filter bag damage, dust removal system failure, and insufficient fan air volume, resulting in inefficient operation of the dust removal system. At the same time, excessively high dust concentration may drift into the tobacco processing workshop, adhere to the surface of tobacco and affect product purity, or change the temperature and humidity of the workshop due to dust absorbing moisture, indirectly causing abnormal moisture content of tobacco and affecting cigarette quality. Summary of the Invention

[0007] The main objective of this application is to provide a method, device, equipment, and storage medium for monitoring dust concentration in tobacco dust removal rooms, in order to solve the problems of outdated monitoring methods, prominent safety risks, and blind operation and process control in the prior art.

[0008] To achieve the above objectives, this application provides the following technical solution: A method for monitoring dust concentration in a tobacco dust removal chamber, wherein dust concentration sensors are deployed inside the dust removal chamber, at pipe joints, and at dust sources; the tobacco dust removal chamber also has a dust removal fan connected to the outside; the dust concentration monitoring method includes: Step S1: Obtain the original dust concentration data of the tobacco dust removal room through the pump-type active sampling of the dust concentration sensor; Step S2: Perform data preprocessing on the original dust concentration data and extract concentration time-series features to obtain a preprocessed concentration dataset; Step S3: Construct a concentration trend prediction model based on the preprocessed concentration dataset, and predict the dust concentration trend through the concentration trend prediction model to construct a dust concentration trend prediction index set; Step S4: Compare the dust concentration trend prediction index set with the preset three-level early warning threshold, and output the corresponding graded early warning instruction; Step S5: Based on the graded early warning command and the current concentration deviation, the frequency of the dust removal fan and the opening of the air valve are dynamically adjusted through the PID control algorithm, and the operating status of the dust generating equipment and the workshop ventilation parameters are adjusted in conjunction, and the linkage control execution result is obtained by combining them. Step S6: Calculate the control deviation statistical index by using the acquired dust concentration feedback data and the linkage control execution result, and update the control parameters of the PID control algorithm and the preset three-level early warning threshold according to the control deviation statistical index to obtain the iterative monitoring parameter set.

[0009] Beneficial effects of steps S1 to S6: This method addresses the issues of insufficient monitoring coverage and delayed response in existing technologies. It simultaneously acquires raw dust concentration data indoors, at pipe nodes, and at dust sources through pump-assisted active sampling. A concentration trend prediction model is constructed to enable early prediction, avoiding the lag in post-event alarms. A three-level early warning threshold system is established and linked with predictive indicators to achieve tiered response and precise early warning. A PID control algorithm, in collaboration with multiple systems, dynamically adjusts the dust collector fan, linked dust-generating equipment, and ventilation parameters for timely handling. Based on feedback data, the PID parameters and early warning thresholds are iteratively updated to construct an adaptive optimization closed loop, ensuring long-term stable operation of the system.

[0010] Step S1 involves acquiring raw dust concentration data through pump-assisted active sampling, enabling simultaneous monitoring and real-time data acquisition at multiple locations. Step S2 performs time-series filtering and outlier removal on the raw data, fills in missing data, and extracts concentration time-series features after normalization, providing high-quality data input for trend prediction. Step S3 constructs a concentration trend prediction model based on the preprocessed concentration data to predict dust concentration changes, achieving early warning and avoiding delayed response. Step S4 compares the predicted indicators with the three-level warning thresholds item by item, determines the warning level, and generates corresponding graded warning instructions, achieving graded response and precise warning. Step S5 calculates the fan frequency and valve opening control quantities using a PID control algorithm, and adjusts the operating status of dust-generating equipment and workshop ventilation parameters in a coordinated manner, merging the results of the coordinated control execution to achieve automatic handling and timely response. Step S6 calculates control deviation statistics based on dust concentration feedback data, updates PID control parameters and the three-level warning thresholds, and achieves adaptive optimization and long-term stable operation of the system.

[0011] As a further improvement to this application, step S1, obtaining the original dust concentration data of the tobacco dust removal chamber through the pump-type active sampling of the dust concentration sensor, includes: Step S1.1: Trigger pump-type sampling at preset sampling intervals, and encapsulate the collected raw dust concentration data in a preset data format to obtain raw encapsulated data; Step S1.2: Remove spike noise and transient interference from the original packaging data to obtain the original dust concentration data.

[0012] Beneficial effects of steps S1.1 to S1.2: This series of steps uses timed triggering of pump-suction active sampling to encapsulate the collected data in a preset format and remove spike noise and transient interference through pre-filtering, enabling simultaneous acquisition of raw dust concentration data from multiple points, ensuring monitoring coverage and data real-time performance, and providing a clean and reliable data source for subsequent data preprocessing.

[0013] In step S1.1, pump-type sampling is triggered at preset sampling intervals, and the collected data is packaged in a preset data format to achieve standardized acquisition of the original packaged data. In step S1.2, peak noise identification and instantaneous interference filtering are used to effectively remove instantaneous jumps in the sensor and environmental interference signals to obtain clean original dust concentration data.

[0014] As a further improvement to this application, step S2 involves preprocessing the original dust concentration data and extracting concentration time-series features to obtain a preprocessed concentration dataset, including: Step S2.1: Perform time-series filtering on the original dust concentration data to obtain the filtered concentration sequence; Step S2.2: Based on the 3σ criterion, outlier detection and removal are performed on the filtered concentration sequence, and abnormal data points with sudden concentration changes are marked. Step S2.3: Perform forward filling on the missing data segments after removing outliers; Step S2.4: Normalize the filled concentration data and map the concentration values ​​to a preset value range. Step S2.5: Extract concentration time-series features based on the normalized concentration sequence to obtain concentration time-series feature data; Step S2.6: Package the concentration time-series feature data into a preprocessed concentration dataset in a unified format.

[0015] Beneficial effects of steps S2.1 to S2.6: This series of steps removes high-frequency noise through time-series filtering, detects and removes outliers based on the 3σ criterion, fills in missing data forward, extracts concentration time-series features after normalization, and packages them into a unified format dataset, thereby improving data quality and standardizing features, and providing high-quality input for trend prediction.

[0016] In this process, step S2.1 smooths the original concentration sequence using moving average filtering or Kalman filtering to effectively suppress random noise; step S2.2 marks outlier data points that exceed the statistical confidence interval based on the 3σ criterion to avoid polluting subsequent analysis with outlier data; step S2.3 performs forward imputation on missing data segments to maintain temporal continuity; step S2.4 maps concentration values ​​to a unified numerical range using Min-Max normalization to eliminate dimensional differences; step S2.5 extracts multi-dimensional time-series features such as rate of change, fluctuation characteristics, and periodic characteristics to enrich the input information of the prediction model; and step S2.6 packages the concentration time-series feature data into a unified format for use in step S3.

[0017] As a further improvement to this application, step S3 involves constructing a concentration trend prediction model based on the preprocessed concentration dataset, and predicting dust concentration trends using the concentration trend prediction model to construct a dust concentration trend prediction index set, including: Step S3.1: Construct a concentration time-series feature matrix based on the preprocessed concentration dataset; Step S3.2: Train and construct the concentration trend prediction model based on the concentration time-series feature matrix; Step S3.3: Predict the dust concentration trend within a preset prediction period using the trained concentration trend prediction model. Step S3.4: Calculate the dust concentration change rate, cumulative offset, and peak prediction value based on the prediction results; Step S3.5: Integrate the rate of change, the cumulative offset, and the peak predicted value into the dust concentration trend prediction index set.

[0018] Beneficial effects of steps S3.1 to S3.5: This series of steps constructs a concentration time-series feature matrix, trains an LSTM or Transformer trend prediction model, predicts the dust concentration trend within a preset prediction period, calculates the rate of change, cumulative offset, and peak prediction value, and integrates them into an index set to achieve early warning and forward-looking decision support.

[0019] In step S3.1, the preprocessed concentration dataset is arranged into a feature matrix according to the time series and used as the input of the prediction model; step S3.2, the temporal dependencies are captured based on the LSTM or Transformer model to construct a trend prediction model with long-term memory capability; step S3.3, the concentration prediction sequence within a preset time period is output through the trained model; step S3.4, three core indicators, namely the rate of change, cumulative offset, and peak prediction value, are calculated to quantify the concentration change trend; step S3.5, the three indicators are integrated into a structured indicator set to provide standardized input for the comparison and judgment in step S4.

[0020] As a further improvement to this application, step S4 compares the dust concentration trend prediction index set with the preset three-level early warning threshold and outputs the corresponding graded early warning instruction, including: Step S4.1: Based on the lower limit of dust explosion, three levels of warning thresholds are set respectively, wherein the first-level warning threshold, the second-level warning threshold, and the third-level warning threshold correspond to different preset percentages of the lower limit of dust explosion; Step S4.2: Compare the dust concentration trend prediction index set with the three-level early warning threshold item by item to determine the current early warning level; Step S4.3: Generate the corresponding graded early warning instruction based on the current early warning level.

[0021] Beneficial effects of steps S4.1 to S4.3: This series of steps sets three levels of early warning thresholds, compares the predicted indicators with the thresholds one by one, determines the early warning level, and generates corresponding graded early warning instructions, thereby achieving refined management and differentiated response of early warning levels.

[0022] Specifically, step S4.1 sets three-level warning thresholds based on the percentage of the lower limit of dust explosion, dividing them into three levels: reminder value, handling value, and emergency value; step S4.2 compares the dust concentration trend prediction index set with the three-level warning thresholds item by item to accurately determine the current warning level; step S4.3 generates corresponding graded warning instructions according to the warning level, providing an execution basis for the linkage control in step S5.

[0023] As a further improvement to this application, step S5 involves dynamically adjusting the frequency of the dust removal fan and the opening of the damper based on the deviation between the graded early warning command and the current concentration using a PID control algorithm, and simultaneously adjusting the operating status of the dust-generating equipment and the workshop ventilation parameters, merging the results of the coordinated control execution to obtain the following: Step S5.1: Calculate the current concentration deviation value based on the graded early warning instruction and the current measured dust concentration; Step S5.2: Based on the current concentration deviation value, calculate the frequency control quantity of the dust removal fan and the opening control quantity of the damper using the PID control algorithm; Step S5.3: Based on the graded early warning instruction and the current concentration deviation value, adjust the operating status of the dust-generating equipment and the workshop ventilation parameters in a coordinated manner; Step S5.4: Summarize the execution results of the dust removal fan, the air valve, the dust generating equipment and the workshop ventilation parameters, and merge them to obtain the linkage control execution result.

[0024] Beneficial effects of steps S5.1 to S5.4: This series of steps calculates the concentration deviation value, calculates the fan frequency and damper opening control quantity based on the PID control algorithm, and adjusts the dust-generating equipment and ventilation parameters in a coordinated manner. The results are then summarized to obtain the coordinated control execution result, thereby achieving automatic adjustment and timely handling of dust concentration.

[0025] Specifically, step S5.1 calculates the current concentration deviation value based on the graded early warning command and the measured concentration, which serves as the input signal for PID control; step S5.2 calculates the frequency control quantity of the dust collector fan and the opening control quantity of the air valve through the PID control algorithm to achieve precise adjustment of continuous quantities; step S5.3 adjusts the start-stop status of the dust-generating equipment and the workshop ventilation parameters in conjunction with the graded early warning command to achieve coordinated handling of multiple systems; and step S5.4 summarizes the execution results of the dust collector fan, air valve, dust-generating equipment, and ventilation parameters, and merges them to obtain the linkage control execution result.

[0026] As a further improvement to this application, in step S6, a control deviation statistical index is calculated using the acquired dust concentration feedback data and the execution result of the linkage control. This index is then used to update the control parameters of the PID control algorithm and the preset three-level early warning threshold, resulting in an iteratively set of monitoring parameters, including: Step S6.1: Obtain the dust concentration feedback data at the current moment through the dust concentration sensor; Step S6.2: Compare the dust concentration feedback data with the target concentration value in the linkage control execution result, and calculate the control deviation statistical index; Step S6.3: Update the proportional coefficient, integral coefficient, and derivative coefficient of the PID control algorithm according to the control deviation statistical index, and adjust the preset three-level early warning threshold according to the triggering of the graded early warning command; Step S6.4: The updated control parameters and the three-level early warning threshold are encapsulated into the iterative monitoring parameter set for use in the next monitoring cycle.

[0027] Beneficial effects of steps S6.1 to S6.4: This series of steps involves acquiring dust concentration feedback data, calculating control deviation statistics, updating PID control parameters and three-level early warning thresholds, and encapsulating them into an iterative monitoring parameter set to achieve adaptive optimization and long-term stable operation of the system.

[0028] Specifically, step S6.1 involves acquiring feedback data in real time through a dust concentration sensor to monitor the linkage control effect; step S6.2 compares the feedback data with the target concentration value to calculate control deviation statistical indicators such as MAE or RMSE; step S6.3 adjusts the three coefficients of the PID controller based on the control deviation statistical indicators and fine-tunes the three-level warning thresholds according to the triggering of the graded warning instructions; and step S6.4 encapsulates the updated control parameters and thresholds into an iterative monitoring parameter set for use in the next monitoring cycle to achieve closed-loop adaptive optimization.

[0029] To achieve the above objectives, this application also provides the following technical solutions: A dust concentration monitoring device for a tobacco dust removal chamber, wherein the dust concentration monitoring device is applied to the dust concentration monitoring method described above, and the dust concentration monitoring device comprises: The raw dust concentration data acquisition module is used to obtain the raw dust concentration data of the tobacco dust removal room through the pump-type active sampling of the dust concentration sensor. The data preprocessing module is used to preprocess the original dust concentration data and extract the concentration time series features to obtain a preprocessed concentration dataset. The dust concentration trend prediction module is used to construct a concentration trend prediction model based on the preprocessed concentration dataset, and to predict the dust concentration trend through the concentration trend prediction model to construct a dust concentration trend prediction index set. The graded early warning instruction acquisition module is used to compare the dust concentration trend prediction index set with the preset three-level early warning threshold and output the corresponding graded early warning instruction. The dust linkage processing module is used to dynamically adjust the frequency of the dust removal fan and the opening of the air valve based on the deviation between the graded early warning command and the current concentration, and to adjust the operating status of the dust generating equipment and the workshop ventilation parameters in a linkage manner, and combine them to obtain the linkage control execution result; The concentration trend prediction model iteration module is used to calculate the control deviation statistical index by acquiring dust concentration feedback data and the linkage control execution result, update the control parameters of the PID control algorithm and the preset three-level early warning threshold according to the control deviation statistical index, and obtain the monitoring parameter set after iteration.

[0030] To achieve the above objectives, this application also provides the following technical solutions: An electronic device includes a processor and a memory coupled to the processor, the memory storing program instructions executable by the processor; when the processor executes the program instructions stored in the memory, it implements the dust concentration monitoring method for a tobacco dust removal chamber as described above.

[0031] To achieve the above objectives, this application also provides the following technical solutions: A computer-readable storage medium storing program instructions, which, when executed by a processor, enable the dust concentration monitoring method for a tobacco dust removal chamber as described above. Attached Figure Description

[0032] Figure 1 This is a schematic flowchart illustrating the steps of an embodiment of a method for monitoring dust concentration in a tobacco dust removal room according to this application; Figure 2 This is a schematic diagram of the functional modules of a dust concentration monitoring device for a tobacco dust removal room according to an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an embodiment of the electronic device of this application; Figure 4 This is a schematic diagram of the structure of one embodiment of the storage medium of this application. Detailed Implementation

[0033] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0034] The terms "first," "second," and "third" in this application are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationships and movements between components in a specific orientation (e.g., as shown in the figures). If the specific orientation changes, the directional indications also change accordingly. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0035] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0036] It should be noted that, due to the limited types and number of symbols or letters that can represent specific meanings, for embodiments with many formulas or codes, there may be situations where symbols or letters cannot meet the usage requirements. Therefore, the interpretation of formula symbols in the steps or sub-steps of the embodiments is only valid for the current step or sub-step.

[0037] If the same symbol has different interpretations in different steps or sub-steps, the interpretation in the current step or sub-step shall prevail; if the same symbol appears in different steps or sub-steps, but no interpretation is given in subsequent steps or sub-steps after its first appearance, the interpretation in the first step or sub-step shall be used.

[0038] For example, "i=1,2,...,n" is a writing convention and a well-known meaning. If "i=1,2,...,n" has already appeared once in an embodiment and needs to be used again in formulas with different scenarios and meanings in the future, then the meaning of "i=1,2,...,n" should be interpreted according to the corresponding formulas. If "i=1,2,...,n" is not used for the first time and is changed to symbols or letters with non-well-known meanings, such as "h=1,2,...,q", to avoid repetition, it is more likely to cause confusion, ambiguity, and unclear problems.

[0039] like Figure 1 As shown, this embodiment provides an example of a method for monitoring dust concentration in a tobacco dust removal room. In this embodiment, dust concentration sensors are deployed in the dust removal room, at pipe nodes, and at dust sources. The tobacco dust removal room also has a dust removal fan that is connected to the outside.

[0040] Specifically, the dust concentration monitoring method includes the following steps: Step S1: Obtain the original dust concentration data of the tobacco dust removal room through active sampling by the dust concentration sensor using a pump-type suction method.

[0041] Furthermore, step S1 specifically includes the following steps: Step S1.1: Obtain the original dust concentration data of the tobacco dust removal room through the pump-suction active sampling of the dust concentration sensor, including: triggering pump-suction sampling at preset sampling intervals, and encapsulating the collected original dust concentration data in a preset data format to obtain the original encapsulated data.

[0042] Preferably, the core of this step is to configure and time the pump-suction active sampling parameters, and to achieve real-time data acquisition through the pump-suction active sampling of the explosion-proof dust concentration sensor.

[0043] The sampling interval is set according to the actual production conditions, preferably between 5 and 30 seconds; the sampling duration is between 10 and 60 seconds to ensure that enough representative samples are collected; the data format adopts a preset encapsulation protocol, including three fields: timestamp, sensor number, and original concentration reading, in the format YYYY-MM-DD-HH:MM:SS-XXX,SENSOR-ID-XXX,CONC-VALUE-XXX.

[0044] The working principle of pump-suction sampling is as follows: after the sampling pump is started, a negative pressure is formed in the sampling pipeline, which continuously draws the dust-laden gas in the dust removal room into the sampling chamber through the pump inlet. The dust concentration sensor measures the concentration of the dust-laden gas in real time in the sampling chamber and outputs a concentration analog signal or digital signal.

[0045] The sampling triggering adopts a dual-mode parallel approach of timed triggering and abnormal triggering: In timed triggering mode, the controller automatically starts pump-type sampling according to the preset sampling interval; in abnormal triggering mode, when the concentration change rate in the previous sampling cycle exceeds the preset threshold, a new round of sampling is immediately triggered without waiting for the timed cycle.

[0046] The relationship between pump suction flow rate and pipeline resistance is as follows: .

[0047] Where Q is the pump suction flow rate, in L / min; The negative pressure value provided by the pump, in Pa; The flow resistance of the sampling pipe is directly proportional to the pipe length and inversely proportional to the fourth power of the pipe's inner diameter.

[0048] The data is encapsulated in a structured CSV format, specifically including a timestamp field, a sensor number field, and a concentration reading field. Fields are separated by commas, and each record ends with a newline character. The data encapsulation format is as follows: "(field1)Timestamp:YYYY-MM-DD-HH:MM:SS-XXX (field2)SensorID:SENSOR-ID-XXX (field3)Concentration:CONC-VALUE-XXXmg / m^{3}".

[0049] Preferably, the sampling frequency is 1Hz, i.e., the sampling interval is 1 second; the duration of each sampling is 30 seconds; the pump flow rate is 2L / min; the inner diameter of the pump pipe is 6mm; the maximum pipe length is 10m; the data encapsulation format is CSV; the sampling triggering mode is a dual-mode parallel system of timed triggering and abnormal triggering, with the abnormal triggering threshold being a concentration change rate exceeding 0.5mg / m³. 3 / s.

[0050] For example, in the dust removal room of a tobacco processing workshop, sampling is triggered at a timed interval of YYYY-MM-DD-HH:MM:SS-001. The sensor SENSOR-DUST-001 continuously collects data within a 30-second sampling period. The instantaneous concentration value at the end of the sampling is taken as the sampling result: CONC-VALUE-001 = 15.7 mg / m³. 3 The timestamp and sensor number are encapsulated into a single record: YYYY-MM-DD-08:30:00-001,SENSOR-DUST-001,15.7.

[0051] Step S1.2: Remove spike noise and transient interference from the original packaging data to obtain the original dust concentration data.

[0052] Preferably, the core of this step is to identify spike noise and filter out instantaneous interference, and obtain clean raw dust concentration data through pre-filtering.

[0053] The sources of spike noise mainly include electronic noise inside the sensor, random collisions of dust particles in the sampling pipeline, and signal jumps caused by electromagnetic interference; the sources of transient interference mainly include sudden increases in dust concentration during dust collector cleaning, airflow disturbances in the pipeline caused by equipment start-up and shutdown, and local dust diffusion caused by personnel moving near the sensor.

[0054] Spike noise is identified using a sliding window mean comparison method: the latest sampled value is compared with the mean of the previous N samples; if the absolute value of the difference exceeds a preset multiple of the mean, it is identified as spike noise. Let the sampling sequence be... The mean of the first N samples is: .

[0055] The criteria for determining the peak are: .

[0056] Where k is the peak determination factor. When the above conditions are met, Marked as spike noise, and used Replace its value.

[0057] The transient interference is determined using a sliding window variance detection method: the variance of M consecutive sampled values ​​is calculated, and if the variance exceeds a preset threshold, transient interference is considered to exist. Let the sampling sequence within the sliding window be... The window variance is: .

[0058] in This is the average of samples taken within the window. When... If the interference is instantaneous, all data within the window will be smoothed.

[0059] The smoothing process uses an exponentially weighted moving average method: .

[0060] in For the smoothed output value, This is a smoothing coefficient, with a value ranging from 0 to 1. The smaller the value, the stronger the smoothing effect.

[0061] The complete pre-filtering process is as follows: For each newly arrived raw encapsulated data, the concentration value is first parsed; then, peak detection is performed, and if a peak is detected, the mean of the previous N values ​​is used instead; next, window variance detection is performed, and if transient interference is detected, exponential weighted smoothing is performed; finally, clean raw dust concentration data is output.

[0062] Preferably, the peak determination window length N is 5 samples; the peak determination factor k is 3; the instantaneous interference determination window length M is 5 seconds; and the variance threshold... It is 0.5 Exponential smoothing coefficient The value is 0.3; the pre-filtering process is completed within the PLC or edge computing unit of the data acquisition terminal, and the processing delay does not exceed one sampling cycle.

[0063] For example, a sensor's five consecutive sample values ​​are 12.3, 12.5, 15.7, 12.8, and 13.1. The average of the first four times The difference between the 5th sample value of 12.8 and the mean is |12.8-13.075|=0.275, which meets the peak determination criteria. This is invalid; it is considered normal data, so 12.8 is output directly. In another scenario, the five sample values ​​were 12.3, 12.5, 35.2, 12.8, and 13.1 mg / m³, respectively. The average of the first four samples was... The difference between the 5th sample value of 35.2 and the mean is |35.2-13.075|=22.125, which meets the peak determination criteria. This also does not hold true, but the window variance is calculated. If the value exceeds the threshold of 0.5, it is determined to be an instantaneous disturbance. Exponential weighted smoothing is then performed, and the output is smoothed with a coefficient of 0.3.

[0064] Beneficial effects of steps S1.1 to S1.2: This series of steps uses timed triggering of pump-suction active sampling to encapsulate the collected data in a preset format and remove spike noise and transient interference through pre-filtering, enabling simultaneous acquisition of raw dust concentration data from multiple points, ensuring monitoring coverage and data real-time performance, and providing a clean and reliable data source for subsequent data preprocessing.

[0065] In step S1.1, pump-type sampling is triggered at preset sampling intervals, and the collected data is packaged in a preset data format to achieve standardized acquisition of the original packaged data. In step S1.2, peak noise identification and instantaneous interference filtering are used to effectively remove instantaneous jumps in the sensor and environmental interference signals to obtain clean original dust concentration data.

[0066] Step S2: Perform data preprocessing on the original dust concentration data and extract the concentration time series features to obtain the preprocessed concentration dataset.

[0067] Furthermore, step S2 specifically includes the following steps: Step S2.1: Perform time-series filtering on the original dust concentration data to obtain the filtered concentration sequence.

[0068] Preferably, the core of this step is to perform time-series filtering, which smooths the original concentration sequence using a moving average filter or a Kalman filter.

[0069] The principle of the moving average filter is as follows: a sliding window of length L is taken, and the arithmetic mean of all sampled values ​​within the window is calculated and output. The calculation is repeated every time the window advances by one sample point. The moving average filter has a good suppression effect on high-frequency random noise, but it introduces a certain phase delay, which is (L-1) / 2 sample points. The Kalman filter is suitable for dynamic systems with process noise and measurement noise. By alternating between prediction and update steps, it can adaptively adjust the filter gain and is widely used in industrial process control.

[0070] The moving average filtering formula is: .

[0071] Where y_n is the filtered output value at time n, and L is the window length. Let be the input value at time ni.

[0072] The recursive formula for Kalman filtering includes a prediction step and an update step:

[0073]

[0074]

[0075]

[0076] .

[0077] in For prior state estimation, For the posterior state estimation, P is the estimation error covariance matrix, K is the Kalman gain, F is the state transition matrix, H is the observation matrix, Q is the process noise covariance, R is the measurement noise covariance, and z is the observed value.

[0078] Preferably, a moving average filter is used, with a window length L of 10 sampling points; when the sampling interval is 5 seconds, the corresponding filtering time is 50 seconds; Kalman filter parameters: state transition matrix F is 1, observation matrix H is 1, process noise covariance Q is 0.01, and measurement noise covariance R is 0.1.

[0079] Step S2.2, based on The criteria perform outlier detection and removal on the filtered concentration sequence, and mark abnormal data points with sudden concentration changes.

[0080] Preferably, the core of this step is completed. Criteria for anomaly detection; calculation of the mean of the filtered concentration sequence. with standard deviation It will exceed Data points within a range are marked as outliers.

[0081] in, The criterion is based on the normal distribution assumption, under which data fall within the range of... The probability within the interval is 99.7%, so data points outside the interval can be considered statistical anomalies. After being marked, abnormal data points are not deleted directly, but are left for forward filling in step S2.3. The mean and standard deviation are calculated using an exponentially weighted moving average update method to adapt to gradual changes in concentration and avoid misjudging normal data as anomalies when the concentration trend changes abruptly.

[0082] The exponentially weighted update formula for the mean and standard deviation is:

[0083] .

[0084] in It is the exponential weighting coefficient, with a value ranging from 0 to 1. The closer it is to 1, the greater the weight of historical data and the slower the update.

[0085] The formula for outlier determination is: .

[0086] When the above conditions are met, Mark as an outlier.

[0087] Preferably, The multiplier is 3; the exponential weighting factor The initial values ​​are 0.95; the mean and standard deviation are calculated based on the first 100 valid sampling points; the anomaly markers include three attributes: anomaly type (upper bound / lower bound), anomaly time, and anomaly value size.

[0088] Step S2.3: Perform forward filling on the missing data segments after removing outliers.

[0089] Preferably, the core of this step is to perform forward filling, whereby missing data points marked as anomalous are filled with the concentration value of the preceding valid data point.

[0090] Forward filling is suitable for sudden missing data in industrial time series data to avoid excessive jumps in filling values ​​that could affect subsequent processing. For data segments with more than a preset number of consecutive missing data segments, average filling is used instead of forward filling to prevent filling values ​​from deviating from reality for a long time due to long-term anomalies.

[0091] The mathematical expression for forward filling is: .

[0092] The formula for imputing the mean when there are more than K consecutive missing values ​​is: .

[0093] Where M is the number of valid historical data points used to calculate the mean, M=10.

[0094] Preferably, the maximum number of consecutive forward filling operations is 3. If there are more than 3 consecutive missing data points, the average of the 10 most recent valid data points is used for filling. After filling is completed, the data point is recorded as "filled" in the dataset for reference in subsequent steps.

[0095] Step S2.4: Normalize the filled concentration data and map the concentration values ​​to a preset value range.

[0096] Preferably, the core of this step is to perform Min-Max normalization, which linearly maps the filled concentration values ​​to a preset interval [0,1] or [0,100], facilitating subsequent feature extraction and model training.

[0097] Among them, normalization eliminates the influence of different physical dimensions on model training, enabling each feature to participate in the calculation on the same scale and accelerating model convergence; and It can be set to a fixed value or use a dynamic update method. The dynamic update method is based on the actual concentration extreme value within the sliding window, which is more suitable for scenarios with large changes in production conditions.

[0098] The Min-Max normalization formula is: .

[0099] in is the normalized concentration value, with a range of [0,1]; x is the original concentration value. and These represent the minimum and maximum concentrations, respectively.

[0100] The inverse normalization formula is used to restore the model prediction results to the actual concentration values: .

[0101] Preferably, the normalized mapping interval is [0,1]; Set to 0 mg / m³; Set to 1.2 times the lower explosion limit concentration of dust; adopt a dynamic update method. Updated based on the maximum concentration value within the most recent 1000 sampling periods, with an update cycle of 1 hour.

[0102] For example, the initial concentration at a certain moment was 18.5. The lower explosive limit benchmark value is 40. =40000 ,but =48000 The normalization result is .

[0103] Step S2.5: Extract concentration time-series features based on the normalized concentration sequence to obtain concentration time-series feature data.

[0104] Preferably, the core of this step is to extract multi-dimensional time-series features, including three dimensions: rate of change features, fluctuation features, and periodic features.

[0105] Among them, the rate of change characteristic is the amount of concentration change per unit time, reflecting the dynamic trend of dust concentration; the fluctuation characteristic is the variance or standard deviation of the concentration sequence, reflecting the degree of dispersion of dust concentration; and the periodicity characteristic is the peak-valley statistics based on a sliding window, reflecting the periodic fluctuation pattern of dust concentration.

[0106] The formula for calculating the rate of change characteristic is: .

[0107] in The rate of change at time n, in units of W represents the length of the calculation window. The sampling interval is denoted as .

[0108] Fluctuation characteristics are calculated using standard deviation: .

[0109] Periodic features employ a peak-valley detection algorithm: the peak value threshold is set as the mean value within the window plus... The threshold for determining the valley value is the mean value within the window minus... The number of peaks and troughs per unit time is used as a periodic feature.

[0110] Preferably, the change rate calculation window length is 6 sampling points (corresponding to 30 seconds); the fluctuation feature statistics window length is 60 sampling points (corresponding to 5 minutes); the periodic feature detection adopts FFT spectrum analysis to extract the main period frequency; the feature vector dimension is 3-dimensional, including change rate, fluctuation feature, and periodic feature.

[0111] Step S2.6: Package the concentration time-series feature data into a preprocessed concentration dataset in a unified format.

[0112] Preferably, the core of this step is to package the data, encapsulating the time-series feature data in a preset format for use in step S3.

[0113] The packaging format includes five fields: timestamp, normalized concentration value, rate of change, fluctuation characteristics, and periodic characteristics, which are encapsulated into a structured data frame. The data frame format is JSON or binary, and it is transmitted to the application layer server using TCP / IP or OPCUA protocol.

[0114] The data frame format is as follows: "(frame)Timestamp:YYYY-MM-DD-HH:MM:SS-XXX (frame)Concentration:NORM-VALUE-XXX (frame)ChangeRate:DELTA-C-XXXmg / (m^{3}\cdots) (frame)Volatility:SIGMA-VALUE-XXX (frame)Periodicity:PERIOD-COUNT-XXX".

[0115] Preferably, the data frame transmission period is 5 seconds; the data frame format is JSON; the transmission protocol is TCP / IP; each frame of data contains 5 fields: timestamp, normalized concentration value, rate of change, fluctuation characteristics, and periodic characteristics; if the network transmission is interrupted, the most recent 1000 frames of data are cached locally and retransmitted after the network is restored.

[0116] Beneficial effects of steps S2.1 to S2.6: This series of steps removes high-frequency noise through time-series filtering, detects and removes outliers based on the 3σ criterion, fills in missing data forward, extracts concentration time-series features after normalization, and packages them into a unified format dataset, thereby improving data quality and standardizing features, and providing high-quality input for trend prediction.

[0117] In this process, step S2.1 smooths the original concentration sequence using moving average filtering or Kalman filtering to effectively suppress random noise; step S2.2 marks outlier data points that exceed the statistical confidence interval based on the 3σ criterion to avoid polluting subsequent analysis with outlier data; step S2.3 performs forward imputation on missing data segments to maintain temporal continuity; step S2.4 maps concentration values ​​to a unified numerical range using Min-Max normalization to eliminate dimensional differences; step S2.5 extracts multi-dimensional time-series features such as rate of change, fluctuation characteristics, and periodic characteristics to enrich the input information of the prediction model; and step S2.6 packages the concentration time-series feature data into a unified format for use in step S3.

[0118] Step S3: Concentration trend prediction model is constructed based on the preprocessed concentration dataset, and dust concentration trend is predicted through the concentration trend prediction model to construct a dust concentration trend prediction index set.

[0119] Furthermore, step S3 specifically includes the following steps: Step S3.1: Construct a concentration time-series feature matrix based on the preprocessed concentration dataset.

[0120] Preferably, the core of this step is to construct the time-series feature matrix, arranging the preprocessed concentration dataset in matrix form according to the time series.

[0121] In this matrix, rows represent time steps and columns represent feature dimensions; the time window length is the number of sampling points required for the prediction model input, and the window length is determined based on the prediction duration and sampling interval; the matrix is ​​constructed using a sliding window method, and adjacent windows can overlap to improve data utilization.

[0122] The mathematical expression of the time series feature matrix is: .

[0123] Where T is the length of the time window (number of rows), d is the feature dimension (number of columns), T=120, d=3 (normalized concentration, rate of change, fluctuation features).

[0124] Preferably, the time window length T is 120 sampling points (corresponding to 10 minutes of historical data); the feature dimension d is 3-dimensional; the matrix is ​​constructed using a sliding window method with a step size of 30 sampling points; the overlap rate of adjacent windows is 75%; when there are fewer than 120 sampling points, zeros are used to fill the window.

[0125] Step S3.2: Train and construct the concentration trend prediction model based on the concentration time-series feature matrix.

[0126] Preferably, the core of this step is to construct and train the trend prediction model, using an LSTM recurrent neural network model.

[0127] Among them, the LSTM model selectively retains and forgets historical information through forget gate, input gate, and output gate mechanisms, effectively solving the gradient vanishing problem of traditional RNNs and is suitable for capturing long-term dependencies; the model training uses historically accumulated time-series feature matrix data, with the objective function being to minimize the mean square error between the predicted and actual values.

[0128] The core formula of the LSTM unit is:

[0129]

[0130]

[0131]

[0132]

[0133] .

[0134] in Output for the forget gate. For input gate output, Candidate cell state, In cellular state, For output gate output, It is in a hidden state; It is the Sigmoid activation function. It is the hyperbolic tangent activation function.

[0135] The loss function uses mean squared error: .

[0136] in This represents the actual concentration value. is the model's predicted value, and N is the number of training samples.

[0137] Preferably, the LSTM model has 2 layers with 64 units per layer; the input feature dimension is 3-dimensional; the training batch size is 32; the initial learning rate is 0.001; the learning rate decay strategy is to decrease by 50% every 10 epochs; the training epochs are 100 to 200; the early stopping strategy is to stop training if the validation set loss does not decrease for 10 consecutive epochs; the optimizer is Adam; and the weights are initialized using Xavier initialization.

[0138] Step S3.3: Predict the dust concentration trend within a preset prediction period using the trained concentration trend prediction model.

[0139] Preferably, the core of this step is to complete the trend prediction output, and to predict the dust concentration within the preset prediction period based on the trained LSTM model.

[0140] The prediction method uses rolling prediction: each time, the feature matrix composed of the most recent T historical sampling points is used as input, and the concentration prediction value for the next H time steps is output; after the prediction is completed, the prediction window is slid forward H time steps and the process is repeated; the prediction output is a normalized value, which needs to be denormalized to obtain the actual concentration prediction value.

[0141] The mathematical expression for rolling forecasts is: .

[0142] in Let n be the input feature matrix at time n. to This is the predicted value for the next H steps.

[0143] The inverse normalization formula is: .

[0144] Preferably, the prediction duration H is 180 sampling points (corresponding to 15 minutes); the prediction step size is 30 seconds; the rolling prediction window is 5 steps (i.e., the input is updated once after 5 consecutive output prediction values); the prediction model is deployed on an edge computing unit or a field application server.

[0145] Step S3.4: Calculate the dust concentration change rate, cumulative offset, and peak prediction value based on the prediction results.

[0146] Preferably, the core of this step is to calculate the prediction indicators and quantify the prediction results into three core early warning indicators.

[0147] Among them, the rate of change The cumulative offset is the ratio of the difference between the predicted end-of-period concentration and the predicted beginning-of-period concentration to the prediction duration, reflecting the overall trend of concentration increase or decrease; The cumulative area of ​​deviation between the concentration curve and the baseline concentration during the prediction period reflects the duration of the exceedance; peak predicted value. The maximum concentration at all predicted points within the prediction period is the core basis for early warning determination.

[0148] The formula for calculating the rate of change is: .

[0149] in This is the predicted concentration value at the end of the prediction period. To predict the actual concentration value at the beginning of the period, The sampling interval is denoted as .

[0150] The cumulative offset is calculated using the trapezoidal integral method: .

[0151] in The baseline concentration is the average of the first 10 sampling points in the prediction window.

[0152] The peak prediction value is: .

[0153] Preferably, baseline concentration The mean of the first 10 sampling points in the prediction window is used; the cumulative offset is calculated using the trapezoidal integral method; all three indicators are numerical, with the following dimensions: , , .

[0154] Step S3.5: Integrate the rate of change, the cumulative offset, and the peak predicted value into the dust concentration trend prediction index set.

[0155] Preferably, the core of this step is to encapsulate the indicator set, packaging the three predictive indicators into a structured dataset for use in step S4.

[0156] The indicator set format includes three fields: rate of change, cumulative offset, and peak prediction, which are encapsulated in a JSON data frame. The indicator set also includes a prediction timestamp and model version number for easy tracking and model management.

[0157] The JSON format of the indicator set is as follows: { "timestamp":"YYYY-MM-DD-HH:MM:SS-XXX", "model_version":"MODEL-VER-XXX", "change_rate":DELTA-C-VALUE, "cumulative_drift":OMEGA-VALUE, "peak_prediction":CMAX-VALUE, "unit_change_rate":"mg / (m^{3}\cdots)", "unit_cumulative_drift":"mg\cdots / m^{3}", "unit_peak":"mg / m^{3}" } Preferably, the indicator set transmission cycle is 5 seconds; it is output in JSON format; it includes prediction timestamp, model version number, and prediction window length information; the indicator set is pushed to the early warning judgment module and the central control monitoring system in step S4.

[0158] Beneficial effects of steps S3.1 to S3.5: This series of steps constructs a concentration time-series feature matrix, trains an LSTM or Transformer trend prediction model, predicts the dust concentration trend within a preset prediction period, calculates the rate of change, cumulative offset, and peak prediction value, and integrates them into an index set to achieve early warning and forward-looking decision support.

[0159] In step S3.1, the preprocessed concentration dataset is arranged into a feature matrix according to the time series and used as the input of the prediction model; step S3.2, the temporal dependencies are captured based on the LSTM or Transformer model to construct a trend prediction model with long-term memory capability; step S3.3, the concentration prediction sequence within a preset time period is output through the trained model; step S3.4, three core indicators, namely the rate of change, cumulative offset, and peak prediction value, are calculated to quantify the concentration change trend; step S3.5, the three indicators are integrated into a structured indicator set to provide standardized input for the comparison and judgment in step S4.

[0160] Step S4: Compare the dust concentration trend prediction index set with the preset three-level early warning threshold, and output the corresponding graded early warning instruction.

[0161] Furthermore, step S4 specifically includes the following steps: Step S4.1: Based on the lower limit of dust explosion, three levels of warning thresholds are set respectively, wherein the first-level warning threshold, the second-level warning threshold, and the third-level warning threshold correspond to different preset percentages of the lower limit of dust explosion.

[0162] Preferably, the core of this step is to set three levels of warning thresholds, defining the warning levels as alert, handling, and emergency based on a percentage of the lower limit of dust explosion.

[0163] The lower explosive limit of dust is determined based on the particle size distribution and combustibility of tobacco dust, and is generally between 20 and 60 kJ / m³. The three threshold levels are 40%, 60%, and 80% of the lower explosive limit, corresponding to the alert value, the handling value, and the emergency value, respectively.

[0164] The formula for calculating the Level 3 warning threshold is:

[0165]

[0166] .

[0167] in , , The warning thresholds for Level 1, Level 2, and Level 3 are respectively (unit: mg / m^{3}). , , The percentage coefficients are 40%, 60%, and 80%, respectively. Lower limit of dust explosion (unit) ).

[0168] Preferably, the first-level warning threshold It is 40% of the lower explosive limit; Level II warning threshold. The lower explosive limit is 60% of the lower explosive limit; the level 3 warning threshold L_3 is 80% of the lower explosive limit; the lower explosive limit benchmark value. Pick ;but =16000 , =24000 , =32000 .

[0169] Step S4.2: Compare the dust concentration trend prediction index set with the three-level early warning threshold item by item to determine the current early warning level.

[0170] Preferably, the core of this step is to determine the early warning level, specifically the peak predicted value among the forecast indicators. Compare with the Level 3 warning threshold.

[0171] The decision-making logic adopts a peak-priority principle: based on the peak prediction value. The primary criterion is used for judgment, corresponding to three threshold levels; if If multiple thresholds are exceeded simultaneously, the highest level will be output; when a level 2 or 3 warning is triggered, the rate of change will be referenced synchronously. With cumulative offset Assess the urgency of the risk.

[0172] The formula for determining the warning level is: .

[0173] Level=0 indicates normal, Level=1 indicates Level 1 warning, Level=2 indicates Level 2 warning, and Level=3 indicates Level 3 warning.

[0174] Preferably, a peak priority judgment principle is adopted; the three warning thresholds are as follows: =16000 , =24000 , =32000 The warning judgment period is 5 seconds; the judgment result is accompanied by... Specific values, , Three indicators are provided for reference in step S5.

[0175] Step S4.3: Generate the corresponding graded early warning instruction based on the current early warning level.

[0176] Preferably, the core of this step is to generate early warning instructions and encapsulate the judgment results into executable hierarchical early warning instructions.

[0177] The warning instruction includes four fields: warning level code, trigger timestamp, predicted indicator value, and suggested action. The warning level code uses numerical encoding, with 0 for normal, 1 for level 1, 2 for level 2, and 3 for level 3. The action is preset according to the warning level, with level 1 being an audible and visual alert, level 2 being a coordinated action, and level 3 being an emergency shutdown.

[0178] The JSON format of the warning instruction is as follows: { "alert_code":LEVEL-CODE, "trigger_time":"YYYY-MM-DD-HH:MM:SS-XXX", "peak_prediction":CMAX-VALUE, "change_rate":DELTA-C-VALUE, "cumulative_drift":OMEGA-VALUE, "action":"ACTION-DESC-XXX" } The corresponding action descriptions for ACTION-DESC are as follows: Level 1 is "Audible and visual alarm to alert on-site personnel", Level 2 is "Increase fan airflow and start dust removal system", and Level 3 is "Cut off equipment power and start explosion relief valve / explosion suppression system".

[0179] Preferably, the warning instruction format is JSON; the transmission method is TCP / IP push to the alarm server and PLC controller; the first-level warning is simultaneously pushed to the alarm window of the central control monitoring system; the second-level warning is pushed to the PLC controller for linkage execution; and the third-level warning is pushed to the emergency stop controller and the audible and visual alarm.

[0180] Beneficial effects of steps S4.1 to S4.3: This series of steps sets three levels of early warning thresholds, compares the predicted indicators with the thresholds one by one, determines the early warning level, and generates corresponding graded early warning instructions, thereby achieving refined management and differentiated response of early warning levels.

[0181] Specifically, step S4.1 sets three-level warning thresholds based on the percentage of the lower limit of dust explosion, dividing them into three levels: reminder value, handling value, and emergency value; step S4.2 compares the dust concentration trend prediction index set with the three-level warning thresholds item by item to accurately determine the current warning level; step S4.3 generates corresponding graded warning instructions according to the warning level, providing an execution basis for the linkage control in step S5.

[0182] Step S5: Based on the graded early warning command and the current concentration deviation, the frequency of the dust removal fan and the opening of the air valve are dynamically adjusted through the PID control algorithm, and the operating status of the dust generating equipment and the ventilation parameters of the workshop are adjusted in conjunction with it to obtain the result of the linkage control execution.

[0183] Furthermore, step S5 specifically includes the following steps: Step S5.1: Calculate the current concentration deviation value based on the graded early warning instruction and the current measured dust concentration.

[0184] Preferably, the core of this step is to calculate the concentration deviation and compare the measured concentration with the target concentration.

[0185] Wherein, the concentration deviation e(t) is the difference between the current measured concentration C(t) and the set target concentration. The difference is positive, indicating that the concentration is higher than the target, and negative, indicating that the concentration is lower than the target. The target concentration is set according to the production conditions and safety margin.

[0186] The formula for calculating concentration deviation is: .

[0187] Where e(t) is the concentration deviation at time t ( C(t) is the measured dust concentration at time t. ), For the target concentration ( ).

[0188] Preferably, the target concentration For 8000 (Corresponding to 20% of the lower explosive limit, with a safety margin reserved below the first-level warning threshold); the concentration deviation calculation cycle is consistent with the sampling interval, which is 5 seconds; the deviation value is retained to two decimal places, and the unit is... .

[0189] Step S5.2: Based on the current concentration deviation value, calculate the frequency control quantity of the dust removal fan and the opening control quantity of the air valve using the PID control algorithm.

[0190] Preferably, the core of this step is to implement the PID control algorithm, which calculates the control quantity through the three stages of proportional, integral, and derivative.

[0191] In this process, PID control provides immediate response through a proportional element, eliminates steady-state error through an integral element, and predicts future trends through a derivative element. The weighted sum of these three elements yields the control output. The control output is then sent to the frequency converter to control the speed of the dust collector fan and to the electric damper controller to adjust the opening degree.

[0192] The formula for continuous PID control is: .

[0193] The formula for discrete PID control (incremental type) is as follows: .

[0194] .

[0195] Where u(t) is the output of the control quantity (frequency Hz or opening percentage), is the proportional coefficient, is the integral coefficient, is the differential coefficient, and e(k) is the concentration deviation at the k-th moment.

[0196] Limiting processing of the control quantity: .

[0197] Preferably, the initial value is 2.0, the initial value is 0.5, the initial value is 0.1; the range of the fan frequency control quantity is 0 to 50 Hz; the range of the air valve opening control quantity is 0 to 100%; the PID control period is 5 seconds; the integral term adopts the trapezoidal integration method; the output of the control quantity is provided with limiting and rate limiting to prevent equipment impact.

[0198] Step S5.3, according to the hierarchical warning instruction and the current concentration deviation value, jointly adjust the operating state of the dust-producing equipment and the ventilation parameters of the workshop.

[0199] Preferably, the core of this step completes multi-system joint control and executes differential disposal actions according to the warning level.

[0200] Among them, the joint control logic is linked to the warning level: the first-level warning triggers audible and visual reminders, and at the same time, the dust-producing equipment maintains normal operation and the ventilation parameters remain unchanged; the second-level warning triggers the dust-producing equipment to operate at a reduced frequency and the workshop ventilation is increased; the third-level warning triggers the dust-producing equipment to stop urgently and the workshop ventilation operates at the maximum gear; the joint adjustment outputs digital or analog signals through the PLC to each subsystem.

[0201] Joint control logic decision table: .

[0202] Preferably, the frequency reduction ratio of the dust-producing equipment (cutter, dryer) is 80% in the second-level warning and 0% (stop) in the third-level warning; the ventilation parameters of the workshop are 60% of the maximum ventilation volume in the second-level warning and 100% in the third-level warning; the joint response delay does not exceed 10 seconds; after the joint action is executed, the central control monitoring system records the joint control log.

[0203] Step S5.4: Summarize the execution results of the dust removal fan, the air valve, the dust generating equipment and the workshop ventilation parameters, and merge them to obtain the linkage control execution result.

[0204] Preferably, the core of this step is to summarize the execution results, packaging the execution status and parameter values ​​of each subsystem into a linkage control execution result.

[0205] The summary includes seven fields: current frequency of dust removal fan, current opening degree of dust removal fan, operating status of dust-generating equipment, status of workshop ventilation parameters, instruction execution timestamp, and execution result code. The execution result code is 0 for success, 1 for partial success, and 2 for failure.

[0206] The JSON format of the linkage control execution result is as follows: { "result_code":RESULT-CODE, "execute_time":"YYYY-MM-DD-HH:MM:SS-XXX", "fan_frequency":FAN-FREQ-VALUEHz, "valve_opening":VALVE-OPEN-VALUE\%, "source_equipment_state":STATE-DESC-XXX, "ventilation_state":VENT-DESC-XXX, "alert_level":LEVEL-CODE } Preferably, the execution result summary period is 10 seconds; the result format is JSON; it is pushed to the central control monitoring system for display, and simultaneously stored in the local database for use in step S6; the execution result includes three subfields: instruction issuance time, target value, and actual value, for easy traceability afterward.

[0207] Beneficial effects of steps S5.1 to S5.4: This series of steps calculates the concentration deviation value, calculates the fan frequency and damper opening control quantity based on the PID control algorithm, and adjusts the dust-generating equipment and ventilation parameters in a coordinated manner. The results are then summarized to obtain the coordinated control execution result, thereby achieving automatic adjustment and timely handling of dust concentration.

[0208] Specifically, step S5.1 calculates the current concentration deviation value based on the graded early warning command and the measured concentration, which serves as the input signal for PID control; step S5.2 calculates the frequency control quantity of the dust collector fan and the opening control quantity of the air valve through the PID control algorithm to achieve precise adjustment of continuous quantities; step S5.3 adjusts the start-stop status of the dust-generating equipment and the workshop ventilation parameters in conjunction with the graded early warning command to achieve coordinated handling of multiple systems; and step S5.4 summarizes the execution results of the dust collector fan, air valve, dust-generating equipment, and ventilation parameters, and merges them to obtain the linkage control execution result.

[0209] Step S6: Calculate the control deviation statistical index by using the obtained dust concentration feedback data and the linkage control execution results. Update the control parameters of the PID control algorithm and the preset three-level early warning threshold according to the control deviation statistical index to obtain the monitoring parameter set after iteration.

[0210] Step S6.1: Obtain the dust concentration feedback data at the current moment through the dust concentration sensor.

[0211] Preferably, the core of this step is to complete the feedback data acquisition and obtain the actual dust concentration value after the linkage control is executed in real time.

[0212] The feedback data is the measured dust concentration value of the dust removal system after a certain response delay under the new operating parameters following the execution of the linkage control in step S5. The response delay time is related to the inertia of the dust removal system and is generally 30 seconds to 5 minutes. Feedback data should be collected only after the system stabilizes. The feedback data is used to calculate the control deviation and update the PID parameters.

[0213] Preferably, the feedback data acquisition delay is 60 seconds (waiting for the system response to stabilize); the feedback data source is the measured concentration value after preprocessing the pump-suction sampling data in step S1.1 in step S2; the feedback data acquisition cycle is 60 seconds; the average value of three consecutive acquisitions is taken as the final feedback value.

[0214] Step S6.2: Compare the dust concentration feedback data with the target concentration value in the linkage control execution result, and calculate the control deviation statistical index.

[0215] Preferably, the core of this step is to calculate the statistical indicators of control deviation and evaluate the effect of the linkage control.

[0216] Among them, the statistical indicators of control deviation include four indicators: MAE (mean absolute error), RMSE (root mean square error), and mean deviation. MAE reflects the overall level of control accuracy, RMSE is more sensitive to large deviations, and mean deviation reflects whether there is a systematic offset.

[0217] Preferably, the statistical window length N is 12 sampling points (corresponding to 60 seconds); the MAE threshold is set to 1000 mg / m³, and exceeding this threshold triggers PID parameter adjustment; the RMSE threshold is set to 1500 mg / m³; the statistical indicators are updated every 60 seconds.

[0218] Step S6.3: Update the proportional coefficient, integral coefficient, and derivative coefficient of the PID control algorithm according to the control deviation statistical index, and adjust the preset three-level early warning threshold according to the triggering of the graded early warning command.

[0219] Preferably, the core of this step is to complete the adaptive adjustment of PID parameters and the fine-tuning of the early warning threshold, and to dynamically optimize the control performance based on the statistical index of control deviation.

[0220] The PID parameter adjustment adopts an incremental adjustment strategy: when MAE exceeds the threshold, K_p is increased to improve the response speed, and Ki is appropriately increased to accelerate the integral action and eliminate steady-state error; when MAE is significantly lower than the threshold and RMSE is stable, K_p can be appropriately decreased to reduce system oscillation; when the three-level warning is frequently triggered, the three-level threshold is finely adjusted synchronously to prevent the threshold from being too dense and causing frequent emergency responses.

[0221] The formula for incremental adjustment of PID parameters is: .

[0222] .

[0223] .

[0224] Among the increments Determine the value based on the MAE size by referring to a table: .

[0225] The conditions for fine-tuning the warning threshold are as follows: If a Level 3 warning is triggered more than 3 times within 1 hour, and the MAE decreases by less than 50% after each trigger, then L_3 will be lowered by 5%, with a maximum cumulative reduction not exceeding [a certain percentage]. 20% of the benchmark value.

[0226] Preferably, the PID parameters are adjusted using the gradient descent method or the table lookup method; the parameter adjustment step size... It is 0.2. It is 0.05. The parameter setting is 0; the parameter adjustment period is 60 seconds; the upper and lower limits of the PID parameter settings are... , , The adjustment step size for the warning threshold is 5% of the baseline value; the threshold adjustment cycle is 1 hour; the maximum cumulative reduction of the Level 3 warning threshold shall not exceed 20% of the baseline value.

[0227] Step S6.4: The updated control parameters and the three-level early warning threshold are encapsulated into the iterative monitoring parameter set for use in the next monitoring cycle.

[0228] Preferably, the core of this step is to encapsulate the iterative parameter set, packaging the updated PID parameters and early warning thresholds for use in the next monitoring cycle.

[0229] The iterative parameter set includes PID control parameters ( , , ) and Level 3 warning threshold ( , , There are a total of 6 parameters; the parameter set includes a version number and update timestamp for easy tracking and comparison; the parameter set is stored in JSON format in a local configuration file and is simultaneously pushed to the PLC controller to update the running parameters.

[0230] The JSON format of the monitoring parameter set after iteration is as follows: { "param_version":"PARAM-VER-XXX", "update_time":"YYYY-MM-DD-HH:MM:SS-XXX", "Kp":KP-VALUE, "Ki":KI-VALUE, "Kd":KD-VALUE, "L1":L1-VALUE, "L2":L2-VALUE, "L3":L3-VALUE, "unit_pressure":"mg / m^{3}" } Preferably, the parameter set version number uses an auto-incrementing number; the parameter set is stored in the local configuration file config.json; it is synchronously pushed to the parameter storage area of ​​the PLC controller; after the parameter is updated, the PLC automatically restarts the PID control loop to load the new parameter; the most recent 10 historical parameter versions are retained for rollback in case of anomalies.

[0231] Beneficial effects of steps S6.1 to S6.4: This series of steps involves acquiring dust concentration feedback data, calculating control deviation statistics, updating PID control parameters and three-level early warning thresholds, and encapsulating them into an iterative monitoring parameter set to achieve adaptive optimization and long-term stable operation of the system.

[0232] Specifically, step S6.1 involves acquiring feedback data in real time through a dust concentration sensor to monitor the linkage control effect; step S6.2 compares the feedback data with the target concentration value to calculate control deviation statistical indicators such as MAE or RMSE; step S6.3 adjusts the three coefficients of the PID controller based on the control deviation statistical indicators and fine-tunes the three-level warning thresholds according to the triggering of the graded warning instructions; and step S6.4 encapsulates the updated control parameters and thresholds into an iterative monitoring parameter set for use in the next monitoring cycle to achieve closed-loop adaptive optimization.

[0233] Beneficial effects of steps S1 to S6: This method addresses the issues of insufficient monitoring coverage and delayed response in existing technologies. It simultaneously acquires raw dust concentration data indoors, at pipe nodes, and at dust sources through pump-assisted active sampling. A concentration trend prediction model is constructed to enable early prediction, avoiding the lag in post-event alarms. A three-level early warning threshold system is established and linked with predictive indicators to achieve tiered response and precise early warning. A PID control algorithm, in collaboration with multiple systems, dynamically adjusts the dust collector fan, linked dust-generating equipment, and ventilation parameters for timely handling. Based on feedback data, the PID parameters and early warning thresholds are iteratively updated to construct an adaptive optimization closed loop, ensuring long-term stable operation of the system.

[0234] like Figure 2 As shown, this embodiment provides an example of a dust concentration monitoring device for a tobacco dust removal room. In this embodiment, the dust concentration monitoring device is applied to the dust concentration monitoring method as described in the above embodiment.

[0235] Specifically, the dust concentration monitoring device includes a raw dust concentration data acquisition module 1, a data preprocessing module 2, a dust concentration trend prediction module 3, a graded early warning instruction acquisition module 4, a dust linkage processing module 5, and a concentration trend prediction model iteration module 6, which are connected electrically or through communication in sequence.

[0236] The system comprises the following modules: a raw dust concentration data acquisition module 1, which acquires raw dust concentration data from the tobacco dust removal chamber through active sampling by a dust concentration sensor; a data preprocessing module 2, which preprocesses the raw dust concentration data and extracts concentration time-series features to obtain a preprocessed concentration dataset; a dust concentration trend prediction module 3, which constructs a concentration trend prediction model based on the preprocessed concentration dataset and predicts the dust concentration trend to build a dust concentration trend prediction index set; a graded early warning instruction acquisition module 4, which compares the dust concentration trend prediction index set with a preset three-level early warning threshold and outputs the corresponding graded early warning instruction; a dust linkage processing module 5, which dynamically adjusts the frequency of the dust removal fan and the opening of the air valve based on the graded early warning instruction and the current concentration deviation using a PID control algorithm, and also adjusts the operating status of the dust-generating equipment and the workshop ventilation parameters in a linkage manner, merging the results of the linkage control execution; and a concentration trend prediction model iteration module 6, which calculates the control deviation statistical index based on the acquired dust concentration feedback data and the linkage control execution results, updates the control parameters of the PID control algorithm and the preset three-level early warning threshold according to the control deviation statistical index, and obtains the iterated monitoring parameter set.

[0237] like Figure 3 As shown, the electronic device 7 includes a processor 71 and a memory 72 coupled to the processor 71.

[0238] The memory 72 stores program instructions for implementing the dust concentration monitoring method of the tobacco dust removal room in any of the above embodiments.

[0239] The processor 71 is used to execute program instructions stored in the memory 72 for monitoring the dust concentration in the tobacco dust removal room.

[0240] The processor 71 can also be referred to as a CPU (Central Processing Unit). The processor 71 may be an integrated circuit chip with signal processing capabilities. The processor 71 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor can be a microprocessor or any conventional processor.

[0241] Furthermore, Figure 4 This is a schematic diagram of the structure of a storage medium according to an embodiment of this application. See also: Figure 4In this embodiment of the application, the storage medium 8 stores program instructions 81 capable of implementing all the above methods. These program instructions 81 can be stored in the storage medium in the form of a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to perform all or part of the steps of the methods in the various embodiments of this application. 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, or terminal devices such as computers, servers, mobile phones, and tablets.

[0242] In the several embodiments provided in this application, it should be understood that the disclosed apparatus, devices, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the dust concentration monitoring of a unit is only a logical function of dust concentration monitoring. In actual implementation, there may be other dust concentration monitoring methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or omitted. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interface, and the indirect coupling or communication connection of the device or unit may be electrical, mechanical, signal, or other forms.

[0243] Furthermore, the functional units in the various embodiments of this application 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. The integrated units described above can be implemented in hardware or as software functional units. The above are merely embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for monitoring dust concentration in a tobacco dust removal chamber, wherein dust concentration sensors are deployed inside the dust removal chamber, at pipe joints, and at dust sources; the tobacco dust removal chamber also has a dust removal fan connected to the outside, characterized in that... The dust concentration monitoring method includes: Step S1: Obtain the original dust concentration data of the tobacco dust removal room through the pump-type active sampling of the dust concentration sensor; Step S2: Perform data preprocessing on the original dust concentration data and extract concentration time-series features to obtain a preprocessed concentration dataset; Step S3: Construct a concentration trend prediction model based on the preprocessed concentration dataset, and predict the dust concentration trend through the concentration trend prediction model to construct a dust concentration trend prediction index set; Step S4: Compare the dust concentration trend prediction index set with the preset three-level early warning threshold, and output the corresponding graded early warning instruction; Step S5: Based on the graded early warning command and the current concentration deviation, the frequency of the dust removal fan and the opening of the air valve are dynamically adjusted through the PID control algorithm, and the operating status of the dust generating equipment and the workshop ventilation parameters are adjusted in conjunction, and the linkage control execution result is obtained by combining them. Step S6: Calculate the control deviation statistical index by using the acquired dust concentration feedback data and the linkage control execution result, and update the control parameters of the PID control algorithm and the preset three-level early warning threshold according to the control deviation statistical index to obtain the iterative monitoring parameter set.

2. The dust concentration monitoring method according to claim 1, characterized by, Step S1, obtaining the original dust concentration data of the tobacco dust removal chamber through the pump-type active sampling of the dust concentration sensor, including: Step S1.1: Trigger pump-type sampling at preset sampling intervals, and encapsulate the collected raw dust concentration data in a preset data format to obtain raw encapsulated data; Step S1.2: Remove spike noise and transient interference from the original packaging data to obtain the original dust concentration data.

3. The dust concentration monitoring method according to claim 1, characterized by, Step S2: Perform data preprocessing on the original dust concentration data and extract concentration time-series features to obtain a preprocessed concentration dataset, including: Step S2.1: Perform time-series filtering on the original dust concentration data to obtain the filtered concentration sequence; Step S2.2: Based on the 3σ criterion, outlier detection and removal are performed on the filtered concentration sequence, and abnormal data points with sudden concentration changes are marked. Step S2.3: Perform forward filling on the missing data segments after removing outliers; Step S2.4: Normalize the filled concentration data and map the concentration values ​​to a preset value range. Step S2.5: Extract concentration time-series features based on the normalized concentration sequence to obtain concentration time-series feature data; Step S2.6: Package the concentration time-series feature data into a preprocessed concentration dataset in a unified format.

4. The dust concentration monitoring method according to claim 1, characterized by, Step S3: Construct a concentration trend prediction model based on the preprocessed concentration dataset, and predict the dust concentration trend using the concentration trend prediction model to construct a dust concentration trend prediction index set, including: Step S3.1: Construct a concentration time-series feature matrix based on the preprocessed concentration dataset; Step S3.2: Train and construct the concentration trend prediction model based on the concentration time-series feature matrix; Step S3.3: Predict the dust concentration trend within a preset prediction period using the trained concentration trend prediction model. Step S3.4: Calculate the dust concentration change rate, cumulative offset, and peak prediction value based on the prediction results; Step S3.5: Integrate the rate of change, the cumulative offset, and the peak predicted value into the dust concentration trend prediction index set.

5. The dust concentration monitoring method according to claim 1, characterized by, Step S4: Compare the dust concentration trend prediction index set with the preset three-level early warning threshold, and output the corresponding graded early warning instruction, including: Step S4.1: Based on the lower limit of dust explosion, three levels of warning thresholds are set respectively, wherein the first-level warning threshold, the second-level warning threshold, and the third-level warning threshold correspond to different preset percentages of the lower limit of dust explosion. Step S4.2: Compare the dust concentration trend prediction index set with the three-level early warning threshold item by item to determine the current early warning level; Step S4.3: Generate the corresponding graded early warning instruction based on the current early warning level.

6. The dust concentration monitoring method according to claim 1, characterized by, Step S5: Based on the deviation between the graded early warning command and the current concentration, the frequency and valve opening of the dust removal fan are dynamically adjusted using a PID control algorithm, and the operating status of the dust-generating equipment and the workshop ventilation parameters are adjusted in conjunction with this adjustment. The combined results of the coordinated control execution are then obtained, including: Step S5.1: Calculate the current concentration deviation value based on the graded early warning instruction and the current measured dust concentration; Step S5.2: Based on the current concentration deviation value, calculate the frequency control quantity of the dust removal fan and the opening control quantity of the damper using the PID control algorithm; Step S5.3: Based on the graded early warning instruction and the current concentration deviation value, adjust the operating status of the dust-generating equipment and the workshop ventilation parameters in a coordinated manner; Step S5.4: Summarize the execution results of the dust removal fan, the air valve, the dust generating equipment and the workshop ventilation parameters, and merge them to obtain the linkage control execution result.

7. The dust concentration monitoring method according to claim 1, characterized by, Step S6: Calculate the control deviation statistical index using the acquired dust concentration feedback data and the execution result of the linkage control. Update the control parameters of the PID control algorithm and the preset three-level early warning threshold based on the control deviation statistical index to obtain the iterative monitoring parameter set, including: Step S6.1: Obtain the dust concentration feedback data at the current moment through the dust concentration sensor; Step S6.2: Compare the dust concentration feedback data with the target concentration value in the linkage control execution result, and calculate the control deviation statistical index; Step S6.3: Update the proportional coefficient, integral coefficient, and derivative coefficient of the PID control algorithm according to the control deviation statistical index, and adjust the preset three-level early warning threshold according to the triggering of the graded early warning command; Step S6.4: The updated control parameters and the three-level early warning threshold are encapsulated into the iterative monitoring parameter set for use in the next monitoring cycle.

8. A dust concentration monitoring device for a tobacco dust removal house, the dust concentration monitoring device being applied to the dust concentration monitoring method according to any one of claims 1 to 7, characterized by, The dust concentration monitoring device includes: The raw dust concentration data acquisition module is used to obtain the raw dust concentration data of the tobacco dust removal room through the pump-type active sampling of the dust concentration sensor. The data preprocessing module is used to preprocess the original dust concentration data and extract the concentration time series features to obtain a preprocessed concentration dataset. The dust concentration trend prediction module is used to construct a concentration trend prediction model based on the preprocessed concentration dataset, and to predict the dust concentration trend through the concentration trend prediction model to construct a dust concentration trend prediction index set. The graded early warning instruction acquisition module is used to compare the dust concentration trend prediction index set with the preset three-level early warning threshold and output the corresponding graded early warning instruction. The dust linkage processing module is used to dynamically adjust the frequency of the dust removal fan and the opening of the air valve based on the deviation between the graded early warning command and the current concentration, and to adjust the operating status of the dust generating equipment and the workshop ventilation parameters in a linkage manner, and combine them to obtain the linkage control execution result; The concentration trend prediction model iteration module is used to calculate the control deviation statistical index by acquiring dust concentration feedback data and the linkage control execution result, update the control parameters of the PID control algorithm and the preset three-level early warning threshold according to the control deviation statistical index, and obtain the monitoring parameter set after iteration.

9. An electronic device, comprising: The method includes a processor and a memory coupled to the processor, the memory storing program instructions that can be executed by the processor; when the processor executes the program instructions stored in the memory, it implements the dust concentration monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions that, when executed by a processor, enable the dust concentration monitoring method as described in any one of claims 1 to 7.

Images