A dust removal equipment filter bag dynamic dust removal method and system

Abstract

The application discloses a dust removal equipment filter bag dynamic ash cleaning method and system, and relates to the technical field of dust removal equipment. The method comprises the following steps: installing a micro hot type mass flow sensor at the outlet of each clean gas chamber of the dust removal equipment corresponding to each filter bag, installing a sound wave exciter on the filter bag ash cleaning blowing pipe, and installing a sound wave receiving sensor at the bottom of the filter bag; starting the sensor under the initial working condition of confirming the health of the filter bag, continuously collecting real-time flow data of each filter bag under different total working conditions, using statistical process control method to establish a personalized dynamic flow health baseline model for each filter bag in linkage with the total working condition, and simultaneously, performing sound wave scanning on each filter bag. Through the establishment of the personalized dynamic baseline model and the sound wave conduction baseline file, and combined with the use of unsupervised anomaly detection algorithm, the application can timely detect and discover early weak anomalies and local faults of the filter bag, and the early warning time is greatly advanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of dust removal equipment, and particularly relates to a dust removal equipment filter bag dynamic ash removal method and system. BACKGROUND

[0002] In modern industrial production, as a key terminal equipment for industrial flue gas purification, the bag type dust removal equipment has been widely applied in the steel, cement, power and flour industries. The working principle of the bag type dust removal equipment is that the dust-containing flue gas passes through the filter bag made of fiber fabric, and the fine particles are captured by the filter effect of the filter bag surface and the gradually formed dust layer, so as to realize efficient purification. In order to ensure the continuous and stable operation of the dust removal equipment, the dust accumulated on the filter bag must be removed regularly to maintain the operating resistance within a reasonable range. This process is called "ash removal". The current ash removal technology for filter bags is pulse blowing to realize ash removal, which has the advantages of high efficiency and high degree of automation.

[0003] At present, the ash removal control strategy of most dust removal systems depends on two traditional modes: timed ash removal and constant pressure difference ash removal. Timed ash removal is to trigger the blowing system to sequentially remove all or part of the filter bags according to the preset time interval without considering the actual working condition. Constant pressure difference ash removal is to monitor the total pressure difference between the inlet and outlet of the dust remover. When the pressure difference rises to the preset upper limit, the ash removal program is started, and the program is stopped until the pressure difference drops to the lower limit.

[0004] However, in application scenarios such as the sintering machine head of a large steel plant, which has extremely complex and volatile working conditions, the inherent disadvantages of the above-mentioned traditional ash removal mode are dramatically magnified, and it is difficult to meet the operating requirements of high efficiency, energy saving and long service life. For example, the flue gas working condition of the sintering machine head has significant complexity, volatility and harmfulness. The sintering flue gas composition is complex, containing not only conventional dust but also high-concentration oil mist, water vapor, sulfur oxide, nitrogen oxide and volatile organic compounds, etc. These components not only cause chemical corrosion of the filter bag, but also affect the physical properties of the dust such as adhesion and particle size distribution, increasing the difficulty of dust removal. The production conditions of the sintering machine (such as raw material ratio, trolley speed and ignition temperature) are frequently adjusted, resulting in real-time and dramatic fluctuations in flue gas flow, temperature, humidity and dust load. This environment is extremely prone to sticky dust and water on the surface of the filter bag, especially in the bottom, creases or areas with loose contact with the cage, forming a sticky and hardened layer. This phenomenon is often local at the initial stage, like a "cecum", which is hidden and develops, and has extremely weak influence on the total pipe pressure difference formed by thousands of filter bags, making it difficult to be detected by the system, so that it cannot be accurately identified and positioned in the early stage of local "cecum" blockage.

[0005] The prior art has the following disadvantages: In addition, the existing ash removal control strategy has fundamental defects in the face of the above problems. For example, the fixed-time ash removal lacks pertinence and cannot perceive the actual working state and dust accumulation of each filter bag. When the dust load is low, unnecessary ash removal not only wastes a large amount of compressed air and increases energy consumption, but also frequently damages the "dust layer" on the surface of the filter bag which plays a key filtering role, resulting in temporary reduction of filtering efficiency, instantaneous emission exceeding the standard, and accelerated damage of the filter bag due to mechanical fatigue effect. When the dust load suddenly increases, the fixed ash removal period may be insufficient to respond in time, resulting in rising resistance; the fixed pressure difference ash removal has a reaction lag. When the total pipe pressure difference is detected to be significantly increased, it usually means that the hardened layer of the local filter bag is quite serious, which may cause aging and deformation of the filter bag material, and even irreversible damage such as "bag sticking". At this time, ash removal not only consumes a lot of energy with little effect, but also cannot identify which filter bag is abnormal, so the entire total pipe is cleaned, which not only wastes energy but also accelerates the wear of healthy filter bags. In the worst case, the belt fails, resulting in large-scale replacement, causing huge economic losses and environmental protection risks of emission exceeding the standard.

[0006] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0007] The purpose of the present application is to provide a dust removal equipment filter bag dynamic ash removal method and system. The present application analyzes individualized flow baseline through unsupervised learning algorithm, performs cross-validation by fusing flow and acoustic multi-sensor data, and accurately removes ash for specific abnormal filter bags. The ash removal parameters are adjusted according to the degree and position of blockage to solve the long-standing "cecum type" blockage problem in the background art.

[0008] In order to achieve the above purpose, the present application provides the following technical scheme: a dust removal equipment filter bag dynamic ash removal method, comprising the following steps:

[0009] S1, a micro thermal mass flow sensor is installed at the outlet of each clean gas chamber of the dust removal equipment corresponding to each filter bag, a sound wave exciter is installed on the filter bag ash removal blowing pipe, and a sound wave receiving sensor is installed at the bottom of the filter bag. In the initial working condition of confirming the health of the filter bag, start the sensor, continuously collect real-time flow data of each filter bag under different total working conditions, use statistical process control method to establish individualized dynamic flow health baseline model for each filter bag in conjunction with total working condition, and at the same time, perform sound wave scanning on each filter bag, record sound wave signal characteristics, and establish sound wave conduction baseline file;

[0010] S2, when the ash removal system is running normally, based on the continuous acquisition of real-time flow data of each filter bag, the deviation of the flow value of each filter bag from the individualized dynamic baseline is calculated, a signal threshold value is pre-set, the deviation data of all filter bags in the continuous time window is input into the isolation forest unsupervised anomaly detection algorithm, the abnormal score is automatically calculated, and then compared with the pre-set signal threshold value to identify and mark the preliminary abnormal filter bags and generate an abnormal list containing the filter bag ID;

[0011] S3, based on the preliminary abnormal filter bag, the sound signal is emitted by the sound wave exciter of the control target filter bag to the sound wave receiving sensor, and the acoustic characteristic parameter change is compared and analyzed with the sound wave conduction baseline file to verify the abnormality, and the preset acoustic characteristic threshold value is compared and analyzed to generate a diagnosis result;

[0012] S4, based on the diagnosis result, a parameterized ash removal instruction is adaptively generated, and when the pulse jet control system receives the ash removal instruction, the ash removal action is performed on the specified target filter bag, and through long-term data accumulation, the machine learning algorithm is used to optimize the ash removal strategy, and the optimal ash removal parameters are adaptively learned.

[0013] Optionally, the real-time flow data acquisition step is as follows:

[0014] At the outlet of each clean gas chamber of the dust removal equipment, a miniature thermal mass flow sensor is installed corresponding to each filter bag, a sound wave exciter is installed on the filter bag ash removal jet pipe, and a sound wave receiving sensor is installed at the bottom of the filter bag, and a total working condition sensor for air volume and temperature is installed at the inlet of the dust removal equipment, all sensors are initialized and set, and are in normal working condition;

[0015] Before starting the sensor to collect data, it is necessary to confirm that the filter bag is in a healthy initial working condition;

[0016] After confirming that the filter bag is in a healthy initial working condition, a synchronous acquisition instruction for starting the miniature thermal mass flow sensor and the total working condition sensor is sent, and the real-time flow data of each filter bag is continuously collected, and the original data of the time stamp, the real-time volume flow value of each filter bag, the total treatment air volume at the corresponding time, and the flue gas inlet temperature at the corresponding time are recorded;

[0017] The real-time flow readings of each sensor are subjected to moving average filtering processing;

[0018] In a self-learning period of 24 hours, as much as possible, repeat the original data collection and processing under various working condition changes, and construct a time series data set of all data collected in the period.

[0019] Optionally, the individualized dynamic flow health baseline model is established as follows:

[0020] Real-time flow data of each filter bag under different total air volume and temperature conditions is extracted from the time series dataset. The data is grouped according to the total air volume and temperature conditions, with each group representing a specific combination of conditions, thus realizing the binning of operating condition parameters.

[0021] All data points in the time series dataset are categorized into the operating chambers with different total air volume and temperature ranges according to their corresponding total air volume and temperature values.

[0022] Using the Statistical Process Control (SPC) method, for each operating chamber, the baseline value and control limits are defined by calculating the sample mean and sample standard deviation of all filter bag flow data falling into each chamber.

[0023] The control limits of each operating condition group are associated with the corresponding total air volume and temperature conditions. After traversing all operating condition boxes, a personalized dynamic flow health baseline model for each filter bag is constructed.

[0024] Based on the personalized dynamic flow health baseline model, input the current total air volume and temperature conditions, and output the expected baseline value and normal fluctuation range of the flow rate of each filter bag under the current conditions by querying the corresponding control limits.

[0025] For the condition where the load falls between the bins, the corresponding upper control limit (UCL), lower control limit (LCL), and control limits of the center mean are calculated using linear interpolation.

[0026] Optionally, the steps for establishing the acoustic wave transmission baseline profile are as follows:

[0027] Under the initial working condition of confirming the health of the filter bags, the acoustic exciter is controlled to perform an acoustic scan on each filter bag. The acoustic exciter emits an acoustic signal of a specific frequency and intensity, which is received by the acoustic receiving sensor after propagating through the filter bag.

[0028] For the Strip filter bag repeat The infrasound signal excitation-acquisition process, and... The received acoustic signals are superimposed and averaged to obtain a high-quality averaged time-domain signal. The formula for calculating the averaged time-domain signal is as follows: In the formula, Represented as the first Strip filter bags in repeated Infrasound signal excitation-acquisition process The average time-domain signal at time t, This represents the total number of repetitive acoustic signal excitation-acquisition processes. Represented as the first The filter bag is in the first Infrasound signal excitation-acquisition process The sound wave signals received at all times;

[0029] The time difference between the transmission and reception times of the acoustic signal, the peak value of the measured signal, and the spectrum obtained by performing a Fast Fourier Transform on the average time-domain signal are calculated from a high-quality average time-domain signal to extract and record the acoustic signal characteristics of propagation time, amplitude, and spectrum. The expression for propagation time is: In the formula, Represented as the first The propagation time of the scanning acoustic signal of the filter bag. This represents the moment when the sound wave receiving sensor receives the signal. This indicates the moment when the acoustic wave exciter emits a signal;

[0030] The expression for amplitude is In the formula, Represented as the first The amplitude of the average time-domain signal of the filter bag. This is expressed as the maximum amplitude taken from the average time-domain signal. This is represented by the specific frequency and intensity of a sound wave signal;

[0031] The expression for the spectrum is In the formula, Represented as the first The spectrum of the acoustic signal of the filter bag. This is represented as a Fast Fourier Transform operation on the average time-domain signal;

[0032] The acoustic signal features extracted from each filter bag are used as the acoustic wave conduction baseline profile of the corresponding filter bag in a healthy state. The acoustic wave conduction baseline profile includes the ID information of the filter bag and the health benchmark value range of the acoustic signal features.

[0033] Optionally, the calculation steps for the deviation are as follows:

[0034] Acquire real-time flow data of each filter bag continuously collected, and at the same time acquire the personalized dynamic baseline established for each filter bag;

[0035] For each filter bag, the real-time flow data is calculated to obtain the deviation value by comparing it with the personalized dynamic baseline centerline.

[0036] Standardize the deviation values ​​and calculate the degree of deviation.

[0037] Optionally, the steps for calculating the anomaly score using the IsolationForest unsupervised anomaly detection algorithm are as follows:

[0038] The deviation data of all filter bags within a continuous sliding time window are combined into a deviation dataset;

[0039] The q isolated trees are constructed by using random sampling method on the deviation degree feature data;

[0040] The deviation degree data set is input into the pre-configured Isolation Forest unsupervised anomaly detection algorithm for training;

[0041] For each sample data point in the deviation degree data set, the path length from the root node to the leaf node in each isolated tree is calculated;

[0042] The average path length of each sample data point in all isolated trees is calculated;

[0043] The average path length is calculated according to the average path length of the corresponding sample data point.

[0044] Optionally, the acoustic feature parameter change verification anomaly step is as follows:

[0045] After receiving the marked preliminary anomaly filter bag, acoustic tests are started on each target filter bag in the anomaly list in turn;

[0046] The acoustic wave exciter of the target filter bag in the anomaly list emits acoustic signals of a specific frequency, and the acoustic wave receiving sensor receives the acoustic signals;

[0047] The acoustic signal excitation-capture process is repeated multiple times for the target filter bag, and the received acoustic signals are time-domain superposition averaged to obtain high signal-to-noise ratio average time-domain signals;

[0048] The processed average time-domain signals are subjected to feature extraction of signal attenuation, propagation time difference, and frequency spectrum distortion to obtain acoustic feature parameters;

[0049] The acoustic feature parameters extracted from the current target filter bag are subtracted from the healthy reference value range stored in the acoustic wave conduction baseline archive to calculate the acoustic feature parameter change;

[0050] The acoustic feature parameter change is compared with the preset acoustic feature threshold to determine whether the target filter bag really has an anomaly.

[0051] Optionally, the generation of the diagnosis result is as follows:

[0052] The acoustic feature parameter changes of each filter bag are compared with the preset acoustic feature threshold one by one, and when the change of any acoustic feature parameter exceeds the corresponding acoustic feature threshold, it is determined that the filter bag has a physical blockage anomaly;

[0053] When the calculated acoustic characteristic parameters of a filter bag are all within the normal range of the preset acoustic characteristic threshold, the preliminary abnormal filter bag is determined as a false alarm, and the filter bag is removed from the abnormal list;

[0054] For the filter bag with physical abnormality verified by acoustic test, the clogging area is roughly located by analyzing the change mode of acoustic characteristic parameters, including A mode rough positioning of filter bag bottom serious clogging, B mode rough positioning of filter bag middle and upper part uniform paste bag, and C mode rough positioning of filter bag local hardening;

[0055] Based on the above all judgments, a structured diagnostic record is generated for each detected target filter bag, which integrates the diagnostic results of filter bag ID information, whether abnormal, whether physical clogging and rough positioning information of abnormal area.

[0056] Optionally, the adaptive generation of the parameterized ash removal instruction is as follows:

[0057] According to the diagnostic result, the abnormality of the target filter bag is determined, and based on the historical ash removal data and experimental results, a mapping relationship between the diagnostic result and the ash removal parameters of the pressure increment, the pulse width extension and the continuous blowing times is established, and an ash removal strategy mapping table is obtained;

[0058] The diagnostic result is divided into different abnormal types and severity, and the ash removal strategy mapping table defines the ash removal parameter adjustment reference value corresponding to different abnormal types and severity;

[0059] According to the total air volume and temperature working condition of the current ash removal system, combined with the historical ash removal record of the target filter bag, the initial ash removal parameters are linearly and adaptively adjusted;

[0060] The adjusted ash removal parameters are combined with the target filter bag to generate a parameterized ash removal instruction.

[0061] A filter bag dynamic ash removal system of a dust removal equipment, comprising a distributed sensing and detection module: a micro thermal mass flow sensor installed at the outlet of each clean gas chamber of the dust removal equipment corresponding to each filter bag, an acoustic wave exciter installed on the filter bag ash removal blowing pipe, and an acoustic wave receiving sensor installed at the bottom of the filter bag, which monitors the filter bag in the dust removal equipment in real time;

[0062] A data acquisition and baseline modeling processing module: the data of the micro thermal mass flow sensor and the acoustic wave receiving sensor are collected, and a personalized dynamic flow health baseline model is established for each filter bag by using statistical process control (SPC) method, the acoustic signal characteristics of the filter bag in healthy state are recorded, and an acoustic wave conduction baseline file is established;

[0063] Diagnosis analysis and ash removal decision module: based on the individualized dynamic flow health baseline model of filter bag, the Isolation Forest unsupervised anomaly detection algorithm is run to identify the preliminary abnormal filter bag, the abnormality is verified according to the comparative analysis result of the acoustic signal, whether the filter bag exists physical blockage is diagnosed and the blockage area is roughly located, and the parameterized ash removal instruction is generated according to the diagnosis result;

[0064] Ash removal parameter control and feedback learning module: control the pulse jet control system to execute the ash removal instruction, collect the effect data after ash removal and store it in association with the ash removal parameter, and use the machine learning algorithm to optimize the ash removal strategy, and feed back the new flow data of the filter bag after ash removal to the individualized dynamic flow health baseline model for fine tuning.

[0065] In the above technical solution, the technical effects and advantages provided by the present application are:

[0066] By establishing an individualized dynamic baseline model and a sound conduction baseline file, and combining the use of an unsupervised anomaly detection algorithm, the present application can detect and discover early weak abnormalities and local faults in filter bags in a timely manner, greatly advancing the early warning time, and in the very early stage of filter bag blockage, the subtle changes in the flow pattern can be used to identify abnormalities, providing the possibility of taking timely measures to prevent further damage to the filter bag; cross verification is performed in a manner of data fusion of multiple sensors of flow monitoring and acoustic detection, effectively avoiding false positives and improving the accuracy and reliability of filter bag blockage fault diagnosis; parameterized ash removal instructions are generated adaptively through diagnosis results, ash removal actions are performed on the specified target filter bag, invalid ash removal is avoided, precise targeted ash removal of abnormal filter bags is achieved, ash removal parameters are adjusted according to the blockage degree and the roughly located position, the ash removal effect is improved while reducing compressed air energy consumption and filter bag damage, and the use of machine learning algorithm to optimize the ash removal strategy can continuously optimize the ash removal strategy, adapt to complex working conditions such as filter bag aging and dust property changes, and further ensure long-term stable and efficient operation of the ash removal system. BRIEF DESCRIPTION OF DRAWINGS

[0067] In order to more clearly illustrate the technical solutions in the embodiments or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0068] Figure 1 Flow chart of the filter bag dynamic ash removal method of the dust removal equipment of the present application.

[0069] Figure 2 Module block diagram of the filter bag dynamic ash removal system of the dust removal equipment of the present application. DETAILED DESCRIPTION

[0070] Example implementations are now described with reference to the drawings. Example implementations can, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the inventive gist to those skilled in the art.

[0071] Embodiment 1

[0072] The present application provides a dynamic dust cleaning method for filter bags of a dust removal equipment as shown in Figure 1 The present application provides a dynamic dust cleaning method for filter bags of a dust removal equipment as shown in

[0073] S1, at the outlet of each clean gas chamber of the dust removal equipment, a micro hot-type mass flow sensor is installed corresponding to each filter bag, an acoustic wave exciter is installed on the filter bag cleaning blowing pipe, and an acoustic wave receiving sensor is installed at the bottom of the filter bag, the sensors are started in the initial working condition of confirming the health of the filter bag, real-time flow data of each filter bag under different total air volume and temperature conditions are continuously collected, a personalized dynamic flow health baseline model linked with the total working condition is established for each filter bag by using statistical process control (SPC) method, at the same time, acoustic wave scanning is performed on each filter bag, acoustic wave signal characteristics such as propagation time, amplitude and frequency spectrum are recorded, and an acoustic wave conduction baseline file is established;

[0074] Specifically, the collection of real-time flow data is as follows:

[0075] At the outlet of each clean gas chamber of the dust removal equipment, a micro hot-type mass flow sensor is installed corresponding to each filter bag, which can accurately measure the clean gas flow through the filter bag, an acoustic wave exciter is installed on the filter bag cleaning blowing pipe and an acoustic wave receiving sensor is installed at the bottom of the filter bag, which are used for acoustic detection and collection of acoustic wave signal characteristics, and a total working condition sensor for air volume and temperature is installed at the air inlet of the dust removal equipment, all sensors are initialized and set to be in normal working condition;

[0076] Before starting the sensors to collect data, it is necessary to confirm that the filter bag is in a healthy initial working condition, which can be ensured by checking the appearance and operating parameters of the filter bag to ensure that the filter bag has no obvious damage or abnormality;

[0077] In the initial working condition of confirming the health of the filter bag, a synchronous collection instruction for starting the micro hot-type mass flow sensor and the total working condition sensor is sent, the real-time flow data of each filter bag is continuously collected, and the original data of the time stamp, the real-time volume flow value of each filter bag, the total treatment air volume at the corresponding time, and the flue gas inlet temperature at the corresponding time are recorded, wherein the expression of the original data is , wherein, represents the original data, represents the time stamp, Represented as the first Filter bags in Real-time volumetric flow rate values ​​collected at specific times. Indicated as in Total air volume handled at any given time Indicated as in The flue gas inlet temperature at any given time;

[0078] The real-time flow readings of each sensor are processed using a moving average filter to eliminate random fluctuations and noise interference. The expression for the moving average filter is as follows: In the formula, Represented as the first Filter bags in Real-time volumetric flow rate values ​​collected at specific times. Represented as the first Filter bags in Real-time volumetric flow rate values ​​collected at specific times. Represented as the first Filter bags in Real-time volumetric flow rate values ​​collected at specific times. Represented as the first Filter bags in Real-time volumetric flow rate values ​​collected at specific times. Represented as the first Filter bags in Real-time volumetric flow rate values ​​collected at specific times. This is expressed as the number of sampling points;

[0079] Within a 24-hour self-learning cycle, the original data collection and processing are repeated to cover as many different working conditions as possible. All data collected within the cycle are then used to construct a time-series dataset for subsequent analysis and processing. The expression for the time-series dataset is as follows: In the formula, Represented as the first Time-series dataset of filter bags during the self-learning period. This is represented as a 24-hour self-learning cycle.

[0080] Specifically, the steps for establishing a personalized dynamic traffic health baseline model are as follows:

[0081] Real-time flow data for each filter bag under different total air volume and temperature conditions was extracted from the time-series dataset. The data was then grouped according to the total air volume and temperature conditions, with each group representing a specific combination of conditions. This process of binning the operating parameters divided the continuously changing total air volume and temperature into M discrete intervals. The expression for binning the operating parameters is as follows: ,and ,as well as , , , , wherein, is expressed as the increment of the total air volume, is expressed as the increment of the temperature, is expressed as the minimum and maximum of the total air volume range, is expressed as the minimum and maximum of the temperature range, is expressed as the first operating condition box, is expressed as the second operating condition box, is expressed as the first operating condition box;

[0082] All data points of the time series data set are classified into the operating condition boxes of the above-mentioned M different total air volume and temperature ranges according to the corresponding total air volume and temperature values;

[0083] By using the statistical process control (SPC) method, the sample mean and sample standard deviation of all filter bag flow data falling into each box are calculated to define the reference value and control limit of each operating condition box, wherein the calculation formula of the sample mean is , and , wherein, is expressed as the flow average value of the first filter bag under the first operating condition box, is expressed as the number of data points falling into the first operating condition box, is expressed as the first flow data point falling into the first operating condition box;

[0084] The calculation formula of the sample standard deviation is , wherein, is expressed as the flow standard deviation of the first filter bag under the first operating condition box;

[0085] The expression of the control limit is , wherein, is expressed as the upper control limit of the first filter bag under the first operating condition box, is expressed as the lower control limit of the first filter bag under the first operating condition box;

[0086] The control limit of each operating condition group is associated with the corresponding total air volume and temperature condition, and after traversing all operating condition boxes, the individualized dynamic flow health baseline model of each filter bag is constructed;

[0087] Based on the personalized dynamic flow health baseline model, input the current total air volume and temperature conditions, and output the expected baseline value and normal fluctuation range of the flow rate of each filter bag under the current conditions by querying the corresponding control limits.

[0088] For the condition where the load falls between the bins, the corresponding upper control limit (UCL), lower control limit (LCL), and control limits of the center mean are calculated using linear interpolation.

[0089] Specifically, the steps for establishing a sound wave transmission baseline profile are as follows:

[0090] Under the initial working condition of confirming the health of the filter bags, the acoustic exciter is controlled to perform an acoustic scan on each filter bag. The acoustic exciter emits an acoustic signal of a specific frequency and intensity, which is received by the acoustic receiving sensor after propagating through the filter bag.

[0091] For the first Strip filter bag repeat The infrasound signal excitation-acquisition process, and... The received acoustic signals are superimposed and averaged to obtain a high-quality averaged time-domain signal. The formula for calculating the averaged time-domain signal is as follows: In the formula, Represented as the first Strip filter bags in repeated Infrasound signal excitation-acquisition process The average time-domain signal at time t, This represents the total number of repetitive acoustic signal excitation-acquisition processes. Represented as the first The filter bag in the first Infrasound signal excitation-acquisition process The sound wave signals received at all times;

[0092] The time difference between the transmission and reception times of the acoustic signal, the peak value of the measured signal, and the spectrum obtained by performing a Fast Fourier Transform on the average time-domain signal are calculated from a high-quality average time-domain signal to extract and record the acoustic signal characteristics of propagation time, amplitude, and spectrum. The expression for propagation time is: In the formula, Represented as the first The propagation time of the scanning acoustic signal of the filter bag. This represents the moment when the sound wave receiving sensor receives the signal. This indicates the moment when the acoustic wave exciter emits a signal;

[0093] The expression for amplitude is: In the formula, Represented as the first an amplitude of the average time domain signal of the filter bag, is represented as a maximum amplitude value taken from the average time domain signal, is represented as a specific frequency and intensity of the sound wave signal;

[0094] The expression of the spectrum is , wherein, is represented as the spectrum of the sound wave signal of the filter bag, is represented as a fast Fourier transform operation on the average time domain signal;

[0095] The sound wave signal feature extracted on each filter bag is taken as a sound wave conduction baseline profile of the corresponding filter bag in a healthy state, and the sound wave conduction baseline profile includes ID information of the filter bag, a healthy reference value range of the sound wave signal feature, which is used in a subsequent monitoring process to compare the sound wave signal feature collected in real time with the sound wave conduction baseline profile to judge the health state of the filter bag, wherein the expression of the sound wave conduction baseline profile is , wherein, is represented as the sound wave conduction baseline profile of the filter bag, is represented as the ID information of the filter bag.

[0096] S2, when the ash removal system is normally running, based on continuously collected real-time flow data of each filter bag, the deviation of the flow value of each filter bag from the personalized dynamic baseline is calculated, a signal threshold value is pre-set, the deviation data of all filter bags in a continuous time window is input into an Isolation Forest unsupervised anomaly detection algorithm, the anomaly score of each data point is automatically calculated, and then compared with the pre-set signal threshold value for analysis, to identify and mark preliminary abnormal filter bags and generate an abnormal list containing the ID of the filter bag;

[0097] Specifically, the calculation steps of the deviation are as follows:

[0098] The continuously collected real-time flow data of each filter bag is obtained, and the personalized dynamic baseline established for each filter bag is obtained;

[0099] For the real-time flow data of each filter bag, the difference from the center line of the personalized dynamic baseline is calculated to obtain a deviation value, wherein the calculation formula of the deviation value is , wherein, is represented as the deviation value of the filter bag, is represented as the real-time flow data of the filter bag, is represented as the reference value of the personalized dynamic baseline of the filter bag;

[0100] The deviation value is standardized to calculate the deviation degree, wherein the calculation formula of the deviation degree is , wherein represents the deviation degree of the first filter bag, represents the upper control limit of the personalized dynamic baseline of the first filter bag, represents the lower control limit of the personalized dynamic baseline of the first filter bag, represents the control range width of the personalized dynamic baseline of the first filter bag.

[0101] Specifically, the steps of calculating the anomaly score by the Isolation Forest unsupervised anomaly detection algorithm are as follows:

[0102] The deviation degree data of all filter bags in a continuous sliding time window is combined into a deviation degree data set;

[0103] q Isolation Trees are constructed by randomly sampling the deviation degree feature data;

[0104] The deviation degree data set is input into the pre-configured Isolation Forest unsupervised anomaly detection algorithm for training;

[0105] For each sample data point in the deviation degree data set, the path length from the root node to the leaf node in each Isolation Tree is calculated;

[0106] The average path length of each sample data point in all Isolation Trees is calculated, wherein the calculation formula of the average path length is , and , wherein represents the average path length of the first sample data point in the Isolation Forest, represents the total number of sample data points in the deviation degree data set, represents the path length of the first sample data point in the first Isolation Tree, represents the number of Isolation Trees in the Isolation Forest;

[0107] The anomaly score of the corresponding sample data point is calculated according to the average path length, wherein the calculation formula of the anomaly score is , and , wherein represents the anomaly score of the first sample data point, represents a given sub-sample size a standardization factor of path length at the time, expressed as a harmonic number over the sample size expressed as a harmonic number over the sample size expressed as a harmonic number over the sample size expressed as a harmonic number over the sample size

[0108] Specifically, the generation step of the anomaly list is as follows:

[0109] For each anomaly score of the deviation degree data point, map back to the corresponding filter bag ID information and time point;

[0110] According to the actual situation and experience, a preset threshold is set for judging whether the sample is abnormal or not;

[0111] Compare the anomaly score of each sample with the preset threshold, and when the anomaly score exceeds the preset threshold, identify and mark the preliminary abnormal filter bag;

[0112] All ID marked as preliminary abnormal filter bag is summarized to generate an anomaly list containing filter bag ID, and the anomaly list contains filter bag ID information, anomaly score, time stamp and deviation degree.

[0113] S3, based on the preliminary abnormal filter bag, the sound wave exciter of the control target filter bag emits a specific frequency of acoustic signal to the sound wave receiving sensor to receive the acoustic signal, compares and analyzes the signal attenuation, propagation time difference, frequency spectrum distortion and acoustic characteristic parameter change amount of the sound wave conduction baseline file to verify the anomaly, compares and analyzes with the preset acoustic characteristic threshold to diagnose whether the filter bag abnormal state is physical blockage, further coarsely locates the abnormal area according to the acoustic characteristic parameter change mode, and generates a diagnosis result;

[0114] Specifically, the steps of verifying the anomaly of the acoustic characteristic parameter change amount are as follows:

[0115] After receiving the preliminary abnormal filter bag, acoustic test is started for each target filter bag in the anomaly list in turn;

[0116] The sound wave exciter of the target filter bag in the anomaly list emits a specific frequency of acoustic signal, and records the acoustic signal received by the sound wave receiving sensor at the same time;

[0117] The sound wave signal excitation-collection process is repeated for the target filter bag for many times, and the time domain superposition average is performed on the acoustic signals received for many times to obtain an average time domain signal with high signal-to-noise ratio;

[0118] The processed average time domain signal is subjected to feature extraction of signal attenuation, propagation time difference and frequency spectrum distortion to obtain acoustic characteristic parameters;

[0119] The acoustic feature parameters extracted from the current target filter bag are subtracted from the range of health baseline values ​​stored in the acoustic wave transmission baseline file to calculate the change in acoustic feature parameters. The formula for calculating signal attenuation is as follows: In the formula, This is expressed as the change in the amount of acoustic signal attenuation. This indicates the target filter bag currently collected. The amount of acoustic signal attenuation, This refers to the filter bag recorded in the acoustic wave transmission baseline file. The baseline value of the amplitude, This indicates the target filter bag that has been marked as having an initial abnormality;

[0120] The formula for calculating the propagation time difference is: In the formula, Indicated as target filter bag The time difference in transmission This indicates the target filter bag currently collected. The propagation time of the sound signal This refers to the filter bag recorded in the acoustic wave transmission baseline file. The baseline value of the propagation time;

[0121] The formula for calculating spectral distortion is: In the formula, Represented as the spectrum of an acoustic signal at a specific frequency The amount of distortion change at that location. This indicates the target filter bag currently collected. The acoustic signal spectrum, This refers to the filter bag recorded in the acoustic wave transmission baseline file. The baseline value of the spectrum;

[0122] The changes in acoustic feature parameters are compared with preset acoustic feature thresholds to determine whether the target filter bag is truly abnormal.

[0123] Specifically, the steps for generating the diagnostic results are as follows:

[0124] The changes in acoustic characteristic parameters calculated for each filter bag are compared one by one with the preset acoustic characteristic thresholds. If the change in any acoustic characteristic parameter exceeds the corresponding acoustic characteristic threshold, it is determined that the filter bag has a physical blockage abnormality.

[0125] If the acoustic characteristic parameters calculated for a filter bag are all within the normal range of the preset acoustic characteristic threshold, the filter bag is determined to be a false alarm and the filter bag is removed from the abnormal list.

[0126] For the acoustic test to verify the physical abnormality of the filter bag, the change mode of the acoustic characteristic parameters is analyzed to roughly locate the blocked area, including A mode for roughly locating the serious blockage at the bottom of the filter bag, B mode for roughly locating the uniform paste bag at the middle and upper part of the filter bag, and C mode for roughly locating the local hardening of the filter bag, wherein the change of the A mode is that the signal attenuation amount significantly increases, and the propagation time difference is significantly positive, which reflects that the signal is strongly attenuated and the propagation is obviously slowed down, and it is determined that the blockage is concentrated at the bottom of the filter bag because a large amount of dust adhered to the bottom increases the equivalent mass of the filter bag and seriously hinders the passage of the acoustic wave;

[0127] The change of the B mode is that the signal attenuation amount moderately increases, the propagation time difference is slightly positive or unchanged, and the spectral distortion value significantly increases, which shows that the spectral shape changes, and it is determined that the uniform paste bag occurs at the middle and upper part of the filter bag because the uniformly distributed dust layer changes the overall acoustic impedance characteristics of the filter bag, which has different absorption effects on acoustic waves of different frequencies, resulting in spectral distortion;

[0128] The change of the C mode is that the signal attenuation amount and the propagation time difference do not change much, but the spectral distortion value increases, which may indicate that the small range of local hardening is not enough to significantly affect the overall propagation path, but it will scatter acoustic waves of specific frequencies;

[0129] Based on the above all judgments, a structured diagnostic record is generated for each detected target filter bag, which integrates the diagnostic results of the filter bag ID information, whether it is abnormal, whether it is a physical blockage, and the rough positioning information of the abnormal area.

[0130] S4, based on the diagnostic results, a parameterized dust removal instruction is adaptively generated, including specifying the target filter bag ID and the dust removal parameters, wherein the dust removal parameters include the pulse width extension amount, the continuous pulse blowing times, when the pulse blowing control system receives the dust removal instruction, the dust removal action is performed on the specified target filter bag, after the dust removal is completed, the system collects the real-time flow data and acoustic signal of the target filter bag as the effect data, and stores it in association with the dust removal parameters in the dust removal instruction, through long-term data accumulation, the dust removal strategy is optimized by using machine learning algorithm, and the optimal dust removal parameters are adaptively learned, at the same time, the new flow data of the filter bag after dust removal is fed back to the individualized dynamic flow health baseline model for fine tuning.

[0131] Specifically, the adaptive generation steps of the parameterized dust removal instruction are as follows:

[0132] According to the diagnostic results, the abnormality of the target filter bag is determined, and based on the historical dust removal data and experimental results, a mapping relationship between the diagnostic results and the dust removal parameters of the pulse width extension amount, the continuous pulse blowing times is established, and a dust removal strategy mapping table is obtained;

[0133] The diagnostic result is divided into different abnormal types and severities, and the ash cleaning strategy mapping table defines the ash cleaning parameter adjustment reference values corresponding to different abnormal types and severities;

[0134] According to the total air volume and temperature condition of the current ash cleaning system operation, in combination with the historical ash cleaning record of the target filter bag, the initial ash cleaning parameter is linearly self-adaptively adjusted, wherein the expression of the linear self-adaptive adjustment of the injection pressure increment is , wherein, represents the injection pressure increment after self-adaptive adjustment, represents the initial value of the injection pressure increment, represents the adjustment coefficient of the influence factor corresponding to the total air volume and temperature condition state of the ash cleaning system operation in the injection pressure increment, represents the influence factor of the total air volume and temperature condition state of the ash cleaning system operation, represents the adjustment coefficient of the historical ash cleaning record influence factor corresponding to the target filter bag in the injection pressure increment, represents the historical ash cleaning record influence factor of the target filter bag;

[0135] The expression of the self-adaptive adjustment of the pulse width extension amount is , wherein, represents the pulse width extension amount after self-adaptive adjustment, represents the initial value of the pulse width extension amount, represents the adjustment coefficient of the influence factor corresponding to the total air volume and temperature condition state of the ash cleaning system operation in the pulse width extension amount, represents the adjustment coefficient of the historical ash cleaning record influence factor corresponding to the target filter bag in the pulse width extension amount;

[0136] The expression of the self-adaptive adjustment of the continuous injection times is , wherein, represents the continuous injection times after self-adaptive adjustment, represents the initial value of the continuous injection times, represents the adjustment coefficient of the influence factor corresponding to the total air volume and temperature condition state of the ash cleaning system operation in the continuous injection times, represents the adjustment coefficient of the historical ash cleaning record influence factor corresponding to the target filter bag in the continuous injection times;

[0137] The adjusted ash cleaning parameter is combined with the target filter bag to generate a parameterized ash cleaning instruction, which clearly indicates the target filter bag object for ash cleaning treatment and specifically controls the injection pressure increment, the pulse width extension amount, and the continuous injection times.

[0138] Specifically, the steps of optimizing the ash cleaning strategy by the machine learning algorithm are as follows:

[0139] The real-time flow data and the sound signal of the target filter bag collected after the dust cleaning are stored in association with the dust cleaning parameters in the dust cleaning instruction adopted, forming a data set;

[0140] The data set is subjected to feature selection and standardization processing for removing irrelevant or redundant features and making different features have the same scale;

[0141] The data set is trained using a machine learning algorithm of a regression tree to establish a regression prediction model between the dust cleaning parameters and the dust cleaning effect;

[0142] The performance of the regression prediction model is evaluated using a cross-validation method, and the parameters of the regression prediction model are adjusted according to the evaluation result to optimize the regression prediction model;

[0143] The running state of the current dust cleaning system and the diagnosis result of the abnormal target filter bag are input into the optimized regression prediction model to predict the optimal dust cleaning parameters.

[0144] Embodiment 2

[0145] The application provides a kind of dust removal equipment filter bag dynamic dust cleaning system as shown in Figure 2 The application provides a kind of dust removal equipment filter bag dynamic dust cleaning system as shown in

[0146] The data acquisition and baseline modeling processing module acquires the data of the micro hot mass flow sensor and the sound wave receiving sensor, establishes a personalized dynamic flow health baseline model for each filter bag using statistical process control (SPC) method, records the sound wave signal characteristics of the filter bag in a healthy state and establishes a sound wave conduction baseline file.

[0147] The diagnostic analysis and dust cleaning decision module identifies preliminary abnormal filter bags based on the personalized dynamic flow health baseline model of the filter bag using the Isolation Forest unsupervised anomaly detection algorithm, verifies the existence of physical blockage in the filter bag and roughly locates the blockage area according to the comparative analysis result of the sound wave signal, and generates a parameterized dust cleaning instruction according to the diagnosis result.

[0148] The dust cleaning parameter control and feedback learning module controls the pulse spray control system to execute the dust cleaning instruction, collects the effect data after dust cleaning and stores it in association with the dust cleaning parameters, optimizes the dust cleaning strategy using a machine learning algorithm, and feeds back the new flow data of the filter bag after dust cleaning to the personalized dynamic flow health baseline model for fine tuning.

[0149] Specifically, the pulse blowing control system connected with the dust removal equipment filter bag dynamic cleaning system is used for regulating and controlling the dust removal parameters of the blowing pressure increment, the pulse width extension and the continuous blowing times, and performing individualized dust removal actions on the target filter bag.

[0150] The dust removal equipment filter bag dynamic cleaning system provided by the embodiment of the application is implemented by the dust removal equipment filter bag dynamic cleaning method, and the specific method and process of the dust removal equipment filter bag dynamic cleaning system are described in the embodiment of the dust removal equipment filter bag dynamic cleaning method, which will not be described here.

[0151] The above formulas are all dimensionless values, and the formulas are obtained by collecting a large amount of data to simulate a formula of the most recent real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0152] The above embodiments can be realized by software, hardware, firmware or any combination thereof, in whole or in part. When realized by software, the above embodiments can be realized in the form of a computer program product in whole or in part. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices. The computer instructions can be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired or wireless (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center and the like containing one or more available medium sets. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD) or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0153] It should be understood that in various embodiments of the application, the size of the sequence number of each process does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the application.

[0154] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0155] The above description is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for dynamic dust removal of filter bags in a dust collector, characterized in that, Includes the following steps: S1. Install a miniature thermal mass flow sensor at the outlet of each clean air chamber of the dust removal equipment corresponding to each filter bag. Install a sound wave exciter on the filter bag cleaning blow pipe and a sound wave receiving sensor at the bottom of the filter bag. Activate the sensor under the initial working condition of confirming the health of the filter bag, and continuously collect the real-time flow data of each filter bag under different overall working conditions. Use the statistical process control method to establish a personalized dynamic flow health baseline model for each filter bag that is linked to the overall working condition. At the same time, perform sound wave scanning on each filter bag, record the characteristics of the sound wave signal, and establish a sound wave transmission baseline file. S2. When the dust removal system is running normally, based on the continuous collection of real-time flow data of each filter bag, the deviation of the flow value of each filter bag from the personalized dynamic baseline is calculated. The default threshold is preset, and the deviation data of all filter bags within the continuous time window is input into the isolated forest unsupervised anomaly detection algorithm. The anomaly score is automatically calculated and then compared with the default threshold to identify and mark the preliminary abnormal filter bags and generate an anomaly list containing the filter bag ID. S3. Based on the initial abnormal filter bag, the acoustic signal emitted by the acoustic exciter of the target filter bag is controlled to be received by the acoustic wave receiving sensor. The change in acoustic characteristic parameters is compared with the acoustic wave transmission baseline file to verify the abnormality. The results are generated by comparing and analyzing the preset acoustic characteristic thresholds. S4. Based on the diagnostic results, the parameterized cleaning command is adaptively generated. When the pulse jet cleaning control system receives the cleaning command, it performs the cleaning action on the specified target filter bag. Through long-term data accumulation, the cleaning strategy is optimized by machine learning algorithm, and the optimal cleaning parameters are adaptively learned. The steps for collecting real-time traffic data are as follows: At the outlet of each clean air chamber of the dust removal equipment, a miniature thermal mass flow sensor is installed corresponding to each filter bag. An acoustic exciter is installed on the filter bag cleaning blowpipe and an acoustic receiving sensor is installed at the bottom of the filter bag. A total operating condition sensor for air volume and temperature is installed at the air inlet of the dust removal equipment. All sensors are initialized and put into normal working condition. Before starting the sensor to collect data, it is necessary to confirm that the filter bag is in a healthy initial operating condition; Under the initial operating condition of confirming the health of the filter bags, a synchronous acquisition command is sent to start the miniature thermal mass flow sensor and the total operating condition sensor, and to start continuously collecting the real-time flow data of each filter bag, and to associate and record the timestamp, the real-time volumetric flow rate value of each filter bag, the total processing air volume at the corresponding time, and the raw data of the flue gas inlet temperature at the corresponding time. The real-time flow readings of each sensor are processed by a moving average filter. Within a 24-hour self-learning cycle, the original data collection and processing are repeated to cover as many different working conditions as possible, and all the data collected within the cycle are used to build a time series dataset. The steps for establishing the personalized dynamic traffic health baseline model are as follows: Real-time flow data of each filter bag under different total air volume and temperature conditions is extracted from the time series dataset. The data is grouped according to the total air volume and temperature conditions, with each group representing a specific combination of conditions, thus realizing the binning of operating condition parameters. All data points in the time series dataset are categorized into the operating chambers with different total air volume and temperature ranges according to their corresponding total air volume and temperature values. Using the Statistical Process Control (SPC) method, for each operating chamber, the baseline value and control limits are defined by calculating the sample mean and sample standard deviation of all filter bag flow data falling into each chamber. The control limits of each working condition box are associated with the corresponding total air volume and temperature conditions. After traversing all working condition boxes, a personalized dynamic flow health baseline model for each filter bag is constructed. Based on the personalized dynamic flow health baseline model, input the current total air volume and temperature conditions, and output the expected baseline value and normal fluctuation range of the flow rate of each filter bag under the current conditions by querying the corresponding control limits. For the working condition of falling between the bins, the corresponding upper control limit UCL, lower control limit LCL and control limits of the center mean are calculated by linear interpolation. The steps for establishing the acoustic wave transmission baseline profile are as follows: Under the initial working condition of confirming the health of the filter bags, the acoustic exciter is controlled to perform an acoustic scan on each filter bag. The acoustic exciter emits an acoustic signal of a specific frequency and intensity, which is received by the acoustic receiving sensor after propagating through the filter bag. For the first Strip filter bag repeat The infrasound signal excitation-acquisition process, and... The received acoustic signals are superimposed and averaged to obtain a high-quality averaged time-domain signal. The formula for calculating the averaged time-domain signal is as follows: In the formula, Represented as the first Strip filter bags in repeated Infrasound signal excitation-acquisition process The average time-domain signal at time t, This represents the total number of repetitive acoustic signal excitation-acquisition processes. Represented as the first The filter bag in the first Infrasound signal excitation-acquisition process The sound wave signals received at all times; The time difference between the transmission and reception times of the acoustic signal, the peak value of the measured signal, and the spectrum obtained by performing a Fast Fourier Transform on the average time-domain signal are calculated from a high-quality average time-domain signal to extract and record the acoustic signal characteristics of propagation time, amplitude, and spectrum. The expression for propagation time is: In the formula, Represented as the first The propagation time of the scanning acoustic signal of the filter bag. This represents the moment when the sound wave receiving sensor receives the signal. This indicates the moment when the acoustic wave exciter emits a signal; The expression for amplitude is In the formula, Represented as the first The amplitude of the average time-domain signal of the filter bag. This is expressed as the maximum amplitude taken from the average time-domain signal. This is represented by the specific frequency and intensity of a sound wave signal; The expression for the spectrum is In the formula, Represented as the first The spectrum of the acoustic signal of the filter bag. This is represented as a Fast Fourier Transform operation on the average time-domain signal; The acoustic signal features extracted from each filter bag are used as the acoustic wave conduction baseline profile of the corresponding filter bag in a healthy state. The acoustic wave conduction baseline profile includes the ID information of the filter bag and the health benchmark value range of the acoustic signal features.

2. The method for dynamic dust removal of filter bags in a dust removal device according to claim 1, characterized in that, The steps for calculating the deviation are as follows: Acquire real-time flow data of each filter bag continuously collected, and at the same time acquire the personalized dynamic baseline established for each filter bag; For each filter bag, the real-time flow data is calculated to obtain the deviation value by comparing it with the personalized dynamic baseline centerline. Standardize the deviation values ​​and calculate the degree of deviation.

3. The dynamic dust removal method for filter bags in a dust removal device according to claim 2, characterized in that, The steps for calculating the anomaly score using the IsolationForest unsupervised anomaly detection algorithm are as follows: The deviation data of all filter bags within a continuous sliding time window are combined into a deviation dataset; q isolated trees were constructed by randomly sampling the deviation feature data; The deviation dataset is input into a pre-configured IsolationForest unsupervised anomaly detection algorithm for training. For each sample data point in the deviation dataset, calculate the path length from the root node to the leaf node in each isolated tree; Next, calculate the average path length of each sample data point across all isolated trees; The anomaly score for the corresponding sample data point is calculated based on the average path length.

4. The dynamic dust removal method for filter bags in a dust removal device according to claim 3, characterized in that, The steps for verifying anomalies in the changes of acoustic feature parameters are as follows: After receiving the filter bags marked as initially abnormal, acoustic tests are initiated for each target filter bag in the abnormality list in sequence; The acoustic exciter of the target filter bag in the anomaly list emits an acoustic signal of a specific frequency, while the acoustic signal received by the acoustic receiving sensor is recorded. The acoustic signal excitation-acquisition process is repeated multiple times on the target filter bag. The received acoustic signals are then time-domain superimposed and averaged to obtain a high signal-to-noise ratio average time-domain signal. Acoustic characteristic parameters are obtained by extracting the amplitude, propagation time, and spectrum of the average time-domain signal from the average time-domain signal. The acoustic feature parameters extracted from the current target filter bag are subtracted from the range of health benchmark values ​​stored in the acoustic wave transmission baseline file to calculate the change in acoustic feature parameters. The changes in acoustic feature parameters are compared with preset acoustic feature thresholds to determine whether the target filter bag is truly abnormal.

5. The dynamic dust removal method for filter bags in a dust removal device according to claim 4, characterized in that, The steps for generating the diagnostic results are as follows: The changes in acoustic characteristic parameters calculated for each filter bag are compared one by one with the preset acoustic characteristic thresholds. If the change in any acoustic characteristic parameter exceeds the corresponding acoustic characteristic threshold, it is determined that the filter bag has a physical blockage abnormality. If the acoustic characteristic parameters calculated for a filter bag are all within the normal range of the preset acoustic characteristic threshold, the filter bag is determined to be a false alarm and the filter bag is removed from the abnormal list. For filter bags that have undergone acoustic testing to verify physical abnormalities, the blockage area can be roughly located by analyzing the change patterns of acoustic characteristic parameters. This includes Mode A, which roughly locates severe blockage at the bottom of the filter bag; Mode B, which roughly locates uniform clogging in the upper and middle parts of the filter bag; and Mode C, which roughly locates localized caking in the filter bag. Based on all the above judgments, a structured diagnostic record is generated for each target filter bag being tested, which integrates the filter bag ID information, whether it is abnormal, whether it is a physical blockage, and the coarse location information of the abnormal area.

6. The method for dynamic dust removal of filter bags in a dust removal device according to claim 5, characterized in that, The adaptive generation steps of the parameterized dust removal command are as follows: Based on the diagnostic results, the abnormality of the target filter bag is determined, and based on historical cleaning data and experimental results, a mapping relationship is established between the diagnostic results and cleaning parameters such as the increase in pulse pressure, the extension of pulse width, and the number of consecutive pulses, resulting in a cleaning strategy mapping table. The diagnostic results are divided into different abnormality types and severity levels, and the dust removal strategy mapping table defines the baseline values ​​for adjusting the dust removal parameters corresponding to different abnormality types and severity levels. Based on the total air volume and temperature conditions of the current dust removal system, and combined with the historical dust removal records of the target filter bag, the initial dust removal parameters are linearly and adaptively adjusted. The adjusted cleaning parameters are combined with the target filter bag to generate parameterized cleaning instructions.

7. A dynamic dust removal system for filter bags in a dust removal device, implemented by a dynamic dust removal method for filter bags in a dust removal device as described in any one of claims 1-6, characterized in that, It includes a distributed sensing and detection module: a miniature thermal mass flow sensor installed at the outlet of each clean air chamber of the dust removal equipment corresponding to each filter bag, an acoustic exciter installed on the filter bag cleaning jet pipe, and an acoustic receiving sensor installed at the bottom of the filter bag, to monitor the filter bags in the dust removal equipment in real time; Data acquisition and baseline modeling processing module: Acquires data from miniature thermal mass flow sensor and acoustic wave receiving sensor, uses statistical process control (SPC) to establish a personalized dynamic flow health baseline model for each filter bag, records the acoustic signal characteristics of the filter bag in its health state and establishes an acoustic wave transmission baseline profile. Diagnostic Analysis and Cleaning Decision Module: Based on the personalized dynamic flow health baseline model of filter bags, the IsolationForest unsupervised anomaly detection algorithm is used to identify preliminary abnormal filter bags. The abnormality is verified by the comparison and analysis results of acoustic signals, the presence of physical blockage in the filter bags is diagnosed and the blockage area is roughly located, and parameterized cleaning instructions are generated based on the diagnostic results. The dust removal parameter control and feedback learning module controls the pulse jet control system to execute dust removal commands, collects the effect data after dust removal and stores it in association with the dust removal parameters, optimizes the dust removal strategy using machine learning algorithms, and feeds back the new flow data of the filter bag after dust removal to the personalized dynamic flow health baseline model for fine-tuning.

Images