Intelligent control method and system for clean bench based on Internet of Things

Abstract

The invention provides a clean bench intelligent control method and system based on the Internet of Things. The method comprises the following steps: acquiring operation data of each clean workbench in a current operation scene, and extracting time sequence data of airflow velocity and airflow direction contained in the operation data; performing data preprocessing on the time sequence data by adopting a preset standardized format to obtain a preprocessed data set; extracting feature vectors of airflow velocity and airflow direction of each clean bench in the pre-processing data set, and respectively calculating Euclidean distances among different feature vectors to determine airflow interaction strength among the clean benches; according to the similarity measurement and the airflow interaction intensity, pollution risk distribution in the current operation scene is determined, the clean bench layout in the current operation scene is optimized, and the optimized clean bench layout is obtained; and scheduling the clean workbenches according to the optimized layout of the clean workbenches. According to the invention, the cross contamination risk of the clean bench can be reduced.

Description

Technical Field

[0001] This application relates to industrial data analysis and edge computing technologies in the field of the Internet of Things, and specifically, to an intelligent control method and system for a clean workbench based on the Internet of Things. Background Art

[0002] With the advent of Industry 4.0, clean benches, as critical equipment in fields like biomedicine and precision manufacturing, face a significant impact on product quality and safety. In this context, clean bench control technology based on the Internet of Things (IoT), such as IoT, has become a research hotspot, aiming to optimize equipment operation through data-driven approaches.

[0003] Clean benches are critical equipment in fields like biomedicine and precision manufacturing, and their operational stability is directly linked to product quality and safety. With the growing demand for intelligent clean benches, IoT-based clean bench control technology has become a research hotspot, aiming to optimize equipment operation through data-driven optimization. However, existing methods mostly rely on centralized data processing, making it difficult to respond to complex environmental changes in real time. They also lack consideration for dynamic interactions between devices, making it difficult to effectively control airflow interference and cross-contamination risks. Furthermore, extensive energy management prevents users from flexibly adjusting equipment layouts based on actual needs, limiting system adaptability. In this context, real-time processing of clean bench operational data and dynamic adjustment of equipment layouts have become key challenges. In an IoT environment, the massive amount of operational data generated by clean benches requires rapid analysis to capture airflow changes, but traditional centralized computing methods, due to high latency, struggle to meet real-time requirements. While the introduction of edge computing technology can alleviate this problem, it further highlights the complexity of the topological relationships between devices. Airflow interference and contamination risks between clean benches evolve dynamically with layout changes. Without an accurate network topology for accurately constructing and analyzing the inter-device topology, it is difficult to predict and optimize airflow distribution. This directly affects the precise control of energy consumption distribution, because the airflow pattern is closely related to the operating power of the equipment, and an unoptimized layout may lead to energy waste.

[0004] Therefore, the present application provides an Internet of Things-based clean bench intelligent control method and system to solve one of the above technical problems. Summary of the Invention

[0005] The purpose of this application is to provide an intelligent control method and system for a clean bench based on the Internet of Things, which can solve at least one of the technical problems mentioned above. The specific solution is as follows: According to the specific implementation of the present application, in a first aspect, the present application provides an intelligent control method for a clean workbench based on the Internet of Things, comprising: For the clean workbench cluster in the current operation scene, the operation data of each clean workbench is obtained, and the time series data of the airflow velocity and airflow direction contained in the operation data is extracted; the time series data is preprocessed in a preset standardized format to obtain a preprocessed data set, and the preprocessed data set contains the preprocessed data corresponding to the identifier of each clean workbench; the characteristic vectors of the airflow velocity and airflow direction of each clean workbench are extracted from the preprocessed data set, and the Euclidean distance between different characteristic vectors is calculated respectively, so as to use the Euclidean distance as the similarity measure of the airflow pattern between different clean workbenches; the airflow interaction intensity between each clean workbench is determined according to the similarity measure; the pollution risk distribution in the current operation scene is determined according to the similarity measure and the airflow interaction intensity, and the clean workbench layout in the current operation scene is optimized according to the pollution risk distribution to obtain an optimized clean workbench layout; according to the optimized clean workbench layout, each clean workbench is scheduled.

[0006] In one embodiment, determining the airflow interaction intensity between each of the clean workbenches according to the similarity measurement includes: performing local clustering on the similarity measurement to obtain a clustering result for each clean workbench; based on the clustering result, calculating the airflow pattern overlap between different clean workbenches; and performing weighted averaging on the airflow pattern overlap to obtain the airflow interaction intensity between each of the clean workbenches.

[0007] In one embodiment, the pollution risk distribution within the current operation scene is determined based on the similarity measurement and the airflow interaction intensity, including: for the target clean workbenches whose airflow interaction intensity is greater than a first intensity threshold, a topological relationship network is constructed based on the airflow interaction intensity between each of the target clean workbenches to obtain a dynamic topological structure between each of the target clean workbenches; according to the dynamic topological structure, an airflow interference coefficient matrix is ​​constructed between each of the target clean workbenches; based on the airflow interference coefficient matrix and the particulate matter concentration data contained in the operation data, the cross-contamination risk probability of particulate matter transfer between each of the target clean workbenches is determined to obtain a cross-contamination risk probability set; based on the cross-contamination risk probability set, a spatial interpolation method is used to generate a spatial distribution of pollution risks between each of the target clean workbenches as the pollution risk distribution within the current operation scene.

[0008] In one embodiment, the cross-contamination risk probability of particle transfer between each of the target clean workbenches is determined based on the airflow interference coefficient matrix and the particle concentration data contained in the operation data to obtain a cross-contamination risk probability set, including: extracting the time series data of the particle concentration data, and preprocessing the time series data of the particle concentration data using a standardized format to obtain a particle concentration data set; for the target matrix elements whose airflow interaction intensity is greater than a second intensity threshold, determining the target clean workbench pairs used to identify the target matrix elements, and extracting the particle concentration data corresponding to each of the clean workbench pairs from the particle concentration data set; simulating the particle propagation path between the clean workbench pairs based on the particle concentration data corresponding to each of the target clean workbench pairs and the airflow interaction intensity of the target matrix elements to obtain particle trajectory data; statistically analyzing the cross-contamination risk probability of particle transfer between the target clean workbench pairs based on the position distribution and time distribution of the particles represented by the particle trajectory data; and summarizing the cross-contamination risk probabilities corresponding to each pair of target clean workbenches to obtain a cross-contamination risk probability set.

[0009] In one embodiment, the clean workbench layout in the current operation scene is optimized according to the pollution risk distribution to obtain an optimized clean workbench layout, including: if the pollution risk probability of any area in the pollution risk distribution is greater than a first probability threshold, then according to the layout of the clean workbench cluster, the spatial coordinates of each clean workbench are encoded and the layout is initialized using a genetic algorithm to obtain a clean workbench layout population; for the encoded clean workbench layout population, in combination with the pollution risk probability and predefined layout constraints, a crossover mutation operation is performed through a genetic algorithm to update the spatial coordinates of each clean workbench while iterating the layout of each clean workbench in the clean workbench layout population according to the updated spatial coordinates; with reference to the updated spatial coordinates of each clean workbench, an updated pollution risk distribution is generated using a spatial interpolation method; with reference to the updated pollution risk distribution, the clean workbench layout with the smallest pollution risk probability and lower than the second probability threshold is screened in the iterated clean workbench layout population as the optimized clean workbench layout.

[0010] In one embodiment, the method further includes: extracting the airflow velocity data, clean workbench operating power data and layout coordinate data contained in the operating data, and generating the airflow velocity matrix, the operating power vector and the spatial coordinate set through standardization processing as the energy consumption distribution of the clean workbench cluster; determining energy consumption optimization parameters for optimizing the energy consumption distribution; if it is determined based on the energy consumption optimization parameters that there is a clean workbench with an energy consumption value greater than an energy consumption threshold, adjusting the operating power of the clean workbench cluster.

[0011] In one embodiment, the energy consumption optimization parameters for optimizing the energy consumption distribution are determined, including: for the energy consumption distribution, using a long short-term memory network to extract time series features of the airflow velocity matrix, the operating power vector and the spatial coordinate set to obtain a sequence of energy consumption prediction values ​​for each clean workbench; if it is determined based on the energy consumption prediction value sequence that there is a clean workbench with an energy consumption value greater than an energy consumption threshold, then according to the spatial coordinate set and the airflow velocity matrix, a linear regression method is used to calculate the adjustment coefficient of the energy consumption distribution to obtain an updated energy consumption distribution parameter; based on the updated energy consumption distribution parameter, combined with the spatial coordinates and pre-collected pollution risk data, the energy consumption distribution parameters and the pollution risk data are fused by a weighted average method to obtain optimized energy consumption optimization parameters.

[0012] In one embodiment, the operating power adjustment of the clean workbench cluster includes: determining an airflow change trend matrix based on the operating data; using a feedback control algorithm to iteratively calculate the airflow change trend matrix and the operating power vector to obtain a power adjustment coefficient sequence; based on the power adjustment coefficient sequence, updating the operating power of the clean workbench cluster by a linear interpolation method to obtain an updated operating power data set; for the updated operating power data set, combined with the airflow velocity data, fusing the energy consumption optimization parameters and the clean workbench operating status by a weighted average method to obtain an optimized operating power data set; according to the optimized operating power data set, adjusting the operating power of the clean workbench cluster.

[0013] In one embodiment, determining the airflow change trend matrix based on the operating data includes: normalizing the airflow velocity data of each clean workbench to obtain normalized airflow velocity data; and using each normalized airflow velocity data as a matrix element to generate an airflow change trend matrix.

[0014] According to the specific implementation of the present application, in a second aspect, the present application provides an Internet of Things-based clean bench intelligent control system, comprising: An acquisition unit is configured to acquire, for a cluster of clean workbenches within a current operation scenario, operating data of each clean workbench, and extract time series data of airflow velocity and airflow direction contained in the operating data; a processing unit is configured to perform data preprocessing on the time series data using a preset standardized format to obtain a preprocessed data set, wherein the preprocessed data set contains preprocessed data corresponding to each clean workbench identifier; the acquisition unit is further configured to extract characteristic vectors of airflow velocity and airflow direction of each clean workbench from the preprocessed data set, and to calculate the Euclidean distance between different characteristic vectors, so as to use the Euclidean distance as a similarity measure of airflow patterns between different clean workbenches; a determination unit is configured to determine the airflow interaction intensity between each clean workbench according to the similarity measure; and to determine the pollution risk distribution within the current operation scenario based on the similarity measure and the airflow interaction intensity, and to optimize the clean workbench layout within the current operation scenario based on the pollution risk distribution to obtain an optimized clean workbench layout; a control unit is configured to schedule each clean workbench according to the optimized clean workbench layout.

[0015] Compared with the prior art, the above-mentioned scheme of the embodiment of the present application has at least the following beneficial effects: The present application provides an intelligent control method for a clean workbench based on the Internet of Things. By obtaining the operating data of the clean workbench cluster and extracting the time series data of the airflow speed and direction, this method first ensures the comprehensiveness and accuracy of the basic data. After preprocessing the data in a standardized format, it can effectively remove noise and outliers, thereby improving the accuracy of subsequent analysis. Based on these preprocessed data sets, the Euclidean distance between the feature vectors is calculated, which not only quantifies the similarity of the airflow patterns between different clean workbenches, but also provides a scientific basis for identifying potential airflow interactions. Furthermore, by determining the intensity of airflow interaction and optimizing the clean workbench layout accordingly, the risk of cross-contamination can be significantly reduced, while improving the cleanliness of the working environment and energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flow chart of an intelligent control method for a clean workbench based on the Internet of Things is shown; Figure 2 A flow chart of a method for determining the airflow interaction intensity between clean workbenches according to similarity measurement is shown; Figure 3 A flow chart of a method for determining the pollution risk distribution within a current operation scene is shown; Figure 4 A flow chart of a method for obtaining a set of cross-contamination risk probabilities is shown; Figure 5A flow chart of a method for obtaining an optimized clean bench layout is shown; Figure 6 A flow chart of a method for adjusting the operating power of a clean bench cluster is shown; Figure 7 A flow chart of a method for determining energy consumption optimization parameters for optimizing energy consumption distribution is shown; Figure 8 A flow chart of a method for adjusting the operating power of a clean bench cluster is shown; Figure 9 A unit block diagram of an IoT-based clean bench intelligent control system according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0017] To make the objectives, technical solutions, and advantages of this application more clear, this application will be further described in detail below with reference to the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0018] The terms used in the examples of this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms "a," "the," and "the" used in the examples of this application and the appended claims are also intended to include plural forms, and unless the context clearly indicates otherwise, "a plurality" generally includes at least two.

[0019] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0020] It should be understood that although the terms first, second, third, etc. may be used to describe in the embodiments of the present application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, without departing from the scope of the embodiments of the present application, the first may also be referred to as the second, and similarly, the second may also be referred to as the first.

[0021] Depending on the context, the words "if" and "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting." Similarly, depending on the context, the phrases "if it is determined" or "if a stated condition or event is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting, such as a stated condition or event" or "in response to detecting, such as a stated condition or event."

[0022] It should also be noted that the terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such product or system. In the absence of further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the product or system comprising the element.

[0023] It should be noted in particular that any symbols and / or numbers in the specification that are not marked in the accompanying drawings are not drawing marks.

[0024] The optional embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0025] The embodiment provided in this application is an embodiment of an intelligent control method for a clean workbench based on the Internet of Things.

[0026] The following combination Figure 1 The embodiments of the present application are described in detail.

[0027] Figure 1 A flow chart of an intelligent control method for a clean workbench based on the Internet of Things is shown. Figure 1 As shown, the following steps are included.

[0028] Step S101 : For a clean bench cluster in a current operation scene, the operation data of each clean bench is acquired, and time series data of airflow velocity and airflow direction included in the operation data are extracted.

[0029] In this application, the operating data obtained for each clean bench may include multiple aspects, such as air flow speed and direction, particle concentration, operating power, spatial coordinates, temperature and humidity, vibration data, noise level, operating time, and filter status.

[0030] Step S102 : preprocessing the time series data using a preset standardized format to obtain a preprocessed data set, wherein the preprocessed data set includes preprocessed data corresponding to each clean bench identifier.

[0031] In step S103 , the eigenvectors of the airflow velocity and airflow direction of each clean workbench are extracted from the preprocessed data set, and the Euclidean distance between different eigenvectors is calculated respectively, so as to use the Euclidean distance as a similarity measure of the airflow patterns between different clean workbenches.

[0032] Step S104 : determining the airflow interaction intensity between the clean workbenches according to the similarity measurement.

[0033] Step S105 , determining the pollution risk distribution in the current operation scene based on the similarity metric and the airflow interaction intensity, and optimizing the clean workbench layout in the current operation scene based on the pollution risk distribution to obtain an optimized clean workbench layout.

[0034] Step S106: scheduling each clean bench according to the optimized clean bench layout.

[0035] In the embodiment of the present application, by obtaining the operating data of the clean bench cluster and extracting the time series data of the airflow speed and direction, this method first ensures the comprehensiveness and accuracy of the basic data. After data preprocessing in a standardized format, noise and outliers can be effectively removed, thereby improving the accuracy of subsequent analysis. Based on these preprocessed data sets, the Euclidean distance between feature vectors is calculated, which not only quantifies the similarity of airflow patterns between different clean benches, but also provides a scientific basis for identifying potential airflow interactions. Furthermore, by determining the intensity of airflow interaction and optimizing the clean bench layout accordingly, the risk of cross-contamination can be significantly reduced, while improving the cleanliness of the working environment and energy efficiency.

[0036] As a specific embodiment, the present application first obtains operational data from sensors installed on each clean bench in a cluster within the current operating scenario. The operational data includes, but is not limited to, parameters such as airflow velocity and direction, particulate matter concentration, operating power, spatial coordinates, temperature and humidity, vibration, and noise levels. Time series data of airflow velocity and direction are then extracted and preprocessed using a preset standardized format to eliminate the impact of data heterogeneity between devices, resulting in a structurally consistent and analyzable preprocessed dataset, with each data entry corresponding to a clean bench identifier.

[0037] Furthermore, the airflow velocity and direction eigenvectors of each clean workbench were extracted from the preprocessed dataset using a sliding window method. The distance between different eigenvectors was calculated using the Euclidean distance formula as a quantitative measure of the similarity of the airflow patterns between the two clean workbenches. Based on this similarity measurement result and the physical location relationship between the clean workbenches, a k-nearest neighbor algorithm or a graph neural network model was used to further derive the airflow interaction intensity matrix between each clean workbench to characterize the possibility of airflow disturbance propagation. Next, the airflow interaction intensity and particulate matter concentration data were combined to calculate the particle transfer path and cross-contamination probability between the clean workbenches using a particle tracking simulation method, and a contamination risk distribution map for the entire work scenario was generated through spatial interpolation. Based on this contamination risk distribution map, a genetic algorithm was used to optimize the spatial layout of the clean workbenches, with the goal of minimizing the contamination probability in high-risk areas and satisfying physical constraints such as minimum spacing, thereby obtaining an optimized clean workbench layout plan.

[0038] On this basis, the optimized layout plan will be sent to the control systems of each clean workbench, and its operating parameters will be adjusted accordingly to achieve scheduling control of the clean workbench cluster, thereby improving the stability and safety of the overall clean environment.

[0039] Figure 2 A flow chart of a method for determining the airflow interaction intensity between clean workbenches according to similarity measurement is shown. Figure 2 As shown, the following steps are included.

[0040] Step S201 : performing local clustering on the similarity measurement to obtain a clustering result for each clean bench.

[0041] Step S202 : calculating the airflow pattern overlap between different clean workbenches based on the clustering result.

[0042] Step S203 : performing weighted averaging processing on the airflow pattern overlap to obtain the airflow interaction intensity between the clean workbenches.

[0043] In the embodiments of the present application, similarity measurement refers to quantifying the similarity of the airflow patterns of different clean workbenches by calculating the Euclidean distance between the airflow velocity and direction feature vectors. The numerical range is generally a continuous value interval (such as 0 to 1 or greater), and the smaller the value, the more similar the airflow patterns of the two clean workbenches are. The clustering result is a classification label obtained after local clustering of each clean workbench based on the similarity measurement. It is used to identify whether a clean workbench belongs to the "similar mode" or "independent mode". It is discrete data, for example, represented by a string or integer (such as the label "Cluster_A" or the number "1" for similar mode, and "Outlier" or "0" for independent mode). The airflow pattern overlap is the degree of overlap in the airflow behavior between clean workbenches, further calculated based on the clustering results, reflecting whether there is a significant mutual influence between their operating states. It is usually represented by a numerical value in the range [0,1]. The higher the value, the more consistent the airflow behavior between the two devices. The airflow interaction intensity is the result of weighted averaging based on the overlap of airflow patterns and factors such as the physical distance between clean workbenches. It is used to characterize the possibility of airflow disturbance propagation between different clean workbenches. It is a comprehensive indicator and is generally expressed in normalized numerical form (such as 0.0 to 1.0). The larger the value, the stronger the airflow interaction between the two, and the higher the risk of interference or contamination propagation.

[0044] In this application's example, local clustering helps identify groups of clean benches with similar airflow patterns when evaluating the intensity of airflow interactions between them. This provides a structured framework for further analysis. By calculating the degree of airflow pattern overlap and performing a weighted average, this not only accurately measures the degree of interaction between clean benches but also provides specific guidance for adjusting equipment layout to reduce unnecessary airflow interference and optimize air quality across the entire work area.

[0045] In some embodiments, for a cluster of clean benches within the current operational scenario, time series data of airflow velocity and direction are first acquired from sensors at each clean bench. This time series data is then preprocessed using a pre-set standardized format to generate a preprocessed dataset corresponding to each clean bench identifier. Based on this preprocessed dataset, a sliding window approach is used to extract feature vectors of airflow velocity and direction for each clean bench. The distance between these feature vectors is then calculated using the Euclidean distance formula, serving as a measure of the similarity of airflow patterns across clean benches. Ensuring that the feature vectors are consistent in dimensionality is crucial, as Euclidean distance is sensitive to each dimension.

[0046] As a feasible embodiment, when obtaining time series data of airflow velocity and airflow direction from the sensors of each clean workbench, it is necessary to ensure the uniformity of data collection. The sensor records the airflow velocity and direction at a fixed frequency, such as once per second, to generate time series data. The raw data may contain noise or missing values, so preprocessing steps including data cleaning and standardization are required. Data cleaning can eliminate outliers, such as data points where the airflow velocity exceeds the reasonable range of 0.3-0.8m / s. Standardization converts the airflow velocity and direction into a unified dimension, such as normalizing the direction angle to the range of 0-1. This preprocessing step ensures the accuracy of subsequent analysis.

[0047] As a specific example, a sliding window with a 10-second step size and a 1-second step size is used to extract feature vectors based on a preprocessed dataset. Within each window, the mean airflow velocity and the standard deviation of the direction are calculated to form a feature vector. For example, within a window, the mean airflow velocity is 0.5 m / s and the standard deviation of the direction is 5°, resulting in the vector [0.5, 5]. This method generates a continuous sequence of feature vectors for each clean bench, reflecting the dynamic changes in the airflow pattern.

[0048] For example, when calculating the similarity measure between feature vectors, the Euclidean distance formula is used. The feature vectors of two clean workbenches in the same time window, such as [0.5, 5] and [0.6, 4], are calculated to have similarities through Euclidean distance. The smaller the distance, the closer the airflow patterns of the two clean workbenches are. For example, if the distances between a clean workbench and multiple other clean workbenches are small, it means that their airflow patterns have commonalities. For example, when using the k-nearest neighbor algorithm and setting the k value to 5 for local clustering, if the average distance between a clean workbench and its neighbors is lower than a threshold, such as 0.1, it is marked as a similar pattern. On the contrary, if it is higher than a threshold, such as 0.3, it is marked as an independent pattern.

[0049] In one embodiment, if the airflow velocity of a clean workbench fluctuates frequently and changes direction greatly, the clustering result shows that it is an independent mode, which indicates that its operating state is unique. In a possible implementation method, the airflow pattern overlap is calculated based on the clustering results, and the overlap reflects the degree of interaction between the airflow patterns between the clean workbenches. For example, if the airflow velocity feature vector sequences of two clean workbenches are highly similar and the overlap reaches 80%, it means that their airflow patterns may affect each other. When the weighted average method is used to process the overlap, weights can be assigned according to the physical distance between the clean workbenches. The closer the distance, the higher the weight, such as 0.6, and the farther the distance, the lower the weight, such as 0.2. After weighted averaging, the airflow interaction intensity is obtained. For example, the interaction intensity between a clean workbench and an adjacent clean workbench is 0.75, indicating that its airflow pattern is greatly affected by the surrounding clean workbenches.

[0050] It's important to note that analyzing airflow interaction intensity can help optimize clean bench layouts. For example, if multiple clean benches have high interaction intensity, their positions can be adjusted to reduce airflow interference, thereby improving cleanliness stability. This method comprehensively characterizes the relationship between airflow patterns between clean benches through multidimensional feature analysis and clustering.

[0051] Figure 3 A flow chart of a method for determining the pollution risk distribution within the current operation scene is shown. Figure 3 As shown, the following steps are included.

[0052] Step S301 : for target clean workbenches with airflow interaction intensities greater than a first intensity threshold, a topological relationship network is constructed according to the airflow interaction intensities between the target clean workbenches to obtain a dynamic topological structure between the target clean workbenches.

[0053] Step S302 : constructing an airflow interference coefficient matrix between target clean workbenches according to the dynamic topology structure.

[0054] Step S303 : determining the cross contamination risk probability of particle transfer between target clean workbenches based on the airflow interference coefficient matrix and the particle concentration data included in the operation data, and obtaining a cross contamination risk probability set.

[0055] Step S304 : Based on the cross-contamination risk probability set, a spatial interpolation method is used to generate a spatial distribution of contamination risks between target clean workbenches as the contamination risk distribution in the current operation scene.

[0056] In the embodiments of this application, the target clean bench is a clean bench with an airflow interaction intensity greater than a first intensity threshold. The process of constructing a topological relationship network in this application makes the dynamic interactions between target clean benches more intuitive and tangible. This model construction method based on actual operating data greatly enhances the understanding of complex airflow environments. By simulating the particle transfer path and calculating the cross-contamination risk probability, this method can identify high-risk areas at an early stage, allowing for timely and effective preventive measures to prevent the spread of pollutants and protect the safety of the production environment.

[0057] In some embodiments, for a cluster of clean workbenches within a current operating scenario, when the airflow interaction intensity is greater than a first intensity threshold, a graph neural network, such as a GNN algorithm, is used to input the airflow interaction intensity between target clean workbenches, construct a topological relationship network, and obtain a dynamic topological structure between target clean workbenches. Specifically, if the airflow interaction intensity corresponding to the similarity metric is greater than the first intensity threshold, the corresponding target clean workbench identification pair is recorded as a candidate node pair, and the interaction intensity of the candidate node pair is processed by a weighted average method to obtain the initial weight of the node pair. A graph neural network algorithm is used to input the initial weight of the node pair and the target clean workbench identification, construct a topological relationship network, and generate a dynamic topological structure. As a feasible embodiment, when calculating the Euclidean distance between feature vectors, the feature vectors of two target clean benches in the same time window are compared, for example, [0.5, 6] and [0.6, 5]. The Euclidean distance reflects the similarity of airflow patterns; the smaller the distance, the closer the patterns. In one possible implementation, if the feature vector distance between one target clean bench and another is 0.08, it indicates that their airflow patterns are highly similar. This similarity metric provides the basis for subsequent interaction strength analysis. As a specific embodiment, the strength of airflow interaction is determined based on a similarity metric. If the Euclidean distance is less than a first strength threshold, such as 0.1, the airflow patterns of the two target clean benches are considered to have strong interaction and are recorded as a candidate node pair. For example, if the Euclidean distance between target clean benches A and B is 0.07, they are marked as a candidate node pair. This screening mechanism only focuses on target clean bench pairs with high interaction, thereby improving analysis efficiency. For example, the weighted average method considers the physical distance between target clean benches when processing the interaction strength of candidate node pairs. Close clean benches, such as those 1 meter apart, are assigned a weight of 0.7, while those farther away, such as those 3 meters apart, are assigned a weight of 0.3. After weighting, the interaction strength between clean benches A and B is 0.65. This method quantifies the degree of interaction and intuitively reflects the influence of airflow between the target clean benches. In one possible implementation, a graph neural network algorithm is used to construct a dynamic topological structure. The initial weights of candidate node pairs and the target clean bench identifiers are input to generate a dynamic topological structure. For example, target clean benches A, B, and C form a network, with the edge weight between A and B being 0.65 and the edge weight between B and C being 0.5. The graph neural network captures the dynamic relationship between the airflow patterns between target clean benches by iteratively updating the weights. This topological structure intuitively reflects the global characteristics of airflow interactions and provides a basis for optimizing the layout of target clean benches. For example, if the airflow velocity of a target clean bench is stable but the direction fluctuates greatly, its edge weight in the topological network is low, indicating that its airflow pattern is relatively independent. This analysis helps to identify special target clean benches and optimize airflow management. In some embodiments, a dynamic topology structure employs fluid dynamics simulation methods to input airflow velocity and direction, and calculate an airflow interference coefficient matrix between target clean workbenches. Specifically, each matrix element Mij in the airflow interference coefficient matrix M represents the airflow interaction intensity between the i-th and j-th target clean workbenches, and the target clean workbenches are identified by this pair. As a feasible embodiment, the fluid mechanics simulation method constructs an airflow interference coefficient matrix M based on the interaction weight and the target clean workbench identifier. For example, the interaction weight between target clean workbenches A and B is 0.57, and Mab is set to 0.57. The simulation process takes into account the airflow propagation path and attenuation to generate a global interference distribution. Compared with simple statistical methods, fluid mechanics simulation can better reflect the physical characteristics of airflow and facilitate the analysis of complex interference patterns. For example, in one possible implementation, the interference coefficient matrix M can be used to identify key interference sources. If the Mij value of a target clean workbench is higher than that of multiple other target clean workbenches, it indicates that its airflow pattern has a wide impact and its position or airflow setting needs to be adjusted first. This matrix analysis intuitively reflects the airflow interaction between the target clean workbenches and provides a basis for optimizing the layout of the target clean workbenches. As a specific example, in a real-world scenario, a cleanroom has 10 target clean benches. After preprocessing the sensor data, a sliding window is used to generate feature vectors. Cosine similarity is then used to filter out five interfering node pairs. After calculating the interaction weights using weighted average, a fluid dynamics simulation generates matrix M, which indicates that target clean bench 3 has high interference with other target clean benches. Based on this information, the fan angle of target clean bench 3 can be adjusted to reduce interference. This multi-step analysis method ensures targeted and efficient airflow management.

[0058] Figure 4 A flow chart of a method for obtaining a set of cross-contamination risk probabilities is shown. Figure 4 As shown, the following steps are included.

[0059] Step S401 : extracting time series data of particulate matter concentration data, and preprocessing the time series data of particulate matter concentration data using a standardized format to obtain a particulate matter concentration dataset.

[0060] Step S402 : for target matrix elements whose airflow interaction intensity is greater than a second intensity threshold, determine target clean bench pairs for identifying target matrix elements, and extract particle concentration data corresponding to each target clean bench pair from the particle concentration dataset.

[0061] In step S403 , according to the particle concentration data corresponding to each target clean bench pair and the airflow interaction intensity of the target matrix elements, a particle propagation path simulation is performed between the target clean bench pairs to obtain particle trajectory data.

[0062] Step S404 : statistically analyzing the cross-contamination risk probability of particle transfer between target clean bench pairs based on the position distribution and time distribution of the particles represented by the particle trajectory data.

[0063] Step S405 , summarizing the cross contamination risk probabilities corresponding to each pair of target clean workbenches to obtain a cross contamination risk probability set.

[0064] In the embodiments of the present application, in order to more accurately assess the risk of particulate matter transmission, the present application standardizes the particulate matter concentration data, combines the key information in the airflow interference coefficient matrix, and performs detailed propagation path simulation for a specific target clean workbench. This method can not only predict the possible movement trajectory of particulate matter, but also quantify its transfer probability, providing scientific support for the formulation of targeted prevention and control strategies, and greatly improving the accuracy and reliability of clean room management.

[0065] In some embodiments, for a cluster of clean workbenches within a current operating scenario, when the airflow interaction intensity is greater than a first intensity threshold, a graph neural network, such as a GNN algorithm, is used to input the airflow interaction intensity between target clean workbenches, construct a topological relationship network, and obtain a dynamic topological structure between target clean workbenches. Specifically, if the airflow interaction intensity corresponding to the similarity metric is greater than the first intensity threshold, the corresponding target clean workbench identification pair is recorded as a candidate node pair, and the interaction intensity of the candidate node pair is processed by a weighted average method to obtain the initial weight of the node pair. A graph neural network algorithm is used to input the initial weight of the node pair and the target clean workbench identification, construct a topological relationship network, and generate a dynamic topological structure.

[0066] As a feasible embodiment, when calculating the Euclidean distance between feature vectors, the feature vectors of two target clean benches in the same time window are compared, for example, [0.5, 6] and [0.6, 5]. The Euclidean distance reflects the similarity of airflow patterns; the smaller the distance, the closer the patterns. In one possible implementation, if the feature vector distance between one target clean bench and another is 0.08, it indicates that their airflow patterns are highly similar. This similarity metric provides the basis for subsequent interaction strength analysis.

[0067] As a specific embodiment, the airflow interaction intensity is determined based on a similarity metric. If the Euclidean distance is less than a first intensity threshold, such as 0.1, it is considered that the airflow patterns of the two target clean workbenches interact strongly and are recorded as candidate node pairs. For example, the Euclidean distance between target clean workbench A and target clean workbench B is 0.07, which is marked as a candidate node pair. This screening mechanism only focuses on target clean workbench pairs with high interaction, thereby improving analysis efficiency. For example, when processing the interaction intensity of candidate node pairs, the weighted average method needs to consider the physical distance between the target clean workbenches. For target clean workbenches that are close, such as 1 meter apart, the weight is set to 0.7, and for those that are far away, such as 3 meters, the weight is set to 0.3. After weighting, the interaction intensity between target clean workbenches A and B is 0.65. This method quantifies the degree of interaction and intuitively reflects the strength of the airflow influence between the target clean workbenches.

[0068] In one possible implementation, a graph neural network algorithm is used to construct a dynamic topological structure. The initial weights of candidate node pairs and the target clean bench identifiers are input to generate a dynamic topological structure. For example, target clean benches A, B, and C form a network, with the edge weight between A and B being 0.65 and the edge weight between B and C being 0.5. The graph neural network captures the dynamic relationship between the airflow patterns between target clean benches by iteratively updating the weights. This topological structure intuitively reflects the global characteristics of airflow interactions and provides a basis for optimizing the layout of target clean benches. For example, if the airflow velocity of a target clean bench is stable but the direction fluctuates greatly, its edge weight in the topological network is low, indicating that its airflow pattern is relatively independent. This analysis helps to identify special target clean benches and optimize airflow management.

[0069] In some embodiments, a dynamic topology structure employs fluid dynamics simulation methods to input airflow velocity and direction, and calculate an airflow interference coefficient matrix between target clean workbenches. Specifically, each matrix element Mij in the airflow interference coefficient matrix M represents the airflow interaction intensity between the i-th and j-th target clean workbenches, and the target clean workbenches are identified by this pair.

[0070] As a feasible embodiment, the fluid mechanics simulation method constructs an airflow interference coefficient matrix M based on the interaction weight and the target clean workbench identifier. For example, the interaction weight between target clean workbenches A and B is 0.57, and Mab is set to 0.57. The simulation process takes into account the airflow propagation path and attenuation to generate a global interference distribution. Compared with simple statistical methods, fluid mechanics simulation can better reflect the physical characteristics of airflow and facilitate the analysis of complex interference patterns. For example, in one possible implementation, the interference coefficient matrix M can be used to identify key interference sources. If the Mij value of a target clean workbench is higher than that of multiple other target clean workbenches, it indicates that its airflow pattern has a wide impact and its position or airflow setting needs to be adjusted first. This matrix analysis intuitively reflects the airflow interaction between the target clean workbenches and provides a basis for optimizing the layout of the target clean workbenches.

[0071] As a specific example, in a real-world scenario, a cleanroom has 10 target clean benches. After preprocessing the sensor data, a sliding window is used to generate feature vectors. Cosine similarity is then used to filter out five interfering node pairs. After calculating the interaction weights using weighted average, a fluid dynamics simulation generates matrix M, which indicates that target clean bench 3 has high interference with other target clean benches. Based on this information, the fan angle of target clean bench 3 can be adjusted to reduce interference. This multi-step analysis method ensures targeted and efficient airflow management.

[0072] Figure 5 A flow chart of a method for obtaining an optimized clean bench layout is shown, as Figure 5 As shown, the following steps are included.

[0073] Step S501: When the contamination risk probability of any area in the contamination risk distribution is greater than a first probability threshold, a genetic algorithm is used to encode the spatial coordinates of each clean bench and initialize the layout according to the layout of the clean bench cluster to obtain a clean bench layout population.

[0074] In step S502, a crossover and mutation operation is performed on the encoded clean bench layout population by combining the contamination risk probability and the predefined layout constraints, so as to update the spatial coordinates of each clean bench and iterate the layout of each clean bench in the clean bench layout population according to the updated spatial coordinates.

[0075] Step S503 : referring to the updated spatial coordinates of each clean workbench, a spatial interpolation method is used to generate an updated contamination risk distribution.

[0076] Step S504 , referring to the updated contamination risk distribution, screening the clean bench layout with the minimum contamination risk probability and lower than the second probability threshold from the iterated clean bench layout population as the optimized clean bench layout.

[0077] In the examples of this application, the application of a genetic algorithm demonstrates unique advantages in optimizing clean bench layouts. It dynamically adjusts spatial coordinates based on the distribution of contamination risks, continuously iterating to find the optimal layout. This process not only minimizes contamination risks but also takes into account the layout constraints in actual operations, ensuring the feasibility and practicality of the final solution. In this way, contamination risks can be effectively reduced and high-quality standards for the production environment can be guaranteed.

[0078] In some embodiments, a genetic algorithm is used to encode the spatial coordinates of the clean workbench and initialize the layout for the pollution risk distribution in the current operation scene to obtain a clean workbench layout population. Specifically, if the pollution risk probability P of any area in the pollution risk distribution is greater than the first probability threshold, the spatial coordinates of each clean workbench are encoded according to the layout of the clean workbench cluster to form an initial population. Through the crossover and mutation operation of the genetic algorithm, combined with the pollution risk probability P and the predefined layout constraints, the spatial coordinates of each clean workbench are updated, and the layout population is iterated according to the updated spatial coordinates. Furthermore, a spatial interpolation method is used to generate an updated pollution risk distribution, and the clean workbench layout with the smallest pollution risk probability and lower than the second probability threshold is screened from the iterated layout population as the optimized clean workbench layout solution.

[0079] As a feasible embodiment, if there is a high-risk area in the initial contamination risk distribution, that is, the contamination risk probability P is greater than the first probability threshold, the genetic algorithm is triggered to adjust the clean workbench layout. The spatial coordinates of the clean workbench are encoded as a gene sequence. For example, the coordinates (5,2) of clean workbench 3 can be represented as a binary string. The genetic algorithm forms a population by initializing multiple layout schemes and generates a new layout scheme based on crossover and mutation operations. For example, in one possible implementation, the coordinates of clean workbenches 3 and 5 are exchanged to form a new layout scheme. Combining the contamination risk probability P and layout constraints, such as the minimum distance between clean workbenches of 1 meter, the genetic algorithm evaluates the risk of each layout scheme. For example, under the new layout, clean workbench 3 is moved to (7,4), and the spatial interpolation calculation shows that the contamination risk probability P of this area is reduced to 0.2. The updated contamination risk distribution intuitively reflects the effect of the layout adjustment.

[0080] As a specific embodiment, the genetic algorithm compares the contamination risk probability P of each layout scheme in the population during the iteration process. If the contamination risk probability P of a certain scheme meets the second probability threshold, such as 0.25, the layout scheme with the lowest P is selected as the optimization result. For example, a certain optimization scheme moves the clean workbench 3 to the corner of the workshop, and the contamination risk probability P35 between it and the clean workbench 5 is reduced to 0.18, and the overall contamination risk distribution shows that there are no high-risk areas. This scheme is selected as the final optimized layout, which can guide the actual adjustment of the clean workbench in the workshop, thereby reducing the possibility of cross-contamination. For example, the entire optimization process forms a closed loop through sensor data collection, standardization processing, spatial interpolation analysis and genetic algorithm iteration. Sensor data provides initial input, standardization processing ensures data consistency, spatial interpolation generates a contamination risk distribution map, and the genetic algorithm accurately locates and optimizes high-risk areas through multiple rounds of iteration. This method not only improves the scientific nature of the clean workbench layout, but also significantly reduces the contamination risk and realizes intelligent dynamic adjustment.

[0081] Figure 6 A flow chart of a method for adjusting the operating power of a clean bench cluster is shown. Figure 6 As shown, the following steps are included.

[0082] Step S601 , extracting the airflow velocity data, clean bench operation power data and layout coordinate data contained in the operation data, and generating an airflow velocity matrix, an operation power vector and a spatial coordinate set through normalization processing as the energy consumption distribution of the clean bench cluster.

[0083] Step S602: determining energy consumption optimization parameters for optimizing energy consumption distribution.

[0084] Step S603 : When it is determined based on the energy consumption optimization parameter that there is a clean bench having an energy consumption value greater than the energy consumption threshold, the operating power of the clean bench cluster is adjusted.

[0085] In the embodiment of this application, considering the importance of energy consumption control for long-term operation, this application proposes a comprehensive analysis method, that is, generating energy consumption distribution parameters by standardizing airflow velocity, operating power and spatial coordinates. Once a clean workbench with high energy consumption is found, it can be adjusted through a series of optimization steps to achieve an improvement in overall energy efficiency. This approach not only helps to save energy and reduce emissions, but also conforms to the concept of modern enterprises pursuing green development.

[0086] Figure 7 A flow chart of a method for determining energy consumption optimization parameters for optimizing energy consumption distribution is shown. Figure 7 As shown, the following steps are included.

[0087] Step S701 : Based on the energy consumption distribution, a long short-term memory network is used to extract time series features of the air flow velocity matrix, the operating power vector, and the spatial coordinate set to obtain a sequence of energy consumption prediction values ​​for each clean workbench.

[0088] Step S702 : When it is determined based on the energy consumption prediction value sequence that there is a clean workbench with an energy consumption value greater than the energy consumption threshold, a linear regression method is used to calculate the adjustment coefficient of the energy consumption distribution according to the spatial coordinate set and the airflow velocity matrix to obtain an updated energy consumption distribution parameter.

[0089] Step S703 : Based on the updated energy consumption distribution parameters, combined with the spatial coordinates and the pre-collected pollution risk data, the energy consumption distribution parameters and the pollution risk data are fused by a weighted average method to obtain optimized energy consumption optimization parameters.

[0090] In this application's examples, a long-short-term memory network (LSTM) is used to extract temporal features, capturing the time-varying trends in clean bench energy consumption, providing strong support for early warning and dynamic adjustments. By combining linear regression to calculate adjustment coefficients and integrating energy consumption and pollution risk data through weighted averaging, this method can develop an optimization strategy that is both energy-efficient and environmentally friendly, fully embodying the concept of multi-objective optimization and promoting efficient resource utilization.

[0091] In some embodiments, for the optimized clean bench layout plan, the airflow velocity, clean bench operating power and spatial coordinates are input into the energy consumption prediction model based on the long short-term memory network, the energy consumption distribution of each clean bench is calculated, and energy consumption optimization parameters are generated. Specifically, airflow velocity data, operating power data and spatial coordinate data are obtained from the sensor, and the airflow velocity matrix, operating power vector and spatial coordinate set are generated through standardization processing as the energy consumption distribution of the clean bench cluster. Furthermore, for the energy consumption distribution, the long short-term memory network is used to extract the time series features of the airflow velocity matrix, operating power vector and spatial coordinate set to obtain a sequence of energy consumption prediction values ​​for each clean bench. If it is determined based on the energy consumption prediction value sequence that the energy consumption value of at least one clean bench is greater than the energy consumption threshold, the linear regression method is used to calculate the adjustment coefficient of the energy consumption distribution based on the spatial coordinate set and the airflow velocity matrix to generate updated energy consumption distribution parameters. Combined with the spatial coordinates and pollution risk data, the energy consumption distribution parameters and pollution risk data are fused through a weighted average method to ultimately obtain optimized energy consumption optimization parameters.

[0092] As a feasible embodiment, in the clean room energy consumption optimization scenario, sensors collect airflow velocity, operating power and spatial coordinate data in real time to provide a basis for energy consumption analysis. For example, the airflow velocity sensor records data at a frequency of 2 times per second, such as in the range of 0.1-1.2m / s, the operating power sensor collects clean bench power, such as in the range of 50-200W, and the spatial coordinates record the position of the clean bench, such as clean bench 1 is located at (2,3). Through standardization to ensure data consistency, the airflow velocity is normalized to the range of 0-1 and outliers are eliminated, such as data greater than 1.5m / s, to generate an airflow velocity matrix, such as elements representing the airflow impact between clean benches. The operating power is standardized into a vector, such as the power of clean bench 1 is normalized to 0.4, and the spatial coordinate set retains the original spatial information to form an energy consumption distribution.

[0093] As a specific embodiment, a long short-term memory network is used to extract time series features. The network analyzes the airflow velocity matrix, operating power vector, and spatial coordinate set to capture the time-varying patterns of clean bench energy consumption. For example, clean bench 2 operates at consistently high power in a high-airflow area, and the network predicts its energy consumption sequence for the next hour to be [120, 125, 130 W]. If the predicted energy consumption of clean bench 2 exceeds an energy consumption threshold, such as 100 W, the optimization process is triggered. This time series analysis accurately identifies high-energy-consuming clean benches, providing a basis for subsequent optimization.

[0094] For example, a linear regression method calculates the energy consumption distribution adjustment coefficient based on a set of spatial coordinates and an airflow velocity matrix. Assuming that clean benches 2 and 3 are 2 meters apart, and the airflow velocity matrix indicates an influence coefficient of 0.7, regression analysis estimates that moving clean bench 2 reduces energy consumption by a factor of 0.85. The adjustment coefficient reflects the impact of the layout change on energy consumption and generates updated energy consumption distribution parameters. This method ensures the rationality of optimization by quantifying the relationship between space and airflow.

[0095] In one possible implementation, a weighted average method combines energy consumption distribution parameters with contamination risk data to generate optimized parameters. Contamination risk data is derived from historical contamination risk distributions. For example, the risk probability near clean bench 2 is 0.4. The energy consumption distribution parameters are assigned a weight of 0.6, and the contamination risk data is assigned a weight of 0.4. The weighted average is calculated, yielding an optimized parameter of 0.32 for clean bench 2. The new layout moves clean bench 2 to a low airflow area, such as (5,5), reducing both energy consumption and contamination risk. This fusion approach balances multi-objective optimization and improves overall workshop efficiency.

[0096] For example, after optimizing parameters to guide the reconfiguration of clean benches, the airflow velocity matrix was updated, showing that the influence coefficient between clean benches 2 and 3 dropped to 0.5, and the predicted energy consumption sequence dropped to [90, 95, 100W], below the energy consumption threshold. Contamination risk data also decreased, for example, to 0.25 near clean bench 2. This closed-loop optimization significantly improves energy efficiency and environmental safety through data-driven decision-making.

[0097] Figure 8 A flow chart of a method for adjusting the operating power of a clean bench cluster is shown. Figure 8 As shown, the following steps are included.

[0098] Step S801: determining an airflow change trend matrix based on operating data.

[0099] Step S802 : Using a feedback control algorithm, iteratively calculate the airflow change trend matrix and the operating power vector to obtain a power adjustment coefficient sequence.

[0100] Step S803 : updating the operating power of the clean bench cluster by a linear interpolation method according to the power adjustment coefficient sequence to obtain an updated operating power data set.

[0101] In step S804 , the updated operating power data set is combined with the airflow velocity data and the energy consumption optimization parameters and the operating status of the clean bench are integrated by a weighted average method to obtain an optimized operating power data set.

[0102] Step S805 : adjusting the operating power of the clean bench cluster according to the optimized operating power data set.

[0103] In the examples of this application, the introduction of a feedback control algorithm enables the system to automatically adjust the clean bench operating power based on real-time data. This method generates a sequence of power adjustment coefficients and applies linear interpolation techniques to gradually update the operating power. This allows for a smooth transition to a more optimal state without affecting normal operation. Furthermore, incorporating airflow velocity data into the optimization process further enhances the system's adaptability and flexibility, ensuring the effectiveness of the optimization results.

[0104] In some embodiments, for the adjustment of the operating power of the clean bench cluster, if the energy consumption value of any clean bench in the energy consumption optimization parameters is greater than the energy consumption threshold, such as T3 = 1.2 times the average energy consumption, the feedback control algorithm is used to input the energy consumption optimization parameters and the airflow change trend matrix, adjust the clean bench operating power, and generate new operating data. Specifically, the airflow velocity data and operating power data are obtained from the sensor, and the airflow change trend matrix and the operating power vector are generated through standardization processing to form an initial input data set. If there is at least one clean bench in the initial input data set with an energy consumption value greater than the energy consumption threshold, the feedback control algorithm is used to iteratively calculate the airflow change trend matrix and the operating power vector to generate a power adjustment coefficient sequence. According to the power adjustment coefficient sequence, the clean bench operating power is updated by the linear interpolation method to generate an updated operating power data set. Furthermore, combined with the airflow velocity data, the energy consumption optimization parameters and the clean bench operating status are fused by the weighted average method, the optimized operating power data set is output, and the operating power of the clean bench cluster is adjusted according to the optimized operating power data set.

[0105] In some embodiments, the airflow velocity data of each clean workbench may be normalized to obtain normalized airflow velocity data, and then an airflow change trend matrix may be generated by using each normalized airflow velocity data as a matrix element.

[0106] As a feasible embodiment, in the clean room energy consumption optimization scenario, the data collected by sensors in real time provides the basis for subsequent analysis. For example, the air velocity sensor records data at a frequency of 2 times per second, such as in the range of 0.2-1.0m / s, and the operating power sensor collects the clean bench power, such as in the range of 60-180W. Through standardization processing to ensure data consistency, the air flow velocity is normalized to the range of 0-1 and outliers are eliminated, such as data greater than 1.2m / s, to generate an air flow change trend matrix, such as matrix elements reflecting the interaction of airflow between clean benches. The operating power is standardized into a vector, such as the power of clean bench 1 is normalized to 0.35, which constitutes the initial input data set. This standardization process ensures a unified data format, which is convenient for subsequent algorithm processing.

[0107] As a specific embodiment, the feedback control algorithm is activated when the energy consumption value of a clean bench in the energy optimization parameters exceeds a threshold. For example, assuming the energy consumption value of clean bench 2 is 120W, which exceeds a threshold, such as 100W, the algorithm iteratively calculates the airflow trend matrix and the operating power vector. The airflow trend matrix shows that the airflow influence coefficient between clean benches 2 and 3 is 0.6, and the operating power vector indicates that the power of clean bench 2 is too high. Through multiple iterations, the algorithm generates a dynamic power adjustment coefficient sequence, such as [0.9, 0.85, 0.8], reflecting the trend of gradually reducing power. This sequence serves as the dynamic basis for adjusting the operating power of the clean benches.

[0108] For example, the linear interpolation method updates the clean bench operating power based on a sequence of power adjustment coefficients. Assuming the original power of clean bench 2 is 120W and the adjustment coefficient sequence is [0.9, 0.85], interpolation calculates the new power values ​​of 108W and 102W, respectively, generating an updated operating power dataset. This method adjusts power through smooth transitions, avoiding sudden changes in operating status and ensuring stability.

[0109] In one possible implementation, a weighted average method combines energy optimization parameters with the clean bench's operating status to generate an optimized result. For example, the energy optimization parameters are based on the adjusted power dataset, while the clean bench's operating status includes information such as operating time and load factor. Assuming that Clean Bench 2's energy optimization parameter is 0.4 and its operating status score is 0.5, with weighted ratios of 0.7 and 0.3, the combined optimization result is 0.43. This optimization result guides further fine-tuning of the clean bench's operating power, for example, stabilizing Clean Bench 2's power at 95W to reduce energy consumption.

[0110] Specifically, airflow velocity data provides environmental constraints during the fusion process. For example, the airflow trend matrix shows that the airflow velocity in the area where clean bench 2 is located is 0.8 m / s, which is too high and could lead to increased energy consumption. The optimization results, combined with the airflow data, suggest reducing the wind speed in this area or adjusting the clean bench's position. This comprehensive analysis ensures that the optimization solution is adapted to the actual environment.

[0111] For example, the iterative calculations of the feedback control algorithm offer dynamic adaptability. When airflow velocity changes due to external air volume adjustments, the algorithm recalculates the airflow trend matrix and updates the power adjustment coefficient sequence. This adaptive capability makes the optimization process more flexible and suitable for dynamic production scenarios.

[0112] As a specific example, the linear interpolation method can be used in power adjustment to optimize the clean bench's operating status. For example, if clean bench 2 is a high-load device, interpolation adjustment prioritizes operational stability, keeping the power reduction within 10%. This refined adjustment balances energy consumption optimization with device performance.

[0113] It's important to note that when integrating energy optimization parameters and operating status, the weighting of the weights can be adjusted based on shop floor priorities. If energy optimization is the core objective, the weighting is biased toward energy optimization parameters. If equipment stability is the core objective, the weighting of operating status is increased. This flexibility allows optimization results to better meet actual needs.

[0114] The present application also provides a system embodiment that is consistent with the above embodiment, which is used to implement the method steps of the above embodiment. The explanation based on the same name meaning is the same as the above embodiment, and has the same technical effect as the above embodiment, which will not be repeated here.

[0115] like Figure 9 As shown, the present application provides a clean bench intelligent control system 900 based on the Internet of Things, including: The acquisition unit 901 is configured to acquire the operation data of each clean workbench in the clean workbench cluster within the current operation scene, and extract the time series data of the airflow velocity and airflow direction contained in the operation data.

[0116] The processing unit 902 is configured to perform data preprocessing on the time series data using a preset standardized format to obtain a preprocessed data set, wherein the preprocessed data set includes preprocessed data corresponding to each clean bench identifier.

[0117] The acquisition unit 901 is further configured to extract the characteristic vectors of the airflow velocity and airflow direction of each clean workbench from the preprocessed data set, and calculate the Euclidean distance between different characteristic vectors, so as to use the Euclidean distance as a similarity measure of the airflow patterns between different clean workbenches.

[0118] Determination unit 93 is configured to determine the airflow interaction intensity between each clean workbench based on the similarity metric. Furthermore, it is configured to determine the contamination risk distribution within the current operation scenario based on the similarity metric and the airflow interaction intensity, and optimize the clean workbench layout within the current operation scenario based on the contamination risk distribution to obtain an optimized clean workbench layout.

[0119] The control unit 904 is used to schedule each clean workbench according to the optimized clean workbench layout.

[0120] Regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0121] Although operations are described in a particular order in the drawings, this should not be understood as requiring that the operations be performed in the particular order shown or in serial order, or that all shown operations be performed to obtain the desired results. In certain circumstances, multitasking and parallel processing may be advantageous.

[0122] The methods and systems of the present application can be implemented using standard programming techniques, using rule-based logic or other logic to implement the various method steps. It should also be noted that the terms "system" and "module" as used herein and in the claims are intended to include implementations using one or more lines of software code and / or hardware implementations and / or devices for receiving input.

[0123] Any steps, operations or procedures described herein may be performed or implemented using one or more hardware or software modules, either alone or in combination with other devices. In one embodiment, the software modules are implemented using a computer program product comprising a computer-readable medium containing computer program code, which can be executed by a computer processor to perform any or all of the steps, operations or procedures described.

[0124] The foregoing description of the implementation of the present application has been provided for purposes of illustration and description. The foregoing description is not intended to be exhaustive or to limit the present application to the precise form disclosed, and various variations and modifications are possible in accordance with the above teachings or may result from the practice of the present application. These embodiments have been selected and described in order to illustrate the principles of the present application and its practical application, so as to enable those skilled in the art to utilize the present application in various embodiments and modifications as appropriate for the particular use contemplated.

[0125] Regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0126] It is further understood that, unless otherwise specified, “connection” includes a direct connection where there are no other components between the two elements, and also includes an indirect connection where there are other elements between the two elements.

[0127] It should be further understood that although operations are described in a particular order in the drawings in the embodiments of the present application, this should not be construed as requiring that these operations be performed in the particular order shown or in a serial order, or that all of the illustrated operations be performed to obtain the desired results. In certain circumstances, multitasking and parallel processing may be advantageous.

[0128] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. This application is intended to encompass any variations, uses, or adaptations of the present application that follow the general principles of the present application and include common knowledge or customary techniques in the field of the present application that are not disclosed herein. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the scope of claims below.

[0129] It should be understood that the present application is not limited to the precise structures described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the scope of the appended claims.

[0130] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A clean bench intelligent control method based on the Internet of Things, characterized in that: include: For the clean workbench cluster in the current operation scene, obtain the operating data of each clean workbench, and extract the time series data of airflow velocity and airflow direction contained in the operating data; Preprocessing the time series data using a preset standardized format to obtain a preprocessed data set, wherein the preprocessed data set includes preprocessed data corresponding to each of the clean bench identifiers; Extracting characteristic vectors of the airflow velocity and airflow direction of each clean workbench from the preprocessed data set, and calculating the Euclidean distance between different characteristic vectors, so as to use the Euclidean distance as a similarity measure of the airflow patterns between different clean workbenches; Determining the airflow interaction intensity between the clean workbenches according to the similarity measurement; Determining a contamination risk distribution within the current operation scene according to the similarity metric and the airflow interaction intensity, and optimizing a clean workbench layout within the current operation scene according to the contamination risk distribution to obtain an optimized clean workbench layout; According to the optimized clean workbench layout, each clean workbench is scheduled.

2. The method according to claim 1, characterized in that Determining the airflow interaction intensity between the clean workbenches according to the similarity measurement includes: Performing local clustering on the similarity measure to obtain a clustering result for each clean bench; Based on the clustering results, calculating the airflow pattern overlap between different clean workbenches; The airflow pattern overlap is weighted averaged to obtain the airflow interaction intensity between the clean workbenches.

3. The method according to claim 1, characterized in that The determining, based on the similarity metric and the airflow interaction intensity, the pollution risk distribution within the current operation scene includes: For the target clean workbenches whose airflow interaction intensity is greater than the first intensity threshold, a topological relationship network is constructed according to the airflow interaction intensity between the target clean workbenches to obtain a dynamic topological structure between the target clean workbenches; According to the dynamic topology, an airflow interference coefficient matrix between the target clean workbenches is constructed; Determining a cross-contamination risk probability of particle transfer between the target clean workbenches based on the airflow interference coefficient matrix and the particle concentration data included in the operation data, and obtaining a cross-contamination risk probability set; According to the cross-contamination risk probability set, a spatial distribution of contamination risks between the target clean workbenches is generated by spatial interpolation to serve as the contamination risk distribution within the current operation scene.

4. The method according to claim 3, characterized in that Determining the cross-contamination risk probability of particle transfer between the target clean workbenches based on the airflow interference coefficient matrix and the particle concentration data included in the operation data to obtain a cross-contamination risk probability set includes: Extracting time series data of the particulate matter concentration data, and preprocessing the time series data of the particulate matter concentration data using a standardized format to obtain a particulate matter concentration dataset; For the target matrix element whose airflow interaction intensity is greater than the second intensity threshold, determining a target clean bench pair for identifying the target matrix element, and extracting the particle concentration data corresponding to each of the clean bench pairs from the particle concentration dataset; According to the particle concentration data corresponding to each of the target clean workbench pairs and the airflow interaction intensity of the target matrix elements, a particle propagation path simulation is performed between the clean workbench pairs to obtain particle trajectory data; Statistically analyzing the cross-contamination risk probability of particle transfer between the target clean workbenches based on the position distribution and time distribution of the particles represented by the particle trajectory data; The cross contamination risk probabilities corresponding to each pair of target clean workbenches are summarized to obtain a cross contamination risk probability set.

5. The method according to claim 1, characterized in that Optimizing the clean bench layout in the current operation scene according to the pollution risk distribution to obtain an optimized clean bench layout includes: If the contamination risk probability of any area in the contamination risk distribution is greater than a first probability threshold, encoding the spatial coordinates of each clean workbench and initializing the layout according to the layout of the clean workbench cluster using a genetic algorithm to obtain a clean workbench layout population; For the encoded clean bench layout population, combining the contamination risk probability and predefined layout constraints, performing a crossover mutation operation through a genetic algorithm to update the spatial coordinates of each clean bench and iterate the layout of each clean bench in the clean bench layout population according to the updated spatial coordinates; Referring to the updated spatial coordinates of each of the clean workbenches, a spatial interpolation method is used to generate an updated contamination risk distribution; With reference to the updated contamination risk distribution, a clean bench layout with the minimum contamination risk probability and lower than a second probability threshold is selected from the iterated clean bench layout population as the optimized clean bench layout.

6. The method according to claim 1 or 5, characterized in that The method further comprises: Extracting the airflow velocity data, clean bench operating power data, and layout coordinate data contained in the operating data, and generating the airflow velocity matrix, the operating power vector, and the spatial coordinate set through normalization processing as the energy consumption distribution of the clean bench cluster; determining energy consumption optimization parameters for optimizing the energy consumption distribution; If it is determined based on the energy consumption optimization parameter that there is a clean bench with an energy consumption value greater than the energy consumption threshold, the operating power of the clean bench cluster is adjusted.

7. The method according to claim 6, characterized in that The determining of energy consumption optimization parameters for optimizing the energy consumption distribution includes: Based on the energy consumption distribution, a long short-term memory network is used to extract time series features of the airflow velocity matrix, the operating power vector and the spatial coordinate set to obtain a sequence of energy consumption prediction values ​​for each clean workbench; If it is determined based on the energy consumption prediction value sequence that there is a clean workbench with an energy consumption value greater than the energy consumption threshold, then according to the spatial coordinate set and the airflow velocity matrix, a linear regression method is used to calculate the adjustment coefficient of the energy consumption distribution to obtain an updated energy consumption distribution parameter; According to the updated energy consumption distribution parameters, combined with the spatial coordinates and pre-collected pollution risk data, the energy consumption distribution parameters and the pollution risk data are fused by a weighted average method to obtain optimized energy consumption optimization parameters.

8. The method according to claim 6, characterized in that The adjusting the operating power of the clean bench cluster includes: determining an airflow change trend matrix based on the operating data; Iteratively calculating the airflow change trend matrix and the operating power vector using a feedback control algorithm to obtain a power adjustment coefficient sequence; According to the power adjustment coefficient sequence, the operating power of the clean bench cluster is updated by a linear interpolation method to obtain an updated operating power data set; For the updated operating power data set, combined with the airflow velocity data, the energy consumption optimization parameters and the operating status of the clean bench are fused by a weighted average method to obtain an optimized operating power data set; The operating power of the clean bench cluster is adjusted according to the optimized operating power data set.

9. The method according to claim 8, characterized in that Determining the airflow change trend matrix according to the operating data includes: Normalizing the airflow velocity data of each clean workbench to obtain normalized airflow velocity data; Each of the normalized airflow velocity data is used as a matrix element to generate an airflow change trend matrix.

10. An intelligent control system for a clean workbench based on the Internet of Things, characterized in that: include: an acquisition unit, configured to acquire, for a cluster of clean workbenches in a current operation scene, operation data of each clean workbench, and extract time series data of airflow velocity and airflow direction contained in the operation data; a processing unit, configured to perform data preprocessing on the time series data using a preset standardized format to obtain a preprocessed data set, wherein the preprocessed data set includes preprocessed data corresponding to each of the clean bench identifiers; The acquisition unit is further configured to extract characteristic vectors of the airflow velocity and airflow direction of each clean workbench from the preprocessed data set, and respectively calculate the Euclidean distance between different characteristic vectors, so as to use the Euclidean distance as a similarity measure of the airflow patterns between different clean workbenches; a determining unit, configured to determine the airflow interaction intensity between the clean workbenches according to the similarity metric; and determining the pollution risk distribution within the current operation scene according to the similarity metric and the airflow interaction intensity, and optimizing the clean workbench layout within the current operation scene according to the pollution risk distribution to obtain an optimized clean workbench layout; A control unit is used to schedule each of the clean workbenches according to the optimized clean workbench layout.

Images