Air cannon cluster cooperative control method and system based on internet of things

Abstract

The application relates to the technical field of Internet of Things and intelligent manufacturing, and discloses an air cannon cluster cooperative control method and system based on the Internet of Things. The method comprises the following steps: obtaining dust density and pressure change data fusion to generate initial state information; detecting adjacent nodes to generate an available link set and a quality index table; performing multi-hop broadcasting and fusion to obtain a shared data set, and performing Kalman filtering denoising to generate a situation model; combining the quality index table to evaluate node urgency and obtain a priority sequence; determining a target device according to the priority sequence to adjust cleaning and unblocking parameters, and adaptively updating an initial threshold value based on an execution result. The method can break information islands, realize bottom layer data anti-destroying synchronization, and realize accurate judgment and dynamic rearrangement of cleaning and unblocking urgency in a multi-concurrent blocking scene, so that the long-term operation stability of an industrial storage and transportation system is maximized.

Description

Technical Field

[0001] This invention relates to the fields of Internet of Things (IoT) and intelligent manufacturing technology, and in particular to a collaborative control method and system for an air cannon cluster based on IoT. Background Technology

[0002] Currently, in the storage and transportation of dust or particulate materials in coal, chemical, and metallurgical industries, material caking and blockage in silos and pipelines are a major bottleneck restricting production efficiency. Air cannons, as devices that use instantaneous high-pressure airflow to break up material arching, are widely used in such scenarios. Moreover, with the expansion of production line scale, cluster operations consisting of multiple air cannons have become an inevitable trend to ensure smooth industrial processes.

[0003] In existing technologies, a centralized industrial control system or fixed manual experience rules are typically used to issue timed trigger commands to control the spraying action of an air cannon cluster. However, this centralized, unidirectional control method is highly susceptible to the negative impact of unstable or even interrupted communication links in complex environments with widely distributed equipment and severe dust interference. Once the underlying network fails, the individual equipment nodes become information islands, unable to perceive the pressure status and blockage severity of adjacent equipment in real time. This control mode, lacking dynamic topology reconfiguration and multi-point coordination, often leads to frequent ineffective air spraying in areas with minor blockages, accelerating equipment mechanical wear and wasting a large amount of air; while critical nodes with truly severe blockages fail to receive a centralized response from the cluster, resulting in increased material accumulation, failure of the clearing task, and ultimately, operational conflicts within the area.

[0004] Existing technologies suffer from low efficiency in clearing blockages and are prone to causing system stagnation. Summary of the Invention

[0005] This invention provides a collaborative control method and system for air cannon clusters based on the Internet of Things, in order to solve the problems of low clearing efficiency and easy system stagnation in the prior art.

[0006] Firstly, to address the aforementioned technical problems, this invention provides a collaborative control method for an air cannon cluster based on the Internet of Things, comprising:

[0007] Dust density data and pressure change data are acquired and weighted to obtain a fusion index. Based on the fusion index, feature fusion is performed in combination with a preset initial threshold to generate initial state information.

[0008] Based on the initial state information, probe packets are sent to adjacent nodes in the preset initial routing table to obtain response time and packet loss rate. The reciprocals of the response time and packet loss rate are calculated respectively, and the obtained reciprocals are weighted to obtain the link score. Based on the link score, an available link set and a quality index table are generated.

[0009] The initial state information is broadcast in multiple hops using the available link set to obtain monitoring values, and the monitoring values ​​are then fused to obtain a shared dataset.

[0010] The shared dataset and the preset noise matrix are denoised using Kalman filtering to obtain the situation model;

[0011] The situation model and the quality index table are weighted to obtain an evaluation score vector. The evaluation score vector is then normalized using the Softmax function to obtain the urgency score of each node. The urgency scores are then sorted in descending order to obtain a priority sequence.

[0012] The target device is determined according to the priority sequence, and the parameters of the target device are adjusted in combination with the preset clearing instruction set to obtain the execution result;

[0013] The threshold compensation amount is obtained by performing compensation calculation based on the execution result, and the preset initial threshold is updated based on the threshold compensation amount to obtain the task completion status.

[0014] Secondly, the present invention provides an Internet of Things-based air cannon cluster collaborative control system, comprising:

[0015] The state generation module is used to acquire dust density data and pressure change data, perform weighted calculations to obtain a fusion index, and perform feature fusion based on the fusion index and a preset initial threshold to generate initial state information.

[0016] The link evaluation module is used to send probe packets to adjacent nodes in the preset initial routing table according to the initial state information to obtain the response time and packet loss rate, calculate the reciprocal of the response time and the packet loss rate respectively, and calculate the link score by weighting the obtained reciprocals, and generate an available link set and a quality index table based on the link score.

[0017] The data synchronization module is used to combine the available link set to perform multi-hop broadcasting of the initial state information to obtain monitoring values, and to perform fusion processing on the monitoring values ​​to obtain a shared dataset;

[0018] The situation denoising module is used to denoise the shared dataset and the preset noise matrix using Kalman filtering to obtain the situation model;

[0019] The priority ranking module is used to perform weighted calculations on the situation model and the quality index table to obtain an evaluation score vector, and to normalize the evaluation score vector using the Softmax function to obtain the urgency score of each node. The urgency scores are then sorted in descending order to obtain a priority sequence.

[0020] The congestion clearing execution module is used to determine the target device according to the priority sequence, and adjust the parameters of the target device in combination with the preset congestion clearing instruction set to obtain the execution result;

[0021] The closed-loop consensus module is used to calculate the threshold compensation amount based on the execution result, and update the preset initial threshold based on the threshold compensation amount to obtain the task completion status.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) This invention generates status information by fusing local device features and performs decentralized multi-hop broadcasting and data sharing based on dynamically detected link quality. This mechanism breaks the information silo dilemma of traditional centralized industrial control systems in harsh communication environments. It can reconstruct the available communication topology in real time when local nodes are disconnected, ensuring the resilience and global transparency of the cluster's underlying status data, and laying a highly reliable data foundation for multi-device collaborative clearing of blockages.

[0024] (2) This invention introduces a Kalman filter model to perform deep denoising on multi-dimensional time-series features, and combines link quality and global situation to perform multi-factor weighting and Softmax normalization. This scheme abandons the traditional single fixed rule judgment, and cleverly integrates the physical severity of global congestion with the control quality of the underlying link in the mathematical implicit space, realizing accurate judgment of the urgency of clearing congestion and dynamic priority reordering in multi-concurrent congestion scenarios, and completely eliminating resource contention and blind empty firing caused by disordered execution.

[0025] (3) This invention independently decouples and outputs multi-dimensional physical adjustment coefficients by using a PID control algorithm based on priority sequence, providing closed-loop guidance for the precise execution of the target equipment, and dynamically and adaptively calibrating the event trigger threshold based on the physical improvement rate. This system not only achieves physical-level customized strikes for different blockage situations, significantly reducing unnecessary mechanical losses, but also endows the entire air cannon cluster with the ability to continuously learn and self-correct under complex and variable working conditions, maximizing the long-term operational stability of the industrial storage and transportation system. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the IoT-based collaborative control method for air cannon clusters provided in the first embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the structure of the Internet of Things-based air cannon cluster collaborative control system provided in the second embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Reference Figure 1 The first embodiment of the present invention provides a collaborative control method for an air cannon cluster based on the Internet of Things, including the following steps:

[0030] S11, acquire dust density data and pressure change data, perform weighted calculation to obtain fusion index, and perform feature fusion based on the fusion index and a preset initial threshold to generate initial state information;

[0031] S12, based on the initial state information, send probe packets to adjacent nodes in the preset initial routing table to obtain response time and packet loss rate, calculate the reciprocal of the response time and the packet loss rate respectively, and perform weighted calculation on the obtained reciprocals to obtain link score, and generate an available link set and quality index table based on the link score;

[0032] S13, combine the available link set to perform multi-hop broadcasting of the initial state information to obtain monitoring values, and perform fusion processing on the monitoring values ​​to obtain a shared dataset;

[0033] S14, the shared dataset and the preset noise matrix are denoised using Kalman filtering to obtain the situation model;

[0034] S15, the situation model and the quality index table are weighted to obtain the evaluation score vector, and the evaluation score vector is normalized using the Softmax function to obtain the urgency score of each node. The urgency scores are sorted in descending order to obtain the priority sequence.

[0035] S16, determine the target device according to the priority sequence, and adjust the parameters of the target device in combination with the preset clearing instruction set to obtain the execution result;

[0036] S17, calculate the threshold compensation amount based on the execution result, and update the preset initial threshold based on the threshold compensation amount to obtain the task completion status.

[0037] In step S11, dust density data and pressure change data are acquired and weighted to obtain a fusion index. Based on the fusion index, feature fusion is performed in conjunction with a preset initial threshold to generate initial state information, including:

[0038] Data on dust density and pressure changes of the air cannon cluster are acquired through sensors.

[0039] The dust density data is multiplied by a preset dust density weight to obtain a first value, and the pressure change data is multiplied by a preset pressure change rate weight to obtain a second value.

[0040] The first value and the second value are added together to obtain a fusion index, and it is determined whether the fusion index is greater than the preset initial threshold.

[0041] If the fusion index is greater than the preset initial threshold, an event trigger flag is generated, and the event trigger flag, the dust density data, and the pressure change data are spliced ​​and recombined to obtain the initial state information.

[0042] It should be noted that this embodiment acquires dust density data using an electrostatic induction dust concentration transmitter, which utilizes the weak charge induction signal generated when dust particles pass through the sensor and measures dust concentration through AC coupling technology; and acquires pressure change data using a high-frequency response pressure transmitter, with a response frequency of up to 2kHz, capable of capturing the instantaneous high-pressure airflow of the air cannon. Regarding the method for determining the preset dust density weight and the preset pressure change rate weight, this embodiment collects time-series data of dust density and pressure change rate from the air cannon cluster during its historical stable operating period, and calculates the distribution variance of the dust density time-series data and the distribution variance of the pressure change rate time-series data respectively. This embodiment divides the distribution variance of the dust density time-series data by the sum of the two distribution variances, and objectively sets the resulting dimensionless ratio as the preset dust density weight; this embodiment also divides the distribution variance of the pressure change rate time-series data by the sum of the two distribution variances, and objectively sets the resulting dimensionless ratio as the preset pressure change rate weight.

[0043] In one implementation, this embodiment multiplies the dust density data by the dust density weight to obtain a first value, and multiplies the pressure change data by the pressure change rate weight to obtain a second value. This embodiment then adds the calculated first value and the second value to obtain a single fusion index. This embodiment uses comparator logic to determine whether the fusion index is greater than a preset initial threshold.

[0044] It is worth noting that, regarding the process of determining the preset initial threshold, this embodiment extracts the time period within ten minutes prior to the occurrence of a real physical blockage event recorded in the historical equipment maintenance log, and extracts all corresponding historical dust density data and historical pressure change data within this time period; using the same weighted summation calculation method as described above, the historical fusion index sequence for each moment within this time period is calculated. This embodiment calculates the arithmetic mean and standard deviation of the historical fusion index sequence, subtracts twice the standard deviation from the arithmetic mean, and strictly sets the calculated difference as the preset initial threshold.

[0045] In one implementation, if the fusion index is determined to be greater than the preset initial threshold, this embodiment generates an event trigger flag. The event trigger flag is represented by a high-level digital signal or a specific status character. In this embodiment, the event trigger flag, the dust density data, and the pressure change data are concatenated and combined in a fixed byte order to generate a one-dimensional data sequence string, which is then output as the initial state information.

[0046] For example, assume that the current dust density data acquired by the sensor is 8.5 mg / m³, and the pressure change data is 7.0 kPa / s. To eliminate the dimensional difference between mg / m³ and kPa / s, this embodiment extracts the physical extreme values ​​recorded by the system during historical stable operation tests. The historical dust density is a minimum of 0.5 mg / m³ and a maximum of 10.5 mg / m³; the historical pressure change is a minimum of 2.0 kPa / s and a maximum of 12.0 kPa / s. This embodiment employs a minimax normalization algorithm for data mapping. The current dust density data (8.5) is subtracted from the minimum value of 0.5 to obtain 8.0, and the maximum value (10.5) is subtracted from the minimum value of 0.5 to obtain a denominator of 10.0. Dividing 8.0 by 10.0 yields a normalized dust density value of 0.8. Similarly, the current pressure change data (7.0) is subtracted from the minimum value of 2.0 to obtain 5.0, and the maximum value (12.0) is subtracted from the minimum value of 2.0 to obtain a denominator of 10.0. Dividing 5.0 by 10.0 yields a normalized pressure change value of 0.5. A preset dust density weight of 0.6 and a preset pressure change rate weight of 0.4 are determined based on the historical variance ratio. This embodiment multiplies the normalized dust density value (0.8) by 0.6 to obtain a first value of 0.48, and multiplies the normalized pressure change value (0.5) by 0.4 to obtain a second value of 0.20. This embodiment adds 0.48 and 0.20 to obtain a fusion index of 0.68. The preset initial threshold, determined by statistical analysis of normalized data from historical congestion events, is 0.65. In this embodiment, it is determined that 0.68 is greater than 0.65, and an event trigger flag with the character "1" is generated. The character "1", the original dust density value "8.5", and the original pressure change value "7.0" are then concatenated in sequence to obtain the sequence "1_8.5_7.0", which is output as the initial state information.

[0047] In step S12, probe packets are sent to neighboring nodes in the preset initial routing table based on the initial state information to obtain response time and packet loss rate. The reciprocals of the response time and packet loss rate are calculated respectively, and the resulting reciprocals are weighted to obtain a link score. Based on the link score, an available link set and a quality index table are generated, including:

[0048] Based on the initial state information, link probe packets are sent to adjacent nodes in the preset initial routing table, and timestamps are recorded to obtain response time and packet loss rate;

[0049] Obtain a preset response time weight and a preset packet loss rate weight, and calculate the reciprocal of the response time and the packet loss rate respectively to obtain the reciprocal of the response time and the reciprocal of the packet loss rate;

[0050] The third value is obtained by multiplying the reciprocal of the response time by the response time weight, and the fourth value is obtained by multiplying the reciprocal of the packet loss rate by the packet loss rate weight.

[0051] The link score is obtained by adding the third value and the fourth value together.

[0052] The adjacent nodes whose link scores are greater than a preset link quality threshold are extracted and connected topologically to obtain an available link set. The link scores of all nodes in the available link set are extracted and merged to obtain a quality index table.

[0053] In one implementation, after generating initial state information, this embodiment uses it as a trigger signal to extract the physical neighbor node addresses recorded in a preset initial routing table. This embodiment concurrently sends a preset number (e.g., 10) of link probe packets to each neighboring node. The link probe packets are encapsulated using the UDP protocol, and the packet header includes a sender timestamp. When a neighboring node receives and returns a response packet, this embodiment records the reception timestamp and calculates the difference between the reception timestamp and the sender timestamp to obtain the round-trip delay. This embodiment also takes the arithmetic average of the delays over a preset number of probe periods to obtain the response time. Simultaneously, this embodiment counts the number of probe packets that do not receive a response within the preset period and divides this number by the total number of probe packets sent to obtain the packet loss rate.

[0054] The preset initial routing table is configured by manually configuring the network address of each air cannon node and the network addresses of its spatially adjacent nodes according to the actual physical topology during the air cannon cluster installation and deployment phase, and storing the configuration in the non-volatile memory of each node; if the network topology changes later, the routing table can be updated through a remote configuration protocol.

[0055] It should be noted that, regarding the method for determining the preset response time weight and the preset packet loss rate weight, this embodiment uses the coefficient of variation method for objective calibration. This embodiment collects multiple sets of response time samples and packet loss rate samples from the network during its historical operating period, calculates the standard deviation and mean of the response time, and divides the standard deviation by the mean to obtain the response time coefficient of variation; similarly, the packet loss rate coefficient of variation is calculated. This embodiment divides the response time coefficient of variation by the sum of the two coefficients of variation, and the result is strictly set as the preset response time weight; the packet loss rate coefficient of variation is divided by the sum of the two, and the result is set as the preset packet loss rate weight.

[0056] In one implementation, this embodiment calculates the reciprocal of the response time and the packet loss rate, respectively, to obtain the reciprocal of the response time and the reciprocal of the packet loss rate. It should be noted that the purpose of using reciprocal calculations is to transform a physical quantity where smaller is better into a larger is better, facilitating subsequent positive weighted scoring. In this embodiment, the reciprocal of the response time is multiplied by the preset response time weight to obtain a third value, and the reciprocal of the packet loss rate is multiplied by the preset packet loss rate weight to obtain a fourth value. This embodiment sums the third value and the fourth value to obtain a link score reflecting the overall communication quality of the link.

[0057] If the packet loss rate is zero, its reciprocal is set to a preset upper limit value, which is twice the maximum reciprocal of all historical non-zero packet loss rate samples, in order to avoid infinite values ​​causing distortion of the link score. Similarly, in the normalization process of step S11, if the current sensor value exceeds the historical extreme value range, it is truncated to the boundary value of the historical extreme value, that is, the minimum value is taken when it is less than the minimum value, and the maximum value is taken when it is greater than the maximum value.

[0058] It is worth noting that this embodiment uses the percentile method to determine the preset link quality threshold. This embodiment extracts all link score data of the air-gun cluster under historical communication environments and constructs a score distribution probability density function. This embodiment extracts the score value corresponding to the 15th percentile cumulative probability from this distribution function and objectively sets it as the preset link quality threshold to ensure the elimination of extremely degraded links from the network noise floor.

[0059] In one implementation, this embodiment iterates through the link scores of all adjacent nodes and determines whether the link score is greater than the preset link quality threshold. If it is greater, the device identifier of the adjacent node is extracted and stored in the available link set. Subsequently, this embodiment extracts the device identifier and link score corresponding to each node from the available link set, merges and reassembles them in the form of key-value pairs, and generates a quality index table reflecting the current local network topology quality.

[0060] It should be noted that if the link score is not greater than the preset link quality threshold, the current communication quality of the adjacent node is determined to be unsatisfactory, and the node will not be added to the available link set, nor will its link score be recorded in the quality index table. In subsequent multi-hop broadcasts, the node will be excluded from the transmission path. At the same time, the node will remain in the initial routing table, and the link quality will be reassessed when the next probe cycle arrives. If its link score recovers to be greater than the threshold, it can be re-added to the available link set.

[0061] For example, assume that 10 probe packets are sent to neighboring node A, with an average response time of 20 milliseconds and a packet loss rate of 0.1%. The response time weight, determined by the coefficient of variation method, is 0.7, and the packet loss rate weight is 0.3. In this embodiment, the reciprocal of the response time is calculated as... The reciprocal of the packet loss rate is Calculate the third value as follows: The fourth value is The sum of these scores yields a link score of 3.035. The preset link quality threshold, determined through historical data statistics, is 0.5. Since 3.035 is greater than 0.5, this embodiment adds node A to the available link set and records the corresponding score of 3.035 in the quality index table.

[0062] In step S13, the initial state information is broadcast multi-hop using the available link set to obtain monitoring values, and the monitoring values ​​are then fused to obtain a shared dataset, including:

[0063] The initial state information is calculated using the MD5 hash algorithm to obtain a checksum, and the initial state information and the checksum are serialized and encapsulated to obtain a transmission data packet;

[0064] Based on the Gossip synchronization protocol, the data packets are broadcast over the network through the set of available links.

[0065] If the transmission delay is less than the preset delay threshold, the data packets returned by each node in the available link set are received, parsed, and the monitoring values ​​are obtained.

[0066] A weighted average algorithm is used to calculate the shared average value of the monitored values, and the shared average value is bound to the timestamp to obtain a shared dataset.

[0067] In one implementation, this embodiment extracts the initial state information generated in the aforementioned steps and inputs it into the core operation module of the MD5 hash algorithm. After nonlinear function transformation and bit operations, a fixed-length feature byte sequence is output and determined as the checksum. Subsequently, this embodiment employs a protocol buffer serialization mechanism, using the initial state information as the data body and the checksum as an appended field, and performs binary encoding compression according to a predefined structure format to generate a transmission data packet.

[0068] In one implementation, this embodiment is based on the Gossip synchronization protocol, and broadcasts the transmission data packet over the network through the available link set. This embodiment randomly selects a preset number (e.g., 3) of adjacent nodes from the available link set as target forwarding nodes, and sends the transmission data packet to the target forwarding nodes; and continuously repeats this random selection and sending operation within a preset synchronization period, relying on the mutual contagion forwarding of nodes until all nodes in the available link set have received the transmission data packet.

[0069] It should be noted that this embodiment monitors the round-trip time of data packets in the network in real time to obtain the transmission delay, and determines whether the transmission delay is less than a preset delay threshold. Regarding the method for determining the preset delay threshold, this embodiment collects one thousand round-trip time samples of message transmission from the air-gun network during its historical stable operation phase, constructs a probability density function for the transmission delay, extracts the time delay value corresponding to the 95th percentile of the cumulative probability in the probability density function, and objectively calibrates it as the preset delay threshold.

[0070] It is worth noting that if it is determined that the transmission delay is less than the preset delay threshold, this embodiment receives the data packets returned by each node in the available link set, and recalculates the hash value of the received data packet body using the MD5 hash algorithm; compares the recalculated hash value with the checksum carried at the end of the data packet, and if the two values ​​are completely consistent, performs deserialization parsing on the data packet body, extracts the local environmental parameters of the remote node contained therein, and determines them as monitoring values.

[0071] In one implementation, this embodiment uses a weighted average algorithm to calculate the shared average value from the monitored values. This embodiment extracts the quality index table generated in step S12, and retrieves the link score of the corresponding node based on the node identifier of the returned data packet. This embodiment multiplies the monitored value corresponding to a single node by its corresponding link score to obtain a single product. The sum of the single products corresponding to all nodes is calculated to obtain a weighted sum, which is then divided by the sum of the link scores of all nodes to obtain the shared average value. Subsequently, this embodiment extracts the current absolute timestamp of the system, combines and concatenates the shared average value with this timestamp using key-value pairs, and generates a shared dataset reflecting the global state characteristics of the network.

[0072] It should be noted that the shared dataset does not simply store a shared average value, but rather a structured data table containing timestamps, node identifiers, and their corresponding pressure change data and dust density data; the weighted average calculation is only used to generate an auxiliary shared average value for anomaly detection, but the complete dataset retains the original monitoring values ​​of each node; the pressure change trend vector and dust density distribution matrix extracted in step S14 are obtained by reorganizing the complete data table according to node identifiers and timestamps.

[0073] For example, assuming this embodiment uses the MD5 algorithm to calculate the checksum of the initial state information and serializes and encapsulates it into a transmission data packet. After broadcasting based on the Gossip protocol, the system receives data packets returned by nodes B and C in the available link set. The system monitors that the transmission delay returned by node B is 20 milliseconds and the transmission delay returned by node C is 30 milliseconds. The preset delay threshold objectively determined by historical sample statistics is 50 milliseconds. Since both 20 milliseconds and 30 milliseconds are less than 50 milliseconds and the hash check is consistent, the system parses and obtains the monitoring value of node B as 10.0 and the monitoring value of node C as 15.0. The system looks up the quality index table and obtains the link score of node B as 3.0 and the link score of node C as 2.0. In this embodiment, 10.0 is multiplied by 3.0 to obtain a single product of 30.0, and 15.0 is multiplied by 2.0 to obtain a single product of 30.0; the two single products are added together to obtain a weighted sum of 60.0. In this embodiment, the sum of the link scores is calculated as 3.0 plus 2.0, which equals 5.0. Dividing 60.0 by 5.0 yields a shared average value of 12.0. Finally, the system binds the value 12.0 with the current timestamp (e.g., 10:25:30) and outputs it as a shared dataset.

[0074] In step S14, the shared dataset and the preset noise matrix are denoised using Kalman filtering to obtain a situational model, including:

[0075] Extract the pressure change trend vector and dust density distribution matrix of adjacent timestamps from the shared dataset;

[0076] The pressure change trend vector, the dust density distribution matrix, and the preset noise matrix are subjected to state estimation iterative update operation using the Kalman filter algorithm to obtain a smooth estimation matrix;

[0077] Extract the smoothed pressure vector from the smoothed estimation matrix, and calculate the error reduction rate of the smoothed pressure vector compared to the pressure change trend vector;

[0078] If the error reduction rate is greater than the preset error convergence threshold, the smooth estimation matrix is ​​reduced in dimension to extract the situation model.

[0079] In one implementation, this embodiment arranges the shared dataset in chronological order according to the timestamps. This embodiment extracts the pressure observation values ​​corresponding to the current timestamp and the previous adjacent timestamp, calculates the difference sequence between the two, and constructs a pressure change trend vector. Simultaneously, it extracts the dust density values ​​returned by each node at the current timestamp, arranges them in two dimensions according to a preset node spatial topological coordinate system, and constructs a dust density distribution matrix.

[0080] It should be noted that the process for determining the preset noise matrix includes both the process noise covariance matrix and the measurement noise covariance matrix. This embodiment uses a combination of engineering calibration and statistical analysis to objectively determine this matrix. First, this embodiment collects long-term sampling sequences from each sensor of the air cannon cluster under historical silent conditions (i.e., a stable period with no physical clearing and blowing actions and ambient wind speed approaching zero). The variance distribution of this sampling sequence is calculated, and this variance distribution is arranged in a diagonal matrix form, objectively set as the measurement noise covariance matrix. Second, this embodiment collects state transition observation data under historical standard blowing conditions, calculates theoretical predicted values ​​using an ideal state transition equation without noise, subtracts the theoretical predicted values ​​from the actual observed values ​​to obtain a residual sequence, calculates the covariance of this residual sequence, and strictly sets it as the process noise covariance matrix.

[0081] The noise-free ideal state transition equation is that the current state prediction value is equal to the preset state transition operator multiplied by the previous state estimate value; where the state is represented by a two-dimensional vector consisting of the pressure change value and the dust density value.

[0082] In one implementation, this embodiment concatenates the pressure change trend vector and the dust density distribution matrix at the feature level to construct the observed state variable at the current moment. This embodiment uses a Kalman filter algorithm to perform iterative state estimation updates on the observed state variable and the preset noise matrix. Specific steps include a time prediction stage and a measurement update stage. In the time prediction stage, the prior state estimate at the current moment is calculated based on the optimal posterior state estimate from the previous moment and a preset state transition operator, and the current prior error covariance is calculated using the process noise covariance matrix. In the measurement update stage, the Kalman gain is calculated based on the prior error covariance and the measurement noise covariance matrix. Then, the difference between the observed state variable at the current moment and the prior state estimate is calculated, and the Kalman gain is used to weight and compensate for this difference in the prior state estimate to obtain the posterior state estimate at the current moment, while simultaneously updating the posterior error covariance. This embodiment extracts the posterior state estimate after iterative convergence and performs inverse matrix reconstruction according to the original concatenated feature dimensions to obtain a denoised smooth estimation matrix.

[0083] It is worth noting that this embodiment uses a system identification method to objectively construct the preset state transition operator. Specifically, this embodiment collects pressure change data and dust density data of the air cannon cluster over more than ten consecutive clearing cycles under historical normal operating conditions, forming a state evolution time series. This series is defined as a set of feature vectors discretely distributed over time. The first 80% of this series is used as training data, and the last 20% is used as validation data. This embodiment assumes that the system state evolution satisfies the first-order linear time-invariant property and establishes a state transition matrix equation containing undetermined coefficients. Subsequently, this embodiment uses a least squares fitting algorithm to iteratively solve the state evolution sequence, calculating the coefficient matrix that minimizes the sum of squared residuals between historical state observations and theoretical predictions, and rigorously calibrating it as the preset state transition operator. This method ensures the objectivity and accuracy of the state prediction stage through mathematical fitting of the inherent evolution law of the physical system.

[0084] In one implementation, this embodiment separates and extracts the corresponding smoothed pressure vector from the smoothed estimation matrix by dimension index. This embodiment calculates the sum of squares of the differences between corresponding elements of the smoothed pressure vector and the original pressure change trend vector, and performs a square root operation to obtain the absolute error distance; simultaneously, it calculates the square root of the sum of squares of each element of the original pressure change trend vector to obtain the original vector magnitude; this embodiment divides the absolute error distance by the original vector magnitude, and determines the resulting dimensionless relative ratio as the error reduction rate.

[0085] It is worth noting that, regarding the method for determining the preset error convergence threshold, this embodiment extracts one thousand Kalman filter operation logs from the historical operation database that successfully eliminated interference noise. Following the same Euclidean distance percentage calculation method described above, one thousand historical error reduction rate samples are extracted. This embodiment arranges these samples in ascending order of value to construct a cumulative probability distribution histogram, extracts the error reduction rate value corresponding to the 80th percentile in this histogram, and objectively labels it as the preset error convergence threshold.

[0086] In one implementation, this embodiment uses comparator logic to determine whether the currently calculated error reduction rate is greater than the preset error convergence threshold. If it is determined to be greater, this embodiment uses principal component analysis (PCA) to perform dimensionality reduction extraction on the smoothed estimation matrix. The eigenvalues ​​and eigenvectors of the covariance matrix of the smoothed estimation matrix are calculated. All eigenvalues ​​are arranged in descending order, and the eigenvectors corresponding to the top few principal eigenvalues ​​with a cumulative variance contribution rate greater than 90% are extracted to form a projection matrix. The smoothed estimation matrix is ​​multiplied by this projection matrix to remove redundant linearly correlated dimensional features, and the low-rank matrix structure of the dimensionality-reduced output is determined as the final situation model.

[0087] For example, assuming the extracted pressure change trend vector contains numerical values, such as [10.5, -2.1], its original vector magnitude is calculated to be approximately 10.7. This vector and the dust density distribution matrix are input into a Kalman filter model, and prior prediction and Kalman gain correction are performed using a measurement noise matrix calibrated based on historical silent data and a process noise matrix calibrated based on residual covariance. After iteration, the extracted smoothed pressure vector is [10.0, -1.8]. The absolute error distance of the difference vector [0.5, -0.3] is calculated to be approximately 0.58. Dividing 0.58 by 10.7 yields an error reduction rate of 0.054. The preset error convergence threshold, determined by historical percentile statistics, is 0.045. Since 0.054 is greater than 0.045, the system performs principal component analysis dimensionality reduction on the smoothed estimation matrix, removing marginal feature vectors with a variance contribution rate of less than 10%, and outputting the dimensionality-reduced situation model.

[0088] In step S15, the situation model and the quality index table are weighted to obtain an evaluation score vector, and the evaluation score vector is normalized using the Softmax function to obtain the urgency score of each node. The urgency scores are then sorted in descending order to obtain a priority sequence, including:

[0089] Extract the smoothed pressure trend value and smoothed density mean of each node from the situation model, and extract the corresponding link score from the quality index table.

[0090] Obtain the preset trend weight, preset density weight, and preset link weight;

[0091] For each of the adjacent nodes, the smoothed pressure trend value is weighted according to the trend weight, the smoothed density mean is weighted according to the density weight, and the link score is weighted according to the link weight. The weighted results of the three are then summed to obtain the single evaluation score of the corresponding node.

[0092] Traverse all nodes to extract the single evaluation score and construct an evaluation score vector;

[0093] The evaluation score vector is normalized using the Softmax function to obtain the urgency score for each node, and the urgency scores of all nodes are sorted in descending order to obtain the priority sequence.

[0094] In one implementation, this embodiment uses the device identifier of the network node as the index key to extract the smoothed pressure trend value and the smoothed density mean of each node from the situation model output in step S14, and performs key-value matching from the aforementioned generated quality index table to extract the corresponding link score, that is, to take the average of the scores of all available links of the node.

[0095] It should be noted that, regarding the determination process of the preset trend weight, the preset density weight, and the preset link weight, this embodiment uses the entropy weight method based on historical real operating data for objective calibration. This embodiment extracts historical environmental characteristic parameters from the central control database of the air cannon cluster, based on actual material blockage events that occurred within the past twelve months. It extracts smoothed pressure trend values, smoothed density mean values, and link scores to construct a historical three-dimensional dataset. This embodiment uses the minimax-maximum-range normalization method to normalize the values ​​of each dimension in this historical three-dimensional dataset, mapping the values ​​to a value range of zero to one. Subsequently, this embodiment calculates the proportion of each dimension's data in the historical sum of that dimension, and calculates the information entropy of each dimension's feature based on this proportion sequence. This embodiment uses the value one minus the information entropy to obtain the difference coefficient of each dimension. Finally, this embodiment divides the difference coefficient of each dimension by the sum of all dimension difference coefficients, and strictly sets the calculated three dimensionless proportion values ​​as the preset trend weight, the preset density weight, and the preset link weight.

[0096] In one implementation, for each of the adjacent nodes, this embodiment multiplies the smoothed pressure trend value by the preset trend weight to obtain a first weight value, multiplies the smoothed density mean by the preset density weight to obtain a second weight value, and multiplies the link score by the preset link weight to obtain a third weight value. This embodiment then sums the first weight value, the second weight value, and the third weight value to calculate the single evaluation score for the corresponding node. Alternatively, this embodiment traverses all adjacent nodes according to the sequential numbering of the network topology nodes, extracts the calculated single evaluation score for each node, and concatenates these scores into a one-dimensional array according to spatial index order to construct an evaluation score vector.

[0097] It is worth noting that this embodiment uses the Softmax function to normalize the evaluation score vector to obtain the urgency score corresponding to each node. Specifically, the calculation logic is as follows: This embodiment extracts each individual evaluation score from the evaluation score vector and calculates the exponential function value with the natural constant as the base and the individual evaluation score as the exponent; this embodiment sums the exponential function values ​​corresponding to all nodes to obtain the global exponential sum; subsequently, this embodiment divides the exponential function value corresponding to the individual evaluation score by the global exponential sum, and determines the floating-point probability value between zero and one, where the sum of all nodes is strictly equal to one, as the urgency score corresponding to each node. This embodiment performs a quicksort algorithm to sort the urgency scores of all nodes in descending order based on their numerical values, extracts the sorted node device identifier sequence, and outputs it as a priority sequence.

[0098] For example, suppose that at a certain moment, local network features containing nodes A, B, and C are extracted. Node A has a smoothed pressure trend value of 8.0, a smoothed density mean of 6.0, and a link score of 2.0. The trend weight, density weight, and link weight obtained by the historical data entropy weight method are 0.5, 0.3, and 0.2, respectively. In this embodiment, the first weight value of node A is calculated to be 4.0, the second weight value to be 1.8, and the third weight value to be 0.4. The sum of these three values ​​gives a single evaluation score of 6.2. Similarly, the single evaluation score of node B is calculated to be 5.8, and the single evaluation score of node C is 6.0. In this embodiment, the three values ​​are extracted to construct an evaluation score vector [6.2, 5.8, 6.0]. In this embodiment, the exponential operation logic of the Softmax function is applied to calculate and normalize the urgency score of node A to be approximately 0.40, the urgency score of node B to be approximately 0.27, and the urgency score of node C to be approximately 0.33. In this embodiment, nodes are sorted in descending order according to their urgency scores, and the final output is a priority sequence consisting of nodes A, C, and B arranged in order.

[0099] In step S16, the target device is determined according to the priority sequence, and the parameters of the target device are adjusted in conjunction with a preset unblocking instruction set to obtain the execution result, including:

[0100] The matching similarity is obtained by comparing the local device identifier with the node identifier corresponding to the highest priority in the priority sequence using a hash matching algorithm.

[0101] If the matching similarity meets the preset consistency condition, then the device corresponding to the local device identifier is determined as the target device;

[0102] The initial vibration frequency and initial airflow velocity are extracted by parsing the preset blockage clearing command set;

[0103] The deviation between the current state value in the situation model and the preset blockage clearing target value is calculated, and the basic control quantity is obtained by using the PID control algorithm to calculate the deviation.

[0104] The basic control quantity is input into a preset decoupling mapping matrix for parameter decomposition to obtain the frequency adjustment coefficient and the speed adjustment coefficient respectively.

[0105] The initial vibration frequency is superimposed and corrected according to the frequency adjustment coefficient, and the initial airflow velocity is superimposed and corrected according to the velocity adjustment coefficient. The correction parameters are combined to obtain the correction parameters, and the output power is calculated based on the conversion of the correction parameters.

[0106] The physical module inside the target device is activated based on the output power to obtain the decrease in pressure and density, and the decrease in pressure and density is determined as the execution result.

[0107] In one implementation, this embodiment extracts the media access control address or factory-fixed serial number of the local device as the local device identifier, and extracts the identifier of the node ranked first in the priority sequence. This embodiment uses a hash matching algorithm to perform hash calculations on the local device identifier and the node identifier, generating two sets of fixed-length hash digests. This embodiment performs a bitwise XOR operation on the two sets of hash digests, counts the absolute number of binary bits with a value of zero in the operation result, and divides this absolute number by the total number of bits in the digest to obtain the matching similarity. It should be noted that, regarding the preset consistency condition, this embodiment strictly sets it to the matching similarity being equal to a value of one, that is, the hash digest of the local device identifier and the highest priority node identifier in the priority sequence are completely consistent. If this consistency condition is met, it indicates that the current device is the node with the highest urgency for clearing congestion in the entire network, and this embodiment then identifies it as the target device to perform subsequent congestion clearing actions.

[0108] In one implementation, this embodiment parses a preset blockage clearing instruction set to extract the initial vibration frequency and initial airflow velocity. Regarding the determination process of the preset blockage clearing instruction set, this embodiment collects equipment control logs from historical physical blockage clearing operations, extracts the records with the top 10% clearing efficiency, maps the corresponding equipment control parameters to blockage status classification tags using key-value pairs, and constructs and stores them as an instruction set configuration file in Extensible Markup Language (XML) format. This embodiment parses the data block corresponding to the current warning status in this configuration file to extract the preset initial vibration frequency and initial airflow velocity.

[0109] It is worth noting that in this embodiment, the smoothed pressure trend value in the situation model is extracted as the current state value, and the current state value is subtracted from the preset clearing target value to obtain the deviation. Regarding the preset clearing target value, this embodiment extracts the natural pressure value sequence of the air cannon cluster under historical no-load and dust-free conditions, calculates the arithmetic mean of this value sequence, and objectively calibrates it as the preset clearing target value. Subsequently, this embodiment uses a PID control algorithm to calculate the deviation to obtain the basic control quantity. Specifically, this embodiment calculates the deviation at the current moment as the proportional term, the discrete sum of historical deviations as the integral term, and the difference between the current deviation and the deviation of the previous sampling period as the differential term; these three are multiplied by the proportional gain, integral gain, and differential gain obtained in advance through the Ziegler-Nichols step response method, and the products of these three terms are summed to obtain a single basic control quantity.

[0110] It should be noted that in this embodiment, the basic control quantity is input into a preset decoupling mapping matrix for parameter decomposition to obtain the frequency adjustment coefficient and the speed adjustment coefficient. Regarding the method for determining the preset decoupling mapping matrix, this embodiment pre-applies step excitation signals to the vibration frequency and airflow velocity of the air cannon on an independent test bench, collects the steady-state cross-response output of the actuator to the other parameter, and constructs a second-order static gain matrix. This embodiment performs matrix inversion on this static gain matrix and strictly sets the calculated inverse matrix as the preset decoupling mapping matrix to eliminate the mechanical cross-coupling interference between the two physical quantities, frequency and speed, during actual execution. In this embodiment, the basic control quantity is treated as a scalar and multiplied by the preset decoupling mapping matrix to decompose and output independent frequency adjustment coefficients and speed adjustment coefficients.

[0111] Specifically, the decoupling mapping matrix is ​​a two-dimensional column vector, with its two elements corresponding to the conversion coefficients of the frequency channel and the speed channel, respectively. The basic control quantity is used as a scalar and multiplied by the two elements of the column vector to obtain the frequency adjustment coefficient and the speed adjustment coefficient. The column vector is used to form a static gain matrix by applying step excitations to the frequency and speed on the test bench and recording the cross-influence coefficients of the two on the unblocking effect. The first column of the pseudo-inverse matrix of this matrix is ​​then used as the decoupling mapping matrix.

[0112] In one implementation, this embodiment performs superposition correction by adding the frequency adjustment coefficient to the initial vibration frequency and the velocity adjustment coefficient to the initial airflow velocity, and then combines the correction results to obtain the correction parameter. This embodiment calculates the output power based on the correction parameter. The squares of the corrected vibration frequency and the corrected airflow velocity are extracted, multiplied, and then multiplied by the inherent mechanical torque constant of the device's motor actuator to obtain the required electrical signal output power.

[0113] In one implementation, this embodiment converts the output power into a pulse-width modulated electrical signal to activate the physical module (including a solenoid valve and a high-pressure nozzle) inside the target device to perform a blockage-clearing operation. In this embodiment, within a fixed observation window after the physical module's action is completed, sensor monitoring values ​​before and after the action are read. The initial pressure value before the action is subtracted from the current pressure value after the action to obtain the pressure drop value, and the initial density value before the action is subtracted from the current density value after the action to obtain the density drop value. This embodiment constructs the pressure drop value and the density drop value into a data pair, which is determined as the final execution result.

[0114] For example, assuming the hash digest of the local device and the highest priority node identifier are completely identical through bitwise XOR comparison, the local device is identified as the target device. Parsing the preset clearing instruction set, the initial vibration frequency is extracted to be 5 Hz and the initial airflow velocity is 10 m / s. The current state value in the situation model is 18 kPa, and the preset clearing target value determined by historical no-load data is 2 kPa. The system calculates the deviation to be 16 kPa. The PID control algorithm with its calibrated input parameters is multiplied and added to obtain a basic control quantity of 8.0. The value 8.0 is multiplied by a preset decoupling mapping matrix composed of the steady-state gain inverse matrix, and the matrix operation decomposes it into a frequency adjustment coefficient of 1.5 and a velocity adjustment coefficient of 3.0. After superposition correction, a corrected parameter combination is obtained, including a frequency of 6.5 Hz and a velocity of 13.0 m / s. The output power is calculated to be 550 watts based on the mechanical constant equation. After the system activates the physical module to spray at this power, the sensor difference calculation shows that the pressure drop is 14 kPa and the density drop is 6 mg / m³. The system outputs the set [14,6] as the execution result.

[0115] In step S17, a threshold compensation amount is calculated based on the execution result, and the preset initial threshold is updated based on the threshold compensation amount to obtain the task completion status.

[0116] The threshold compensation amount is obtained by compensation calculation based on the execution result, including:

[0117] Extract the pressure decrease and density decrease values ​​from the execution results;

[0118] The pressure improvement rate is obtained by dividing the pressure drop value by the initial pressure value, and the density improvement rate is obtained by dividing the density drop value by the initial density value.

[0119] The pressure improvement rate and the density improvement rate are used to construct an effect evaluation vector;

[0120] The effect evaluation vector is input into a preset threshold compensation mapping matrix for multiplication and addition operations, and the corresponding threshold compensation amount is output.

[0121] The process of updating the preset initial threshold based on the threshold compensation amount to obtain the task completion status includes:

[0122] The updated threshold is obtained by adding the preset initial threshold to the threshold compensation amount;

[0123] The update threshold is calculated using an information digest algorithm to obtain the digest data;

[0124] Based on the Paxos consensus protocol, the digest data is broadcast as a threshold update proposal to neighboring nodes in the set of available links;

[0125] If a confirmation response is received from the majority of nodes, consensus is determined to have been reached and a global synchronization identifier is generated, which is then used to identify the task as complete.

[0126] In one implementation, this embodiment extracts the pressure drop and density drop values ​​from the execution results generated in the aforementioned steps, and obtains the initial pressure and initial density values ​​collected by sensors before the unblocking action. This embodiment uses the pressure drop value as the dividend and the initial pressure value as the divisor to perform a division operation to obtain the pressure improvement rate; similarly, this embodiment divides the density drop value by the initial density value to obtain the density improvement rate. This embodiment arranges the pressure improvement rate and the density improvement rate in a fixed dimensional order to construct a two-dimensional effect evaluation vector.

[0127] It should be noted that, regarding the method for determining the preset threshold compensation mapping matrix, this embodiment uses a multiple linear regression algorithm for objective construction. This embodiment extracts one thousand successful physical clearing records from the historical equipment maintenance logs of the central control system, extracting the historical pressure improvement rate, historical density improvement rate, and the ideal threshold correction amount (manually calibrated offline) corresponding to each record to construct a training sample set. This embodiment uses the historical pressure improvement rate and historical density improvement rate as independent variables, and the ideal threshold correction amount as the dependent variable, employing the least squares method for multiple linear regression fitting to solve for the regression coefficients corresponding to the independent variables. This embodiment extracts the two calculated regression coefficients, arranges them according to the dimensional order corresponding to the effect evaluation vector, and objectively sets them as the preset threshold compensation mapping matrix.

[0128] In one implementation, this embodiment inputs the effect evaluation vector into the preset threshold compensation mapping matrix for multiplication and addition operations. Specifically, this embodiment extracts each element of the effect evaluation vector, multiplies it by the corresponding regression coefficient in the matrix, and sums all the product results to output a scalar value as the corresponding threshold compensation amount.

[0129] In one implementation, this embodiment extracts the preset initial threshold determined in step S11, adds it to the calculated threshold compensation amount, and completes the numerical superposition to obtain the updated threshold.

[0130] It should be noted that this embodiment uses an information digest algorithm to calculate the digest data from the update threshold. Specifically, this embodiment uses the SHA-256 algorithm to convert the floating-point value of the update threshold into a standard binary string and input it into the hash function module. After multiple rounds of nonlinear transformations and shift operations, a fixed-length 256-bit hash feature string is output. This embodiment determines this hash feature string as the digest data.

[0131] It is worth noting that, in this embodiment, based on the Paxos consensus protocol, the digest data is broadcast as a threshold update proposal to neighboring nodes in the available link set. This embodiment uses the local node as the proposer, assigning the digest data a globally incrementing proposal number, encapsulating it as a preparation request to be sent to all recipient nodes in the available link set; after receiving responses from more than half of the recipient nodes, this embodiment sends an acceptance request containing the digest data. This embodiment continuously counts the absolute number of nodes returning confirmation responses and determines whether this absolute number is greater than half of the total number of nodes in the available link set. If it is greater than this proportion, this embodiment determines that consensus has been reached. This embodiment then generates a global synchronization identifier containing the current absolute timestamp and the local device identifier, outputs this global synchronization identifier, and determines the task completion status.

[0132] For example, assume the pressure drop in the execution result is 10 kPa and the density drop is 5 mg / m³; the initial pressure before unclogging is 50 kPa and the initial density is 25 mg / m³. This embodiment calculates a pressure improvement rate of 0.2 (10 divided by 50) and a density improvement rate of 0.2 (5 divided by 25). This embodiment constructs a two-dimensional effect evaluation vector. The preset threshold compensation mapping matrix, calibrated by multiple linear regression of historical data, contains coefficients -2.0 and -1.5. This embodiment performs multiplication and addition operations: multiplying 0.2 by -2.0 yields -0.4, and multiplying 0.2 by -1.5 yields -0.3; adding the two yields a threshold compensation of -0.7. The original preset initial threshold is 7.5. This embodiment adds -0.7 to 7.5 to obtain an updated threshold of 6.8. The system uses the SHA-256 algorithm to hash the value 6.8 to generate a digest, and broadcasts this digest as Paxos proposal number 101 to eight neighboring nodes in the available link set. The system subsequently receives confirmation responses from five nodes. Since 5 is greater than half of 8, the system determines that consensus has been reached, generates a global synchronization flag containing the current timestamp and device identifier, and outputs the task completion status.

[0133] In summary, this invention generates status information by fusing local equipment dust density and pressure characteristics, and performs multi-hop broadcasting and data sharing based on dynamically detected link scores. This breaks the information silo dilemma under centralized control, achieving real-time reconstruction of the communication topology and robust synchronization of underlying status data. By introducing a Kalman filter model to denoise temporal features, and combining link quality and global situational awareness with multi-factor weighting and Softmax normalization, it achieves accurate judgment of the urgency of congestion clearing and dynamic priority reordering in multi-concurrent congestion scenarios, eliminating blind firing caused by disordered execution. Furthermore, by using a PID control algorithm to decouple and output physical adjustment coefficients to guide target equipment execution in a closed loop, and dynamically adaptively calibrating event trigger thresholds based on the physical improvement rate using the Paxos consensus protocol, it achieves customized attacks and cluster self-correction for congestion situations. This effectively solves the problems of low congestion clearing efficiency and easy system stagnation, maximizing the long-term operational stability of industrial storage and transportation systems.

[0134] Reference Figure 2 The second embodiment of the present invention provides an IoT-based collaborative control system for air cannon clusters, comprising:

[0135] The state generation module is used to acquire dust density data and pressure change data, perform weighted calculations to obtain a fusion index, and perform feature fusion based on the fusion index and a preset initial threshold to generate initial state information.

[0136] The link evaluation module is used to send probe packets to adjacent nodes in the preset initial routing table according to the initial state information to obtain the response time and packet loss rate, calculate the reciprocal of the response time and the packet loss rate respectively, and calculate the link score by weighting the obtained reciprocals, and generate an available link set and a quality index table based on the link score.

[0137] The data synchronization module is used to combine the available link set to perform multi-hop broadcasting of the initial state information to obtain monitoring values, and to perform fusion processing on the monitoring values ​​to obtain a shared dataset;

[0138] The situation denoising module is used to denoise the shared dataset and the preset noise matrix using Kalman filtering to obtain the situation model;

[0139] The priority ranking module is used to perform weighted calculations on the situation model and the quality index table to obtain an evaluation score vector, and to normalize the evaluation score vector using the Softmax function to obtain the urgency score of each node. The urgency scores are then sorted in descending order to obtain a priority sequence.

[0140] The congestion clearing execution module is used to determine the target device according to the priority sequence, and adjust the parameters of the target device in combination with the preset congestion clearing instruction set to obtain the execution result;

[0141] The closed-loop consensus module is used to calculate the threshold compensation amount based on the execution result, and update the preset initial threshold based on the threshold compensation amount to obtain the task completion status.

[0142] It should be noted that the IoT-based air cannon cluster collaborative control system provided in this embodiment of the invention is used to execute all process steps of the IoT-based air cannon cluster collaborative control method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0143] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0144] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A collaborative control method for an air cannon cluster based on the Internet of Things, characterized in that, include: Dust density data and pressure change data are acquired and weighted to obtain a fusion index. Based on the fusion index, feature fusion is performed in combination with a preset initial threshold to generate initial state information. Based on the initial state information, probe packets are sent to adjacent nodes in the preset initial routing table to obtain response time and packet loss rate. The reciprocals of the response time and packet loss rate are calculated respectively, and the obtained reciprocals are weighted to obtain the link score. Based on the link score, an available link set and a quality index table are generated. The initial state information is broadcast in multiple hops using the available link set to obtain monitoring values, and the monitoring values ​​are then fused to obtain a shared dataset. The shared dataset and the preset noise matrix are denoised using Kalman filtering to obtain the situation model; The situation model and the quality index table are weighted to obtain an evaluation score vector. The evaluation score vector is then normalized using the Softmax function to obtain the urgency score of each node. The urgency scores are then sorted in descending order to obtain a priority sequence. The target device is determined according to the priority sequence, and the parameters of the target device are adjusted in combination with the preset clearing instruction set to obtain the execution result; The threshold compensation amount is obtained by performing compensation calculation based on the execution result, and the preset initial threshold is updated based on the threshold compensation amount to obtain the task completion status.

2. The IoT-based collaborative control method for air cannon clusters according to claim 1, characterized in that, The process involves acquiring dust density data and pressure change data, performing weighted calculations to obtain a fusion index, and then performing feature fusion based on the fusion index and a preset initial threshold to generate initial state information, including: Data on dust density and pressure changes of the air cannon cluster are acquired through sensors. The dust density data is multiplied by a preset dust density weight to obtain a first value, and the pressure change data is multiplied by a preset pressure change rate weight to obtain a second value. The first value and the second value are added together to obtain a fusion index, and it is determined whether the fusion index is greater than the preset initial threshold. If the fusion index is greater than the preset initial threshold, an event trigger flag is generated, and the event trigger flag, the dust density data, and the pressure change data are spliced ​​and recombined to obtain the initial state information.

3. The IoT-based collaborative control method for air cannon clusters according to claim 1, characterized in that, The process involves sending probe packets to neighboring nodes in a preset initial routing table based on the initial state information to obtain response time and packet loss rate. The reciprocals of the response time and packet loss rate are calculated, and the resulting reciprocals are weighted to obtain a link score. Based on the link score, a set of available links and a quality index table are generated, including: Based on the initial state information, link probe packets are sent to adjacent nodes in the preset initial routing table, and timestamps are recorded to obtain response time and packet loss rate; Obtain a preset response time weight and a preset packet loss rate weight, and calculate the reciprocal of the response time and the packet loss rate respectively to obtain the reciprocal of the response time and the reciprocal of the packet loss rate; The third value is obtained by multiplying the reciprocal of the response time by the response time weight, and the fourth value is obtained by multiplying the reciprocal of the packet loss rate by the packet loss rate weight. The link score is obtained by adding the third value and the fourth value together. The adjacent nodes whose link scores are greater than a preset link quality threshold are extracted and connected topologically to obtain an available link set. The link scores of all nodes in the available link set are extracted and merged to obtain a quality index table.

4. The IoT-based collaborative control method for air cannon clusters according to claim 1, characterized in that, The process of combining the available link set to perform multi-hop broadcasting of the initial state information to obtain monitoring values, and then fusing the monitoring values ​​to obtain a shared dataset, includes: The initial state information is calculated using the MD5 hash algorithm to obtain a checksum, and the initial state information and the checksum are serialized and encapsulated to obtain a transmission data packet; Based on the Gossip synchronization protocol, the data packets are broadcast over the network through the set of available links. If the transmission delay is less than the preset delay threshold, the data packets returned by each node in the available link set are received, parsed, and the monitoring values ​​are obtained. A weighted average algorithm is used to calculate the shared average value of the monitored values, and the shared average value is bound to the timestamp to obtain a shared dataset.

5. The IoT-based collaborative control method for air cannon clusters according to claim 1, characterized in that, The process of denoising the shared dataset and a preset noise matrix using Kalman filtering to obtain the situation model includes: Extract the pressure change trend vector and dust density distribution matrix of adjacent timestamps from the shared dataset; The pressure change trend vector, the dust density distribution matrix, and the preset noise matrix are subjected to state estimation iterative update operation using the Kalman filter algorithm to obtain a smooth estimation matrix; Extract the smoothed pressure vector from the smoothed estimation matrix, and calculate the error reduction rate of the smoothed pressure vector compared to the pressure change trend vector; If the error reduction rate is greater than the preset error convergence threshold, the smooth estimation matrix is ​​reduced in dimension to extract the situation model.

6. The IoT-based collaborative control method for air cannon clusters according to claim 1, characterized in that, The evaluation score vector is obtained by weighting the situation model and the quality index table, and the evaluation score vector is normalized using the Softmax function to obtain the urgency score of each node. The urgency scores are then sorted in descending order to obtain a priority sequence, including: Extract the smoothed pressure trend value and smoothed density mean of each node from the situation model, and extract the corresponding link score from the quality index table. Obtain the preset trend weight, preset density weight, and preset link weight; For each of the adjacent nodes, the smoothed pressure trend value is weighted according to the trend weight, the smoothed density mean is weighted according to the density weight, and the link score is weighted according to the link weight. The weighted results of the three are then summed to obtain the single evaluation score of the corresponding node. Traverse all nodes to extract the single evaluation score and construct an evaluation score vector; The evaluation score vector is normalized using the Softmax function to obtain the urgency score for each node, and the urgency scores of all nodes are sorted in descending order to obtain the priority sequence.

7. The IoT-based collaborative control method for air cannon clusters according to claim 1, characterized in that, The step of determining the target device according to the priority sequence and adjusting the parameters of the target device in conjunction with a preset clearing instruction set to obtain the execution result includes: The matching similarity is obtained by comparing the local device identifier with the node identifier corresponding to the highest priority in the priority sequence using a hash matching algorithm. If the matching similarity meets the preset consistency condition, then the device corresponding to the local device identifier is determined as the target device; The initial vibration frequency and initial airflow velocity are extracted by parsing the preset blockage clearing command set; The deviation between the current state value in the situation model and the preset blockage clearing target value is calculated, and the basic control quantity is obtained by using the PID control algorithm to calculate the deviation. The basic control quantity is input into a preset decoupling mapping matrix for parameter decomposition to obtain the frequency adjustment coefficient and the speed adjustment coefficient respectively. The initial vibration frequency is superimposed and corrected according to the frequency adjustment coefficient, and the initial airflow velocity is superimposed and corrected according to the velocity adjustment coefficient. The correction parameters are combined to obtain the correction parameters, and the output power is calculated based on the conversion of the correction parameters. The physical module inside the target device is activated based on the output power to obtain the decrease in pressure and density, and the decrease in pressure and density is determined as the execution result.

8. The IoT-based collaborative control method for air cannon clusters according to claim 1, characterized in that, The step of calculating the threshold compensation amount based on the execution result includes: Extract the pressure decrease and density decrease values ​​from the execution results; The pressure improvement rate is obtained by dividing the pressure drop value by the initial pressure value, and the density improvement rate is obtained by dividing the density drop value by the initial density value. The pressure improvement rate and the density improvement rate are used to construct an effect evaluation vector; The effect evaluation vector is input into a preset threshold compensation mapping matrix for multiplication and addition operations, and the corresponding threshold compensation amount is output.

9. The IoT-based collaborative control method for air cannon clusters according to claim 1, characterized in that, The step of updating the preset initial threshold according to the threshold compensation amount to obtain the task completion status includes: The updated threshold is obtained by adding the preset initial threshold to the threshold compensation amount; The update threshold is calculated using an information digest algorithm to obtain the digest data; Based on the Paxos consensus protocol, the digest data is broadcast as a threshold update proposal to neighboring nodes in the set of available links; If a confirmation response is received from the majority of nodes, consensus is determined to have been reached and a global synchronization identifier is generated, which is then used to identify the task as complete.

10. A collaborative control system for an air cannon cluster based on the Internet of Things, characterized in that, include: The state generation module is used to acquire dust density data and pressure change data, perform weighted calculations to obtain a fusion index, and perform feature fusion based on the fusion index and a preset initial threshold to generate initial state information. The link evaluation module is used to send probe packets to adjacent nodes in the preset initial routing table according to the initial state information to obtain the response time and packet loss rate, calculate the reciprocal of the response time and the packet loss rate respectively, and calculate the link score by weighting the obtained reciprocals, and generate an available link set and a quality index table based on the link score. The data synchronization module is used to combine the available link set to perform multi-hop broadcasting of the initial state information to obtain monitoring values, and to perform fusion processing on the monitoring values ​​to obtain a shared dataset; The situation denoising module is used to denoise the shared dataset and the preset noise matrix using Kalman filtering to obtain the situation model; The priority ranking module is used to perform weighted calculations on the situation model and the quality index table to obtain an evaluation score vector, and to normalize the evaluation score vector using the Softmax function to obtain the urgency score of each node. The urgency scores are then sorted in descending order to obtain a priority sequence. The congestion clearing execution module is used to determine the target device according to the priority sequence, and adjust the parameters of the target device in combination with the preset congestion clearing instruction set to obtain the execution result; The closed-loop consensus module is used to calculate the threshold compensation amount based on the execution result, and update the preset initial threshold based on the threshold compensation amount to obtain the task completion status.

Images