A method and system for microcontamination control in the production of high purity materials

Abstract

The application relates to the technical field of high-purity material preparation, in particular to a micro-pollution control method and system in a high-purity material preparation process. The method comprises the following steps: obtaining a real-time environment data stream and extracting a micro-pollution feature region, calculating a feature point information contribution coefficient to divide a controlled region block; positioning a target block at a matching moment, calculating a pollution matching degree based on a feature information vector and a spatial position reasonable coefficient, and integrating a region block meeting a preset threshold into a continuous diffusion path sequence; integrating the path sequence and mapping to a three-dimensional geographic coordinate system to generate a dynamic diffusion track; and according to the track, controlling an isolation valve, a fan filter unit and a negative pressure regulating valve to execute a blocking action. The application can overcome the associated errors caused by the variable working conditions, realize real-time tracking and accurate blocking of the diffusion path of the trace impurities, significantly improve the micro-pollution protection capability, and guarantee the material purity.

Description

Technical Field

[0001] This invention relates to the field of high-purity material preparation technology, and in particular to a method and system for controlling micro-contamination during the preparation of high-purity materials. Background Technology

[0002] With the rapid development of high-end manufacturing industries such as semiconductors, photovoltaics, and aerospace, the purity requirements for high-purity materials have reached extremely high levels. Even minute amounts of impurities introduced during material preparation, purification, and packaging can significantly affect the physical properties and chemical stability of the final product. Therefore, establishing a scientific and rigorous micro-contamination control system is of significant practical value for improving the yield of high-purity materials, ensuring the reliability of downstream precision manufacturing, and optimizing process flows.

[0003] Existing methods for controlling microcontamination typically rely on maintaining a cleanroom environment and regularly cleaning production equipment. However, in actual production scenarios, dynamic wear and tear during material transport, trace emissions from operators, and instantaneous fluctuations in environmental parameters render static defense measures limited in the face of complex and ever-changing production conditions. When monitoring frequency is insufficient or feedback mechanisms are lagging, there is often room for improvement in the precise location and immediate blocking of microcontamination sources, which can easily lead to fluctuations or substandard material purity. Summary of the Invention

[0004] To address the technical problems of existing microcontamination control methods in high-purity material preparation processes being too static, having lagging monitoring and feedback mechanisms, and insufficient accuracy in locating contamination sources, this invention aims to provide a microcontamination control method and system for high-purity material preparation processes. This invention maintains extremely high purity of the preparation environment by constructing a dynamic regional monitoring model and a contamination characteristic matching mechanism, and by real-time tracking and precise blocking of the diffusion paths of trace impurities in the preparation environment. The specific technical solution adopted is as follows:

[0005] A method for controlling microcontamination during the preparation of high-purity materials includes the following steps:

[0006] The real-time environmental data streams collected by each monitoring sensor within the preparation area are acquired, the micro-pollution feature regions in the data frames at each sampling time are extracted, the information contribution coefficient of each monitoring feature point within each micro-pollution feature region is calculated, and then each micro-pollution feature region is divided into an independent controlled region block.

[0007] At the matching moment when the initial micro-pollution disturbance is detected, the corresponding target controlled area block is located, and a matching process is performed between any target controlled area block at the matching moment and each area block to be matched at the matching moment. The matching process includes calculating the pollution matching degree based on the feature information vectors and spatial location rationality coefficients of the target controlled area block and the area blocks to be matched.

[0008] If the maximum value of the pollution matching degree is greater than the preset matching threshold, the corresponding area block to be matched is included in the continuous diffusion path sequence of the micro-pollution source and used as a new matching reference frame. The matching process is repeated until all data frames in the preparation cycle have completed the matching analysis.

[0009] The continuous diffusion path sequence is integrated and mapped to the real three-dimensional geographic coordinate system of the preparation workshop to generate the dynamic diffusion trajectory of the micro-pollution source. The corresponding actuator is then controlled to perform blocking actions based on the dynamic diffusion trajectory.

[0010] Furthermore, before acquiring the real-time environmental data stream collected by each monitoring sensor within the preparation area, the following steps are also included:

[0011] The sampling clocks of each monitoring sensor distributed within the preparation area are calibrated using a network synchronization protocol, so that the data streams collected by each monitoring sensor have a timestamp with a unified time reference.

[0012] The real-time environmental data stream is preprocessed. The preprocessing includes: using a preset compensation algorithm to eliminate nonlinear deviations caused by sensor drift due to environmental temperature and humidity changes, and standardizing and normalizing monitoring data of different dimensions.

[0013] Furthermore, the micro-pollution feature regions in the real-time environmental data streams collected by each monitoring sensor within the preparation area include:

[0014] The gradient change field of environmental parameters between adjacent sampling periods is obtained. The gradient change field is optimized and clustered based on the direction of the gradient vector to obtain feature point groups with consistent gradient direction. Each feature point group is defined as a micro-pollution feature region.

[0015] Furthermore, the process of obtaining the information contribution coefficient includes:

[0016] Determine the reference neighborhood feature points of the current monitoring feature point. The current monitoring feature point and the reference neighborhood feature points together form a multidimensional parameter analysis region.

[0017] Obtain the spatial physical distance between the current monitored feature point and any of the reference neighbor feature points, the difference in monitored values, and the maximum spatial scale within the multidimensional parameter analysis area.

[0018] Based on the spatial physical distance, the difference in the monitored values, and the maximum spatial scale, the information contribution coefficient of the current monitored feature point is obtained through a preset contribution function.

[0019] Furthermore, the process of dividing the controlled region block includes:

[0020] A local monitoring point set is constructed based on the monitoring feature points. The process of constructing the local monitoring point set includes: for the micro-pollution feature area that does not contain the assigned feature points, the feature points with the largest and second largest distance values ​​from the geometric center of the area are assigned to the local monitoring point set.

[0021] Calculate the rationality of adding each unassigned feature point to the local monitoring point set. If the rationality value of the partition is greater than a preset rationality threshold, then assign the unassigned feature point to the local monitoring point set.

[0022] When the construction process ends and there are still unassigned feature points, the construction process is repeated to construct a new set of local monitoring points until all monitoring feature points have a unique local monitoring point set to which they belong. A set of local monitoring points is defined as a controlled region block.

[0023] Furthermore, the process of obtaining the rationality of the division includes:

[0024] The multidimensional spatial volume enclosed by the outer contours of the unassigned feature points after they have been assigned to the local monitoring point set is obtained. Geometric compactness is then determined based on the ratio of this spatial volume to the volume of its smallest circumscribed envelope.

[0025] The average difference between the information contribution coefficient and the mean information contribution of each monitoring feature point within the local monitoring point set is obtained, as well as the single-point difference between the information contribution coefficient and the mean information contribution of the unassigned feature points. Information consistency is determined based on the ratio of the average difference to the single-point difference.

[0026] The rationality of the partition is obtained based on the weighted summation result of the geometric compactness and the information consistency.

[0027] Furthermore, the process of selecting the time to be matched includes:

[0028] When the matching time is a non-terminating frame in its data sequence, the time to be matched is the next adjacent sampling time that is incremented by the time step of the matching time.

[0029] When the matching time is the termination frame in its data sequence, the time to be matched includes the starting sampling time of any unmatched sequence in other associated monitoring dimensions.

[0030] Furthermore, the process of obtaining the contamination matching degree includes:

[0031] When the matching time is a non-terminating frame in its data sequence, the cosine similarity of the feature information vectors between any target controlled region block at the matching time and any controlled region block at an adjacent matching time is obtained.

[0032] Calculate the spatial distance between the physical center of the target controlled region block and the physical center of any controlled region block in the adjacent matching time, and define the inverse proportional function value of the ratio of the spatial distance to the equivalent diameter of the target controlled region block as the first position rationality coefficient.

[0033] The contamination matching degree is calculated based on the product of the cosine similarity and the reasonableness coefficient of the first position.

[0034] Furthermore, the process of obtaining the contamination matching degree also includes:

[0035] When the matching time is the termination frame in its data sequence, the cosine similarity of the feature information vectors between any target controlled region block at the matching time and any controlled region block at the start time of any unmatched sequence is obtained.

[0036] The inverse proportional function value of the time deviation between the actual response time of the micro-pollution source in the unmatched sequence and the response time interval predicted based on the diffusion model is calculated and defined as the second location rationality coefficient.

[0037] The contamination matching degree is calculated based on the product of the cosine similarity and the second position reasonable coefficient.

[0038] This invention also provides a micro-contamination control system in the preparation process of high-purity materials, the system comprising:

[0039] The region block division module is configured to acquire real-time environmental data streams collected by each monitoring sensor within the preparation area, extract micro-pollution feature regions from the data frames at each sampling time, calculate the information contribution coefficient of each monitoring feature point within each micro-pollution feature region, and then divide each micro-pollution feature region into independent controlled region blocks.

[0040] The matching calculation module is configured to locate the corresponding target controlled area block at the matching time when the initial micro-pollution disturbance is detected, and to perform a matching process between any target controlled area block at the matching time and each area block to be matched at the matching time. The matching process includes calculating the pollution matching degree based on the feature information vectors and spatial location rationality coefficients of the target controlled area block and the area blocks to be matched.

[0041] The path tracking module is configured to include the corresponding unmatched area block into the continuous diffusion path sequence of the micro-pollution source if the maximum value of the pollution matching degree is greater than a preset matching threshold, and use it as a new matching reference frame, repeatedly calling the matching calculation module until all data frames are matched.

[0042] The execution control module is configured to integrate the continuous diffusion path sequence and map it to a real geographic coordinate system to generate a dynamic diffusion trajectory, and output control signals to the execution mechanism according to the coordinates of the dynamic diffusion trajectory.

[0043] The actuator includes an isolation valve located on the material transport channel, a fan filter unit located at the top of the preparation chamber, and a negative pressure regulating valve located at the exhaust outlet. The execution control module changes the airflow direction and pressure gradient in the region where the dynamic diffusion trajectory is located by adjusting the opening and closing state of the isolation valve, the rotational speed frequency of the fan filter unit, and the valve opening degree of the negative pressure regulating valve.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] This invention acquires real-time environmental data streams from various monitoring sensors within a preparation area and divides the micro-pollution characteristic region into subdivided controlled area blocks. By independently analyzing these controlled area blocks and utilizing the stability of local features under dynamic conditions, the dependence of the micro-pollution control system on a single global concentration index is reduced, improving monitoring sensitivity and the continuity of diffusion path identification in complex airflow environments. This invention locates the target controlled area block at the matching time and introduces feature information vectors and spatial location rationality coefficients for multi-dimensional matching. By calculating the pollution matching degree, false alarm interference caused by instantaneous fluctuations in environmental parameters can be eliminated. The higher the pollution matching degree, the stronger the physical correlation between the area block to be matched and the known pollution source, thereby achieving accurate source tracing of trace amounts of impurities. This invention employs a recursive matching mechanism based on a preset matching threshold. By continuously incorporating eligible area blocks to be matched into a continuous diffusion path sequence, a complete evolutionary chain of micro-pollution sources in the spatiotemporal dimension is constructed. This frame-by-frame matching method ensures that even under conditions of extremely low pollutant concentration and extremely fast diffusion speed, the system can still maintain real-time tracking of the pollution source and its impact range. This invention transforms abstract data associations into concrete spatial physical trajectories by integrating continuous diffusion path sequences and mapping them to a real geographic coordinate system. By directly applying these trajectories to actuators such as isolation valves, fan filter units, and negative pressure regulating valves, the system can dynamically adjust the pressure field and airflow organization within the preparation environment based on the actual path of contaminant diffusion. The continuous path sequence of each controlled area block reflects the infiltration process of impurities at different process nodes. By fusing these local sequences to generate a global diffusion trajectory, precise execution criteria are provided for instantly blocking contaminant sources and ensuring material purity. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 A flowchart of a micro-contamination control method in the high-purity material preparation process provided in the first embodiment of the present invention;

[0048] Figure 2 A flowchart illustrating the process of obtaining the information contribution coefficient provided in the second embodiment of the present invention;

[0049] Figure 3 A flowchart illustrating the process of dividing the controlled region block according to the third embodiment of the present invention;

[0050] Figure 4 A flowchart illustrating the process of obtaining the rationality of the division provided in the fourth embodiment of the present invention;

[0051] Figure 5 A flowchart illustrating the process of obtaining the contamination matching degree as provided in the fifth embodiment of the present invention;

[0052] Figure 6 A flowchart illustrating another process for obtaining the contamination matching degree provided in the sixth embodiment of the present invention;

[0053] Figure 7 This is a schematic diagram of the micro-contamination control system in the high-purity material preparation process provided in the seventh embodiment of the present invention. Detailed Implementation

[0054] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a micro-contamination control method and system in the high-purity material preparation process proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0056] The specific scheme of the micro-contamination control method in the preparation process of high-purity materials provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0057] Please see Figure 1 The diagram illustrates a flowchart of a micro-contamination control method in the high-purity material preparation process provided in the first embodiment of the present invention, the method comprising:

[0058] S101. Obtain the micro-pollution feature regions of each data frame in the real-time environmental data stream collected by each monitoring sensor in the preparation area, calculate the information contribution coefficient of each monitoring feature point in each micro-pollution feature region, and then divide each micro-pollution feature region into controlled area blocks.

[0059] Before acquiring the micro-pollution feature regions of each data frame in the real-time environmental data stream collected by each monitoring sensor within the preparation area, the following steps are also included:

[0060] The sampling period of each monitoring sensor in the preparation area is synchronized by a precision clock protocol, so that the data streams collected by each monitoring sensor have a timestamp with a unified time reference.

[0061] The real-time environmental data stream is preprocessed, including: eliminating sensor nonlinear drift using a temperature and humidity compensation model and performing dimensional standardization.

[0062] The micro-pollution feature regions in the real-time environmental data streams collected by each monitoring sensor within the preparation area include:

[0063] The gradient change field between consecutive sampling frames is obtained, the gradient change field is optimized, and the spatial sampling points in each frame data are clustered based on the gradient vector direction to obtain the sampling point group with the same gradient direction. Each sampling point group corresponds to a micro-contamination feature region.

[0064] In detail, the core objective of this invention is to construct a complete and continuous diffusion trajectory of micro-pollutants within the controlled area (i.e., the preparation area) by cross-monitoring and matching the identities of impurity sources based on environmental parameter sequences collected by multiple monitoring sensors distributed throughout the preparation workshop. To achieve this goal, the data segments collected by each monitoring sensor must first be processed. Specific steps include:

[0065] a. Data acquisition: Acquire environmental data (such as particle concentration, chemical composition, airflow velocity, etc.) synchronously collected by all monitoring sensors in the preparation area, and ensure that the sampling frequency of each sensor is accurately synchronized;

[0066] b. Time alignment: Based on the timestamp information carried by the data stream, sensor data from different physical locations are uniformly mapped to the same time axis to achieve time step synchronization between multi-dimensional parameters;

[0067] c. Data preprocessing: Preprocessing operations are performed on each data stream, mainly including: using environmental compensation parameters to calibrate sensor sensitivity, and normalizing monitoring data with different ranges or different response characteristics to reduce the interference of environmental noise on feature extraction.

[0068] d. Feature Separation: For the monitoring data of each sensor, parametric gradient field analysis is used to separate micro-pollution targets. Since sampling points belonging to the same pollution source diffusion path have a consistent diffusion trend in the same data frame, a preset 3×3 spatial convolution kernel is first used to enhance the gradient response of the pollution evolution direction. Then, all sampling points are clustered based on the gradient vector direction to obtain sampling point groups with consistent directions. Each sampling point group corresponds to an independent micro-pollution feature region. Sampling points whose gradients remain at the background noise level in the data frame are determined to be clean background and are excluded.

[0069] At this point, the characteristic regions corresponding to each micro-pollution source in each data sequence have been obtained. Further analysis is needed to match pollution between different monitoring dimensions.

[0070] For any given monitoring node, since the sensor placement is fixed, the airflow direction and pollution diffusion perspective detected by that node are basically stable. For example, if the material transport channel is a unidirectional flow environment from south to north and the sensor is installed on the side wall of the channel, then the detected impurity transport is all lateral diffusion distribution, and the pollution evolution perspective sensed by the sensor is basically consistent. That is, the characteristic surface exposed by the micro-pollution source at each monitoring point can be determined.

[0071] The visual and physical composition of different diffusion interfaces of pollution sources varies, resulting in highly distinctive information distribution patterns. Therefore, from the perspective of different monitoring nodes, these distribution differences can be used to identify different diffusion surfaces of micro-pollution sources.

[0072] First, extreme value detection is performed on each micro-pollution feature region in each frame of data, and the obtained response points are defined as monitoring feature points of that region. Existing gradient extreme value extraction techniques can be used for feature point detection. Since there are significant parametric gradient fluctuations at the boundary between the pollution blobs and the clean airflow, the richer the physical information and concentration gradient changes contained in the diffusion surface, the more monitoring feature points are usually detected. Subsequently, based on the spatial distribution of feature points and the differences in their neighboring values, the complexity of their information content is assessed and quantified as an information contribution coefficient. This coefficient reflects the complexity and uniqueness of the pollution structure near the feature point. The calculation principle is: the closer the physical distance between a monitoring feature point and its neighboring feature points, and the greater the difference in the monitoring content of their neighborhood, the richer its information content is considered.

[0073] The process of obtaining the information contribution coefficient will be described in detail in the second embodiment, and will not be repeated here.

[0074] To address the pollution matching bias caused by local airflow disturbances in multi-dimensional monitoring, this invention does not directly rely on the easily disturbed overall feature vector. Instead, it utilizes the distribution characteristics of polluted surface information for matching to generate path segments for each monitoring dimension. The specific implementation is as follows:

[0075] Since the same pollution source may exhibit different diffusion surfaces at different monitoring times, and the information distribution patterns of each surface vary significantly, to achieve accurate cross-time matching, the micro-pollution characteristic regions are first segmented. Segmentation is based on the distribution of monitoring feature points and the differences in their information content (information contribution coefficient), thereby dividing the polluted area into several controlled region blocks that approximate its different diffusion surfaces (such as the front, core, and wake). This strategy aims to ensure that subsequent matching can target the local characteristics of the same pollution interface, thereby improving the robustness of matching under complex airflow conditions.

[0076] The process of dividing the controlled region into controlled region blocks will be described in detail in the third embodiment, and will not be repeated here.

[0077] S102. Locate the target controlled area block corresponding to the initial micro-pollution disturbance at the matching time, and execute the matching process between any target controlled area block at the matching time and each area block to be matched at the matching time. The matching process includes calculating the pollution matching degree based on the feature information vector and spatial location rationality coefficient of the target controlled area block and the area block to be matched.

[0078] The matching time, as the name suggests, is a sampling time used for matching. First, it is necessary to determine which pollution patch in the data stream is the target pollution source, i.e., the pollution source whose diffusion trajectory needs to be acquired. The matching time should display the diffusion surface information of the target pollution source as much as possible. The initial matching time and the target controlled region block that appears therein can be preset by the system or identified by the algorithm.

[0079] The time to be matched is a data frame time point used to match the matching time.

[0080] The process of selecting the time to be matched includes:

[0081] When the matching time is a non-terminal frame in its data sequence, the time to be matched is the next sampling time adjacent to the matching time;

[0082] When the matching time is the end frame of its data sequence, the time to be matched also includes the start time of any unmatched sequence.

[0083] Specifically, when matching within the same monitoring sequence, if the matching time is a non-terminating frame in its sequence, the time to be matched can be the adjacent preceding and following sampling times. After determining the matching time, the matching process is bidirectional. When the time to be matched fails to match the matching time, the time to be matched becomes an unmatched frame. Next, the adjacent times of the unmatched frame are then matched with the matching time. When the time to be matched successfully matches the matching time, the time to be matched becomes a matched time, and the matching process repeats. If every frame after a certain matching time in a sequence fails to match, then the last matching time of that sequence becomes the terminating frame.

[0084] When performing cross-monitoring matching, that is, when the matching time is the end frame of its own sequence, the time to be matched also includes the start time of any unmatched sequence. In other words, when the matching time is the end frame of its own sequence, the matching time can not only be matched with adjacent times within the same sequence, but also with the start time of any unmatched sequence across dimensions.

[0085] When any of the target controlled region blocks at the matching time successfully matches a region block to be matched at the matching time, it can be known that the remaining region blocks of the feature region to which the matching time belongs also become the new target controlled region blocks.

[0086] Each controlled region block corresponds to a feature information vector. The direction of the feature information vector is determined by the spatial gradient direction of the information contribution coefficients of all feature points within the controlled region block, while its magnitude is determined by the average information contribution coefficient of all feature points within the controlled region block.

[0087] Since the surface of the same pollution source may appear within the sensing range of multiple monitoring nodes, cross-node tracking of the pollution source can be achieved by matching corresponding controlled area blocks in different monitoring sequences. The specific implementation involves the following two steps:

[0088] Step 1: Path construction within a single monitoring dimension;

[0089] For any controlled region block, a continuous diffusion path of that controlled region block within a single monitoring node can be constructed by tracing its position sequence in the time step sequence. Considering the continuity of impurity evolution between adjacent time points, the paths of the controlled region blocks in the same sequence can be chained together by calculating the similarity between the position of the controlled region block (represented by the geometric center of the feature points within the block) and its feature information vector, i.e., the contamination matching degree.

[0090] When the matching time is a non-terminating frame in its sequence, the process of obtaining the contamination matching degree will be described in detail in the fifth embodiment, and will not be repeated here.

[0091] Step 2: Path association across monitoring dimensions.

[0092] For different monitoring nodes, due to differences in physical coordinates and dimensional characteristics, it is impossible to directly determine whether a region belongs to the same pollution source based on its spatial location. Therefore, it is necessary to introduce spatiotemporal rationality constraints for comprehensive judgment. That is, the matched controlled region should not only conform to the logical sequence of impurity diffusion in terms of response time, but its diffusion path should also have physical continuity in the three-dimensional coordinate system of the preparation workshop.

[0093] When the matching time is the termination frame in its sequence, the process of obtaining the pollution matching degree will be described in detail in the sixth embodiment, and will not be repeated here.

[0094] S103. If the maximum pollution matching degree is greater than the preset matching threshold, the corresponding area block to be matched is included in the continuous diffusion path sequence of the micro-pollution source and used as the new matching time. The matching process is repeated until all frames are matched.

[0095] The preset matching threshold can be set independently according to actual working conditions, preferably 0.65.

[0096] A successful match is defined as a pollution matching degree value greater than a preset matching threshold. The area block to be matched corresponding to the pollution matching degree value value is the target controlled area block corresponding to the micro-pollution source.

[0097] Regardless of whether the matching is successful or not, the data frames of each monitoring dimension need to be matched to obtain the path sequence of the target controlled area block under all sensors. The path sequence of a target controlled area block represents the evolution state of a certain local feature of the corresponding micro-pollution source throughout the entire production cycle.

[0098] S104. Integrate the continuous diffusion path sequence and map it to a real three-dimensional geographic coordinate system to generate the dynamic diffusion trajectory of the micro-pollution source.

[0099] Based on step S103, a continuous diffusion path sequence for each target controlled area block was obtained. According to the prior deployment coordinates of the monitoring sensors, these paths can be transformed from sensor coordinates to actual three-dimensional physical coordinates, i.e., the real geographic coordinate system. For the same micro-pollution source, the trajectories of different controlled areas may be interrupted due to local airflow shielding, but the paths of each block are physically complementary. By fusing these multiple area block paths belonging to the same pollution source, the shielded path segments can be completed, ultimately forming a complete and continuous dynamic diffusion trajectory for the micro-pollution source. Thus, the tracking of micro-pollution trajectories in high-purity material preparation based on real-time data streams was achieved, and further output signals were used to control actuators (such as isolation valves, FFU rotation speed, etc.) to implement blocking.

[0100] Figure 2 The flowchart below shows the process for obtaining the information contribution coefficient according to the second embodiment of the present invention. The process for obtaining the information contribution coefficient includes:

[0101] S201. Determine the reference neighborhood feature points of the current monitoring feature point, wherein the current monitoring feature point and the reference neighborhood feature points together form a reference region.

[0102] The reference neighborhood feature points can be the k other monitored feature points that are spatially closest to the current feature point for comparison. The number of k is preferably 8.

[0103] S202. Obtain the spatial physical distance between the current monitored feature point and any of the reference neighbor feature points, the difference in monitored values, and the maximum spatial scale within the reference area.

[0104] D f,m,r,i This represents the spatial physical distance between the r-th feature point and the i-th reference point within the m-th feature region at the f-th sampling time.

[0105] V f,m,r,i This represents the difference in monitoring values ​​between the r-th feature point and the i-th reference point within the m-th feature region at the f-th sampling time.

[0106] L max,f,m This represents the maximum distance between any two feature points within the m-th feature region at the f-th sampling time.

[0107] S203. Based on the spatial physical distance, the difference in the monitored values, and the maximum spatial scale, obtain the information contribution coefficient of the current monitored feature point.

[0108] The information contribution coefficient can be expressed by the formula:

[0109]

[0110] Wherein, the C f,m,r The ε represents the information contribution coefficient of the r-th feature point in the m-th feature region at the f-th sampling time, and the ε represents a preset minimum constant.

[0111] Similarly, the information contribution coefficient of each feature point can be obtained, which represents the reliability and uniqueness of each sampling point as the basis for locating pollution sources.

[0112] Figure 3 The flowchart below shows the process of dividing the controlled region block according to the third embodiment of the present invention. The process of dividing the controlled region block includes:

[0113] S301. Construct a local monitoring point set based on the monitored feature points. The construction process of the local monitoring point set includes: for the feature region that does not contain the assigned feature points, assign the feature points corresponding to the maximum and second largest distance values ​​from the geometric center of the region to the local monitoring point set; calculate the rationality of adding each unassigned feature point to the local monitoring point set; if the value of the rationality of the division is greater than a preset rationality threshold, then the unassigned feature points are assigned to the local monitoring point set.

[0114] The geometric center of the region is determined by the mean coordinates of the unassigned feature points. Taking the m-th feature region at the f-th sampling time as an example, a set of points with similar information representation within the region is constructed, and the specific steps are as follows:

[0115] The local monitoring point set is constructed starting from the feature points farthest and second farthest from the geometric center in the m-th feature region. Feature points belonging to the same point set are approximately considered to be response points of the same diffusion interface on the pollution source.

[0116] To determine whether it is reasonable to add new feature points to the local monitoring point set, the following conditions must be met: the spatial distribution is compact and the information consistency is high, and the envelope of the point set after addition still conforms to the regularity of the impurity diffusion front.

[0117] The process of obtaining the rationality of the division will be described in detail in the fourth embodiment, and will not be repeated here.

[0118] The reasonableness threshold can be set independently according to the actual situation, preferably 0.7.

[0119] S302. When there are still unassigned feature points after the construction process ends, repeat the construction process to construct a new set of local monitoring points until each feature point has its unique local monitoring point set to which it belongs, and one set of local monitoring points corresponds to one controlled region block.

[0120] The feature region is further divided into several controlled region blocks, which physically represent different diffusion gradient surfaces perceived by the sensor. Even when the local concentration fails due to environmental disturbances, accurate matching results can still be obtained based on other blocks.

[0121] Figure 4 This is a flowchart of the process for obtaining the rationality of the partitioning provided in the fourth embodiment of the present invention. The process for obtaining the rationality of the partitioning includes:

[0122] S401. Obtain the spatial volume enclosed by the outer contour line after the unassigned feature points are assigned into the local monitoring point set, and obtain the geometric compactness based on the spatial volume and its minimum circumscribed envelope volume.

[0123] U m,r It represents the volume enclosed by the outer contour line after the r-th feature point in the m-th feature region is added to the local monitoring point set.

[0124] U env,m,r This represents the volume of the corresponding minimum circumscribed envelope. The geometric compactness can be expressed as:

[0125]

[0126] The value of geometric compactness G m,rThe larger the value, the more regular the outline, and the more reasonable the inclusion.

[0127] S402. Obtain the average difference between the information contribution coefficient and the average information contribution of each monitoring feature point in the local monitoring point set, and the single-point difference between the information contribution coefficient and the average information contribution of the unassigned feature points, and obtain information consistency based on the average difference and the single-point difference.

[0128] The average difference can be expressed as:

[0129]

[0130] Wherein, the C i The coefficient representing the information contribution of the original feature points within the point set. This represents the average value of the information contribution coefficients within the point set.

[0131] The single-point difference can be expressed as:

[0132]

[0133] Wherein, the C r This represents the information contribution coefficient of the feature points to be added.

[0134] The consistency of the information can be represented as:

[0135]

[0136] S403. Determine the rationality of the partitioning based on the geometric compactness and the information consistency.

[0137] The rationality of the division can be expressed by the following formula:

[0138]

[0139] Wherein, α and β are preset weighting coefficients.

[0140] Figure 5 The flowchart below shows the process for obtaining the contamination matching degree according to the fifth embodiment of the present invention. The process for obtaining the contamination matching degree includes:

[0141] S501. When the matching time is a non-terminating frame in its sequence, obtain the cosine similarity S between the feature information vectors of the target controlled region block g at the matching time and any controlled region block g' in the adjacent matching time. cos .

[0142] S502, Calculate the physical distance D between the center of the target controlled region block and the center of the region block to be matched.g,g' Equivalent diameter W of the target controlled region block g The inverse proportional function value of the ratio is used as the first position reasonable coefficient P1.

[0143] The rationality coefficient of the first position can be expressed by the formula:

[0144]

[0145] S503. Based on the cosine similarity and the first position reasonable coefficient, the pollution matching degree M=Scos\P1 is obtained.

[0146] Figure 6 The flowchart below shows another process for obtaining the contamination matching degree provided in the sixth embodiment of the present invention. The process for obtaining the matching degree includes:

[0147] S601. When the matching time is the termination frame in its sequence, obtain the cosine similarity Scos between the feature information vectors of the target controlled region block and the controlled region block at the start time of the cross-dimensional unmatched sequence.

[0148] S602. Calculate the inverse proportional function value of the time deviation between the actual response time Tact and the estimated response time interval [Tmin, Tmax] of the micro-pollution source in different monitoring dimensions, and use it as the second position reasonable coefficient P2.

[0149] The estimated response time interval is defined by the path length L and the average diffusion velocity. calculate:

[0150]

[0151] S603. Based on the cosine similarity and the second position reasonable coefficient, the pollution matching degree M = Scos * P2 is obtained.

[0152] Figure 7 This is a schematic diagram of a micro-pollution control system provided in the seventh embodiment of the present invention. The system includes:

[0153] The region block division module 100 is used to collect real-time environmental data streams, extract micro-pollution feature regions, calculate the information contribution coefficient of feature points, and divide controlled region blocks.

[0154] The matching calculation module 200 is used to perform the matching process between the target controlled region block and the region block to be matched at the matching time, and to calculate the pollution matching degree based on the feature vector and the location rationality coefficient;

[0155] The path tracing module 300 is used to include the block to be matched into a continuous diffusion path sequence when a match is successful, and to recursively perform the matching until it is completed.

[0156] The execution control module 400 is used to integrate the diffusion path mapping to the three-dimensional coordinate system, generate dynamic diffusion trajectory, and output control signals to the actuators (isolation valves, FFUs, negative pressure regulating valves, etc.).

[0157] In summary, this invention overcomes the problems of lagging and insufficient positioning accuracy of traditional monitoring methods by focusing on highly discriminative local feature blocks for path association, and significantly enhances the micro-pollution defense capability of high-end manufacturing environments.

[0158] The following section, using a specific application scenario in a high-purity semiconductor silicon wafer fabrication workshop, further supplements the present invention:

[0159] Step 1, Micro-contamination Feature Extraction and Synchronization, firstly, particle concentration and airflow parameters are collected in real time by monitoring sensors distributed at various process nodes in the preparation workshop. Due to differences in the physical location and sampling frequency of different sensors, the system uses a precision clock protocol to align the timestamps of each sensor, and the region block partitioning module 100 calls a temperature and humidity compensation model to perform nonlinear drift correction on the original data stream. Subsequently, the system uses a 3x3 spatial convolution kernel to perform gradient enhancement processing on each frame of data, identifying areas with drastic parameter fluctuations in the environment as micro-contamination feature regions. By extracting the spatial distribution and monitoring values ​​of each monitoring feature point within this region, the information contribution coefficient C is calculated. f,m,r This coefficient characterizes the information abundance of a specific sampling point on the pollution diffusion surface, thus providing a physical weighting basis for subsequent refined segmentation.

[0160] Step 2, during the dynamic division and representation of controlled area blocks, the area block division module 100 further deconstructs the micro-pollution feature regions based on the distance of feature points relative to the geometric center of the region. The system first selects the farthest and second farthest feature points as seed points to construct a local monitoring point set, and introduces a partitioning rationality R... m,r Iterative selection is performed. During this process, the system calculates the geometric compactness G. m,r (i.e., the ratio of actual volume to envelope volume) and information consistency I m,r This ensures that each controlled region block conforms to the regularity of the impurity diffusion front or core region in terms of physical morphology, and that the internal parametric gradients have a high degree of homology. Finally, each controlled region block is mapped to a feature information vector containing the diffusion evolution direction and average information intensity, thereby transforming the complex contamination clump into a quantifiable topological description unit.

[0161] Step 3: During recursive path matching across spatiotemporal dimensions, the matching calculation module 200 locates the target controlled area block at the matching time in response to the initial disturbance. When a micro-pollution source moves within the same monitoring sequence, the system selects the next adjacent frame as the matching time, calculates the cosine similarity (Scos) of the feature information vectors between the target block and the block to be matched, and combines it with the first positional rationality coefficient P1 determined by physical displacement to obtain the pollution matching degree M. If the micro-pollution source crosses the coverage of different monitoring sensors, the path tracking module 300 initiates cross-dimensional association, uses the estimated response time interval [Tmin, Tmax] to constrain the logic of cross-node matching, and calculates the second positional rationality coefficient P2. Through this recursive matching mechanism, the system strings together the scattered area block responses into a continuous diffusion path sequence spanning multiple monitoring nodes, effectively solving the tracking interruption problem caused by local airflow obstruction.

[0162] Step four, during the coordinated blocking of the trajectory mapping and actuator, the execution control module 400 maps the acquired continuous diffusion path sequence from the sensor's local coordinate system to the real three-dimensional geographic coordinate system of the preparation workshop, generating a dynamic diffusion trajectory of the micro-pollution source. When the generated trajectory shows that the pollution source is penetrating into the high-purity material packaging area, the execution control module 400 outputs control commands in real time based on the spatial coordinates of the trajectory. Specifically, the system blocks the pollution penetration path by adjusting the isolation valve on the material transfer channel, while simultaneously increasing the rotational frequency of the blower filter unit (FFU) above the diffusion trajectory to enhance the suppression of local clean airflow, and simultaneously opening the negative pressure regulating valve at the exhaust outlet. By changing the pressure gradient within the controlled area, the pollutants are guided to migrate directionally towards the exhaust outlet. Through this feedback control based on dynamic trajectory, the real-time location and physical isolation of the micro-pollution source are achieved, ensuring the stability of the high-purity material preparation environment.

[0163] All content not described in detail in this specification belongs to the prior art known to those skilled in the art. Electrical control components not mentioned in this technical solution are also prior art and will not be described further here. The above descriptions are merely preferred embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling micro-contamination in the preparation process of high-purity materials, characterized in that, Includes the following steps: The real-time environmental data stream collected by each monitoring sensor in the preparation area is obtained, the micro-pollution feature regions in the data frames at each sampling time are extracted, the information contribution coefficient of each monitoring feature point in each micro-pollution feature region is calculated, and each micro-pollution feature region is divided into an independent controlled region block. At the matching moment when the initial micro-pollution disturbance is detected, the corresponding target controlled area block is located, and the matching process between any target controlled area block at the matching moment and each area block to be matched at the matching moment is executed. The matching process includes calculating the pollution matching degree based on the feature information vectors and spatial location rationality coefficients of the target controlled area block and the area blocks to be matched. If the maximum value of the pollution matching degree is greater than the preset matching threshold, the corresponding area block to be matched is included in the continuous diffusion path sequence of the micro-pollution source and used as a new matching reference frame. The matching process is repeated until all data frames in the preparation cycle have completed the matching analysis. The continuous diffusion path sequence is integrated and mapped to the real three-dimensional geographic coordinate system of the preparation workshop to generate the dynamic diffusion trajectory of the micro-pollution source. The corresponding actuator is then controlled to perform blocking actions based on the dynamic diffusion trajectory.

2. The method for controlling micro-contamination in the high-purity material preparation process according to claim 1, characterized in that, Before acquiring the real-time environmental data streams collected by each monitoring sensor within the preparation area, the method further includes: The sampling clocks of each monitoring sensor distributed within the preparation area are calibrated using a network synchronization protocol, so that the data streams collected by each monitoring sensor have a timestamp with a unified time reference. The real-time environmental data stream is preprocessed, including using a preset compensation algorithm to eliminate nonlinear deviations caused by sensor drift due to environmental temperature and humidity, and standardizing and normalizing monitoring data of different dimensions.

3. The method for controlling micro-contamination in the high-purity material preparation process according to claim 1, characterized in that, The micro-pollution feature regions in the real-time environmental data streams collected by each monitoring sensor within the preparation area include: The gradient change field of environmental parameters between adjacent sampling periods is obtained, the gradient change field is optimized, and the monitoring feature points at each sampling time are clustered based on the direction of the gradient vector to obtain feature point groups with consistent gradient direction. Each feature point group is defined as a micro-pollution feature region.

4. The method for controlling micro-contamination in the high-purity material preparation process according to claim 1, characterized in that, The process of obtaining the information contribution coefficient includes: Determine the reference neighborhood feature points of the current monitoring feature point, and the current monitoring feature point and the reference neighborhood feature points together form a multidimensional parameter analysis region; Obtain the spatial physical distance between the current monitored feature point and any of the reference neighbor feature points, the difference in monitored values, and the maximum spatial scale within the multidimensional parameter analysis area; Based on the spatial physical distance, the difference in the monitored values, and the maximum spatial scale, the information contribution coefficient of the current monitored feature point is obtained through a preset contribution function.

5. The method for controlling micro-contamination in the high-purity material preparation process according to claim 1, characterized in that, The process of dividing the controlled region block includes: A local monitoring point set is constructed based on the monitoring feature points. For the micro-pollution feature area that does not contain the assigned feature points, the feature points with the largest and second largest distances from the geometric center of the area are assigned to the local monitoring point set. Calculate the rationality of adding each unassigned feature point to the local monitoring point set. If the value of the rationality of the partition is greater than a preset rationality threshold, then assign the unassigned feature point to the local monitoring point set. When the construction process ends and there are still unassigned feature points, the construction process is repeated to construct a new set of local monitoring points until all monitoring feature points have a unique local monitoring point set to which they belong, and a set of local monitoring points is defined as a controlled region block.

6. The method for controlling micro-contamination in the high-purity material preparation process according to claim 5, characterized in that, The process for obtaining the rationality of the division includes: After the unassigned feature points are assigned to the local monitoring point set, the multidimensional space volume enclosed by their outer contour lines is obtained, and the geometric compactness is obtained based on the ratio of the space volume to the volume of its smallest circumscribed envelope. The average difference between the information contribution coefficient and the average information contribution of each monitoring feature point in the local monitoring point set is obtained, as well as the single-point difference between the information contribution coefficient and the average information contribution of the unassigned feature points. Information consistency is obtained based on the ratio of the average difference to the single-point difference. The rationality of the partition is obtained based on the weighted summation result of the geometric compactness and the information consistency.

7. The method for controlling micro-contamination in the high-purity material preparation process according to claim 1, characterized in that, The process of selecting the time to be matched includes: When the matching time is a non-terminal frame in its data sequence, the time to be matched is the next sampling time adjacent to the matching time, which is incremented by the time step. When the matching time is the termination frame in its data sequence, the time to be matched includes the starting sampling time of any unmatched sequence in other associated monitoring dimensions.

8. The method for controlling micro-contamination in the high-purity material preparation process according to claim 1, characterized in that, The process of obtaining the contamination matching degree includes: Obtain the cosine similarity of the feature information vectors between any target controlled region block at the matching time and any controlled region block at the time to be matched; If the matching time is a non-terminating frame in its data sequence, calculate the spatial distance between the physical center of the target controlled region block and the physical center of any controlled region block in an adjacent matching time. Define the inverse proportional function value of the ratio of the spatial distance to the equivalent diameter of the target controlled region block as the first position reasonable coefficient. Calculate the contamination matching degree based on the product of the cosine similarity and the first position reasonable coefficient. If the matching time is the termination frame in its data sequence, calculate the inverse proportional function value of the time deviation between the actual response time of the micro-pollution source in the unmatched sequence and the response time interval predicted based on the diffusion model, and define it as the second position reasonable coefficient. Calculate the pollution matching degree based on the product of the cosine similarity and the second position reasonable coefficient.

9. A micro-contamination control system in the preparation process of high-purity materials, used to execute the method according to any one of claims 1 to 8, characterized in that, include: The region block division module (100) is configured to acquire the real-time environmental data stream collected by each monitoring sensor in the preparation area, extract the micro-pollution feature area in the data frame at each sampling time, calculate the information contribution coefficient of each monitoring feature point in each micro-pollution feature area, and then divide each micro-pollution feature area into an independent controlled region block. The matching calculation module (200) is configured to locate the corresponding target controlled area block at the matching time when the initial micro-pollution disturbance is detected, and to execute the matching process between any target controlled area block at the matching time and each area block to be matched at the matching time. The matching process includes calculating the pollution matching degree based on the feature information vector and spatial location rationality coefficient of the target controlled area block and the area block to be matched. The path tracking module (300) is configured to include the corresponding area block to be matched into the continuous diffusion path sequence of the micro-pollution source if the maximum value of the pollution matching degree is greater than the preset matching threshold, and use it as a new matching reference frame, and repeatedly call the matching calculation module (200) until all data frames are matched. The execution control module (400) is configured to integrate the continuous diffusion path sequence and map it to a real geographic coordinate system to generate a dynamic diffusion trajectory, and output control signals to the execution mechanism according to the coordinates of the dynamic diffusion trajectory.

10. The micro-contamination control system in the high-purity material preparation process according to claim 9, characterized in that: The actuator includes an isolation valve installed on the material transport channel, a fan filter unit installed on the top of the preparation chamber, and a negative pressure regulating valve installed at the exhaust outlet; the execution control module (400) changes the airflow direction and pressure gradient in the area where the dynamic diffusion trajectory is located by adjusting the opening and closing state of the isolation valve, the rotational speed frequency of the fan filter unit, and the valve opening degree of the negative pressure regulating valve.

Images