Semiconductor manufacturing gaseous pollutant intelligent monitoring and early warning system and method

Abstract

The invention discloses an intelligent monitoring and early warning system and method for gaseous pollutants in semiconductor manufacturing, and relates to the technical field of semiconductor manufacturing management.The intelligent monitoring and early warning system comprises a window screening module, a curve construction module and a real-time early warning module; calculating an exclusive differential pressure attenuation period, and meanwhile, locking a window with the highest risk score through dynamic time window division; the curve construction module disturbs phase alignment through pressure difference, divides effective disturbance sections and synthesizes a window total pressure difference curve; and the real-time early warning module calculates the disturbance accumulated impulse of the window total pressure difference curve, and performs early warning linkage by combining with the predicted pollutant permeation amount. According to the method, the pressure difference superposition risk of multiple door opening can be accurately identified, the accuracy and pertinence of monitoring and early warning are improved, and gaseous pollutant permeation is effectively prevented and controlled.

Description

Technical Field

[0001] This invention relates to the field of semiconductor manufacturing management technology, specifically to an intelligent monitoring and early warning system and method for gaseous pollutants in semiconductor manufacturing. Background Technology

[0002] In semiconductor manufacturing, the control of gaseous contaminants in cleanrooms is directly related to product yield. Door opening is a key factor disrupting the regional pressure gradient and causing contaminant infiltration. Existing intelligent monitoring and early warning technologies for gaseous contaminants have the following shortcomings: First, it cannot correlate key parameters such as the door opening angle and the degree of gap opening, making it impossible to accurately quantify the intensity of pressure difference disturbance caused by a single door opening, resulting in a one-sided disturbance assessment. Secondly, the lack of a dynamic time window division mechanism based on the pressure difference decay period makes it difficult to identify the pressure difference superposition effect of multiple door openings, and it is easy to miss high-risk working conditions caused by high-frequency door opening. Third, the estimated pollutant infiltration volume is not deeply correlated with the pressure differential impulse and the actual gap condition, resulting in low early warning efficiency; Therefore, there is an urgent need for intelligent monitoring and early warning systems and methods for gaseous pollutants in semiconductor manufacturing. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an intelligent monitoring and early warning system and method for gaseous pollutants in semiconductor manufacturing, which solves the problems of limited monitoring, crude early warning, and inability to accurately identify the risks of differential pressure superposition.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an intelligent monitoring and early warning system for gaseous pollutants in semiconductor manufacturing, comprising: The window filtering module collects the real-time differential pressure curve and gap opening degree of the same door in the cleanroom for each door opening event, calculates the exclusive differential pressure decay period of the door opening event, and sets a fixed monitoring period and opening number threshold. When the number of door opening events in the fixed monitoring period exceeds the corresponding opening number threshold, the module dynamically divides all door opening events in the fixed monitoring period into time windows, calculates the risk score of each window, and selects the window with the highest risk score as the target window. The curve construction module extracts the independent differential pressure curves of all door opening events within the target window, corrects the timing deviation of each independent differential pressure curve by using differential pressure disturbance phase alignment, divides the effective disturbance segment from the corresponding differential pressure curve by the exclusive differential pressure decay period of each door opening event, and superimposes the effective disturbance segments of all door opening events according to the same phase fluctuation to synthesize the total differential pressure curve of the window. The real-time early warning module calculates the cumulative impulse of the disturbance of the total pressure difference curve of the window, and predicts the amount of pollutant penetration by combining the pollutant penetration coefficient of the semiconductor process in the corresponding region. It then executes early warning linkage according to the pollutant penetration level.

[0005] As a further aspect of the present invention, the specific operation for calculating the pressure differential attenuation period specific to the door opening event is as follows: From the collected real-time differential pressure curves, the peak point of differential pressure for this door opening event is located, and the baseline of differential pressure in this area under the condition of no door opening and no equipment disturbance is taken as the steady-state differential pressure. Record the time it takes for the differential pressure to drop from the peak point back to the steady-state differential pressure, and use this time as the basic differential pressure decay period for this opening event; The maximum opening degree of the gap and the duration of the maximum value were counted in this door opening event, and the product of the two was used as the gap characteristic parameter. The reference parameters of the door under standard gap conditions are retrieved, the ratio of the gap characteristic parameters of this opening to the reference parameters is calculated, and the ratio is multiplied by the basic pressure difference decay period to obtain the specific pressure difference decay period corresponding to this opening event. The reference parameters represent the product of the maximum gap opening degree and the duration when the door body is unworn and the sealing strip is intact.

[0006] As a further aspect of the present invention, the specific steps for dynamically dividing all door opening events within a fixed monitoring period into time windows are as follows: Extract all door opening events within a fixed monitoring period in chronological order, and retrieve the dedicated differential pressure attenuation period corresponding to each door opening event to establish an association list between the event occurrence sequence and the dedicated differential pressure attenuation period. Based on the association list, if the time interval between two adjacent door opening events is less than the pressure differential decay period of the previous door opening event, it is determined that there is a pressure differential superposition effect between the two door opening events; if the time interval between two adjacent door opening events is greater than or equal to the pressure differential decay period of the previous door opening event, it is determined that there is no pressure differential superposition effect between the two events. Adjacent door-opening events with overlapping effects are grouped into the same window, while door-opening events without overlapping effects are grouped into a new window, until all door-opening events within the fixed monitoring period are covered.

[0007] As a further aspect of the present invention, the specific operation for calculating the window risk score is as follows: Extract the maximum opening angle, maximum gap opening degree, and event interval difference for each door opening event. If the current door opening event is the last door opening event within the window, the difference is taken as the dedicated attenuation period itself. Simultaneously, record the disturbance impact duration for each door opening event. The disturbance impact duration is selected from the dedicated differential pressure attenuation period and the event interval difference. Specifically, when the dedicated differential pressure attenuation period ≤ the event interval difference, the disturbance impact duration = the dedicated differential pressure attenuation period; when the dedicated differential pressure attenuation period > the event interval difference, the disturbance impact duration = the event interval difference. The event interval difference is the difference between the dedicated differential pressure attenuation period of this door opening event and the time interval of the next adjacent door opening event. For each door opening event, the maximum opening angle is multiplied by the maximum gap opening degree to obtain the basic disturbance value, and then the basic disturbance value is divided by the disturbance influence duration to obtain the dynamic disturbance intensity value. The summation of the dynamic disturbance intensity values ​​of all door opening events within the window, multiplied by the window superposition coefficient, yields the final window risk score.

[0008] As a further aspect of the present invention, the specific method for obtaining the window overlay coefficient is as follows: Extract the timing data of all door opening events within the window and label them as E according to their order of occurrence. j And simultaneously retrieve the dedicated differential pressure decay period T of the previous n-1 door opening events. i Where j∈[1,n], i∈[1,n-1], and n is the number of door opening events within the window; Calculate the superposition density factor f for each group of adjacent door opening events. i Its specific expression is: f i =T i / △t i , where △t i For the door opening event E i With E i+1 The time interval; When the number of door opening events within the window is n=1, the window overlay coefficient is... When the number of door opening events n≥2 within the window, the window stacking coefficient is... .

[0009] As a further aspect of the present invention, the timing deviation of each independent differential pressure curve is corrected by phase alignment of differential pressure disturbance, specifically including: Based on the steady-state pressure difference Ps of each independent pressure difference curve, the disturbance initiation characteristic point of the curve is located. The time axis of each independent differential pressure curve is shifted so that the disturbance initiation characteristic point of all differential pressure curves corresponds to the same time zero point. The specific operation for locating the initiation feature point of the disturbance on the independent differential pressure curve is as follows: On the independent differential pressure curve, select the first sampling point that simultaneously meets the following two conditions as the initiation feature point of the disturbance: Condition 1: The differential pressure sample value at this sampling point is greater than Ps, and the differential pressure sample values ​​at the next three consecutive sampling points are all greater than Ps. Condition 2: Calculate the pressure difference between the sampling point and the previous adjacent sampling point, and divide it by the time interval between the two sampling points to obtain the first derivative of the sampling point. The first derivative value must be greater than 0.

[0010] As a further aspect of the present invention, the specific operation of dividing the effective disturbance segment is as follows: retrieve the exclusive differential pressure decay period corresponding to the door opening event for each independent differential pressure curve, and take the disturbance start feature point of the differential pressure curve as the starting point, and delineate the curve segment with a duration equal to the exclusive differential pressure decay period as the effective disturbance segment.

[0011] As a further aspect of the present invention, the specific steps for calculating the cumulative impulse of the disturbance on the total pressure difference curve of the window are as follows: The total pressure difference curve is divided into two stages: a disturbance rise segment and a disturbance decay segment. The pressure difference deviation at each sampling moment within each stage is calculated, representing the difference between the pressure difference at that moment and Ps. The disturbance rise segment is the period from the steady-state pressure difference Ps to the total peak pressure P. max The curve segment, where the disturbance attenuation segment is the total pressure difference from P max The curve segment that falls back to Ps; Integrate the pressure difference deviations of the rising segment and the decay segment of the disturbance with respect to time to obtain the impulse of the rising segment and the impulse of the decay segment. Add the two together to obtain the cumulative impulse of the disturbance on the total pressure difference curve of the window.

[0012] As a further aspect of the present invention, the specific steps for predicting pollutant infiltration are as follows: The basic pollutant permeability coefficient of the corresponding cleanroom area is retrieved, and the maximum value of the gap opening degree of all door opening events in the target window is extracted. The average value is used as the window gap characteristic parameter. The basic pollutant permeability coefficient refers to the amount of pollutant permeating through a unit area gap under a unit pressure difference impulse. The cumulative impulse of the disturbance is multiplied by the basic pollutant permeability coefficient to obtain the theoretical permeability under the standard gap condition. The theoretical permeability is then multiplied by the window gap characteristic parameters to predict the pollutant permeability.

[0013] Intelligent monitoring and early warning methods for gaseous pollutants in semiconductor manufacturing include: Step 1: For each door opening event of the same door in the cleanroom, collect the real-time differential pressure curve and gap opening degree of the door, calculate the exclusive differential pressure decay period of the door opening event, and set a fixed monitoring period and opening number threshold. When the number of door opening events in the fixed monitoring period exceeds the corresponding opening number threshold, divide all door opening events in the fixed monitoring period into dynamic time windows, calculate the risk score of each window, and take the window with the highest risk score as the target window. Step 2: Extract the independent differential pressure curves of all door opening events within the target window, use differential pressure disturbance phase alignment to correct the timing deviation of each independent differential pressure curve, and divide the effective disturbance segment from the corresponding differential pressure curve through the exclusive differential pressure decay period of each door opening event. Then, combine the effective disturbance segments of all door opening events in the same phase to synthesize the total differential pressure curve of the window. Step 3: Calculate the cumulative impulse of the disturbance of the total pressure difference curve of the window, and predict the amount of pollutant penetration by combining the pollutant penetration coefficient of the semiconductor process in the corresponding region. Execute early warning linkage according to the pollutant penetration level.

[0014] This invention provides an intelligent monitoring and early warning system and method for gaseous pollutants in semiconductor manufacturing. Compared with the prior art, it has the following advantages: (1) This invention collects real-time differential pressure curves and gap opening degree, uses gap opening degree to calibrate a dedicated differential pressure attenuation cycle, and then performs dynamic time window division and risk score calculation based on this cycle to accurately identify the differential pressure superposition effect of multiple door openings, thus solving the problem that existing technologies cannot easily lock high-risk working conditions. (2) The present invention uses pressure difference disturbance phase alignment to correct timing deviation, combines a dedicated pressure difference attenuation period to divide the effective disturbance segment, and synthesizes the total pressure difference curve by superimposing the same phase in time unit, which fully preserves the peak value and duration characteristics of the superimposed disturbance, ensuring that the total pressure difference curve can truly reflect the cumulative disturbance intensity. (3) This invention calculates the cumulative impulse of disturbance by segmented integration, and estimates the amount of pollutant infiltration by combining the basic permeability coefficient and the characteristic parameters of the window gap. It implements early warning linkage according to the permeability level, realizing closed-loop control from disturbance quantification to early warning execution, avoiding over-warning or under-warning caused by experience threshold. Attached Figure Description

[0015] Figure 1 This is the system principle block diagram of the present invention; Figure 2 This is a flowchart illustrating the steps involved in calculating the window risk score in this invention. Figure 3 This is a flowchart of the steps of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] like Figure 1 This invention provides an intelligent monitoring and early warning system for gaseous pollutants in semiconductor manufacturing; As an embodiment of this application, it includes: The window filtering module collects the real-time differential pressure curve and gap opening degree of the same door in the cleanroom for each door opening event, calculates the exclusive differential pressure decay period of the door opening event, and sets a fixed monitoring period and opening number threshold. When the number of door opening events in the fixed monitoring period exceeds the corresponding opening number threshold, the module dynamically divides all door opening events in the fixed monitoring period into time windows, calculates the risk score of each window, and selects the window with the highest risk score as the target window. The curve construction module extracts the independent differential pressure curves of all door opening events within the target window, corrects the timing deviation of each independent differential pressure curve by using differential pressure disturbance phase alignment, divides the effective disturbance segment from the corresponding differential pressure curve by the exclusive differential pressure decay period of each door opening event, and superimposes the effective disturbance segments of all door opening events according to the same phase fluctuation to synthesize the total differential pressure curve of the window. The real-time early warning module calculates the cumulative impulse of the disturbance of the total pressure difference curve of the window, and predicts the amount of pollutant penetration by combining the pollutant penetration coefficient of the semiconductor process in the corresponding region. It then executes early warning linkage according to the pollutant penetration level.

[0018] As a second embodiment of this application, it is implemented based on the first embodiment, except that this embodiment includes: The window filtering module collects the real-time differential pressure curve and gap opening degree of the same door in the cleanroom for each opening event; Cleanrooms can prevent external pollutants from penetrating by maintaining a pressure gradient between the area and the outside world. Opening the door is the direct cause of disrupting this gradient. By collecting real-time pressure differential curves, the continuous changes in pressure differential can be dynamically recorded throughout the entire door opening process, which more intuitively reflects the occurrence, development and dissipation of disturbances. At the same time, when the door opening angle is the same, factors such as aging of the door sealing strip and wear of the hinges will cause the actual gap size to vary. Moreover, the gap is the only channel for airflow and pollutants to penetrate, and its dynamic changes directly determine the penetration efficiency of pollutants. The specific differential pressure decay period for the door opening event is calculated based on the above two parameters. The specific operation is as follows: From the collected real-time differential pressure curves, the peak point of differential pressure for this door opening event is located, and the baseline of differential pressure in this area under the condition of no door opening and no equipment disturbance is taken as the steady-state differential pressure. Record the time taken for the differential pressure to drop from its peak point back to the steady-state differential pressure, and use this time as the basic differential pressure decay period for this opening event; The maximum opening degree of the gap during this door opening event and the duration of the maximum value are statistically analyzed. The product of the two is used as the gap characteristic parameter, which directly characterizes the actual airflow infiltration channel characteristics during this door opening. Retrieve the reference parameters of the door under standard gap conditions, calculate the ratio of the gap characteristic parameters of this door opening to the reference parameters, multiply the ratio by the basic pressure difference decay period, and the result is the exclusive pressure difference decay period corresponding to this door opening event. The reference parameters represent the product of the maximum gap opening degree and the duration when the door body is unworn and the sealing strip is intact. The gap conditions vary under different door opening scenarios. For example, aging of the door sealing strip and wear of the hinges can cause the gap to increase. By correcting the standard parameter ratio, the cycle can be adapted to the actual sealing characteristics of a single door. At the same time, a fixed monitoring period and a threshold for the number of times a door is opened are set. When the number of door opening events within the fixed monitoring period exceeds the corresponding threshold for the number of times a door is opened, a dynamic time window is divided for all door opening events within the fixed monitoring period. Cleanroom differential pressure monitoring needs to be running in real time, but continuously and indiscriminately analyzing all door opening events would consume a lot of system computing power, and the differential pressure disturbance caused by a single door opening usually dissipates quickly. In addition, setting a fixed monitoring cycle can integrate discrete door opening events into fixed time periods for analysis, and setting a threshold for the number of openings can distinguish between normal door opening frequency and high-frequency door opening risks. In-depth analysis can be performed on the time periods of high-frequency operating conditions to avoid unnecessary consumption of computing power. The specific operation of dynamically dividing all door opening events within the fixed monitoring period into time windows is as follows: Extract all door opening events within the fixed monitoring period in chronological order, and retrieve the dedicated differential pressure attenuation period corresponding to each door opening event to establish a one-to-one correspondence list between the event occurrence sequence and the dedicated differential pressure attenuation period. The core basis for dynamic window division is whether the pressure difference disturbance of the previous door opening has completely dissipated, and the dedicated pressure difference decay period is an accurate parameter characterizing the duration of a single door opening disturbance. Establishing a correlation list can directly provide data support for subsequent judgments. Based on the associated list, the relationship between the time interval of two adjacent door opening events and the pressure differential decay period of the previous door opening event is determined one by one. The specific judgment rule is as follows: if the time interval of two adjacent door opening events is less than the pressure differential decay period of the previous door opening event, it is determined that there is a pressure differential superposition effect between the two door opening events; if the time interval of two adjacent door opening events is greater than or equal to the pressure differential decay period of the previous door opening event, it is determined that there is no pressure differential superposition effect between the two events. The superposition effect of pressure difference is the core cause of the sharp increase in the risk of regional pollutant infiltration. The premise for the superposition effect is that the pressure difference disturbance caused by the previous opening has not completely dissipated, and the next opening has already occurred. By directly comparing the interval with the dedicated pressure difference decay period, the existence of the superposition effect can be accurately determined. Adjacent door opening events with overlapping effects are grouped into the same window, while door opening events without overlapping effects are grouped into a new window, until all door opening events within the fixed monitoring period are covered. All door-opening events within the same window have a superposition effect of pressure difference, and their disturbance intensity on the regional steady-state pressure difference is much greater than that of a single door-opening event. Integrating these events into an evaluation unit can enable focused analysis of high-risk operating conditions.

[0019] The curve construction module extracts the independent differential pressure curves for all door opening events within the target window. It then uses differential pressure disturbance phase alignment to correct the timing deviations of each independent differential pressure curve. The specific operation is as follows: Based on the steady-state pressure difference Ps of each independent pressure difference curve, the disturbance initiation characteristic point of the curve is located, which corresponds to the moment when the door is opened until the gap opening degree reaches the critical value of airflow exchange. The specific steps involved in locating the initial feature point of the curve disturbance are as follows: On this independent differential pressure curve, locate the first sampling point that simultaneously meets the following two conditions; this point is the disturbance initiation characteristic point: Condition 1: The differential pressure sample value at this sampling point is greater than Ps, and the differential pressure sample values ​​at the next three consecutive sampling points are all greater than Ps, indicating that the differential pressure has been rising continuously from the steady-state reference value. Condition 2: Calculate the pressure difference between the sampling point and the previous adjacent sampling point, and divide it by the time interval between the two sampling points to obtain the first derivative of the sampling point. The first derivative value must be greater than 0. The action characteristics of different door opening events within the window are naturally different, which will lead to different disturbance start times, rising slopes, and peak occurrence times for each independent differential pressure curve. The key to in-phase superposition is to make the differential pressure disturbances caused by multiple door openings exhibit a cumulative effect in the same time dimension. Among them, the disturbance start feature point is the common source of disturbances for all differential pressure curves. This point is the only reference point that can accurately anchor the disturbance period of all curves. The time axis of each independent differential pressure curve is shifted so that the disturbance initiation characteristic point of all differential pressure curves corresponds to the same zero point in time; after the shift, the rising segment and the decay segment of each differential pressure curve are accurately correlated on the time axis. By using the exclusive differential pressure decay period for each door opening event, an effective disturbance segment is delineated from the corresponding differential pressure curve. The specific operation for delineating the effective disturbance segment is as follows: retrieve the exclusive differential pressure decay period for each door opening event corresponding to each differential pressure curve, take the disturbance start feature point of the differential pressure curve as the starting point, and delineate the curve segment with a duration equal to the exclusive differential pressure decay period as the effective disturbance segment. Starting from the disturbance initiation feature point, a curve segment with a duration equal to the corresponding differential pressure decay period of the differential pressure curve is delineated, which can accurately define the differential pressure disturbance range caused only by this door opening action; Meanwhile, by using the disturbance initiation feature point and the exclusive differential pressure decay period, each differential pressure curve can be divided into three parts: the steady-state section before the disturbance, the disturbance section caused by the door opening, and the redundant recovery section after the disturbance. Among them, the steady-state section has no disturbance information, and the differential pressure in the redundant recovery section is close to the steady state and is not related to pollutant infiltration. If the effective sections are not divided and directly superimposed, the intensity of the disturbance signal will be greatly diluted. The specific operation is as follows: The total pressure difference curve of the window is synthesized by superimposing the effective disturbance segments of all door-opening events with in-phase fluctuations. For the effective perturbation segment of all differential pressure curves, it is divided into several consecutive time units with the same time granularity, and each time unit corresponds to the same time interval of all differential pressure curves. For each time unit, calculate the instantaneous pressure difference values ​​of all effective disturbance segments within that time unit, and directly superimpose the instantaneous values ​​to obtain the total instantaneous pressure difference value of that time unit. Connect the total instantaneous pressure difference values ​​of all time units in chronological order to generate the window total pressure difference curve. The effective disturbance segment duration may vary for different door opening events within the window (due to the different differential pressure decay periods for each door opening event). If the effective disturbance segments are directly superimposed, there will be data loss or redundant misalignment issues in some time periods. Dividing continuous time units by the same fixed time granularity can force the instantaneous pressure difference data of all effective disturbance segments to be anchored in a completely consistent time interval, ensuring that the object of the superposition operation is the pressure difference disturbance value at the same moment.

[0020] The real-time early warning module calculates the cumulative impulse of disturbances on the total differential pressure curve within the window. The specific operation is as follows: The total pressure difference curve is divided into two stages: the disturbance rise segment and the disturbance decay segment. The disturbance rise segment is when the total pressure difference increases from the steady-state pressure difference Ps to the total peak value P. max The curve segment, where the disturbance attenuation segment is the total pressure difference from P maxThe curve segment that falls back to Ps is calculated, and the pressure difference deviation value at each sampling time in the two stages is calculated, that is, the difference between the pressure difference value at that time and Ps. The pressure difference disturbances in the rising and decaying stages of the disturbance have different mechanisms of action on pollutant infiltration. The rising stage is a process of active influx of external airflow, while the decaying stage is a process of passive recovery of regional airflow. Segmented calculation can distinguish the disturbance contribution of different stages and avoid the overall integration from masking the stage characteristics. Integrate the pressure difference deviations of the rising segment and the decay segment of the disturbance with respect to time to obtain the impulse of the rising segment and the impulse of the decay segment, and add the two together to obtain the cumulative impulse of the disturbance in the total pressure difference curve of the window. The amount of contaminant penetration is predicted by combining the contaminant penetration coefficient of the corresponding semiconductor process area. The specific operation is as follows: The basic pollutant permeability coefficient of the corresponding cleanroom area is retrieved. At the same time, the maximum value of the gap opening degree of all door opening events in the target window is extracted, and its average value is used as the window gap characteristic parameter. The basic pollutant permeability coefficient refers to the amount of pollutant permeating through a unit area gap under a unit pressure difference impulse. This value is determined by the regional airflow characteristics and pollutant type. Multiply the cumulative disturbance impulse by the basic pollutant permeability coefficient to obtain the theoretical permeability under standard gap conditions. Then multiply the theoretical permeability by the window gap characteristic parameters to predict the pollutant permeability. The core influencing factors of pollutant infiltration are pressure differential impulse, gap size, and regional infiltration characteristics. The first step is to calculate the theoretical infiltration under standard gap conditions. The second step is to correct the theoretical value with the actual gap level to obtain the final real infiltration. Early warning and linkage mechanisms are implemented based on the classification of pollutant infiltration levels, specifically including: Retrieve the contaminant tolerance limit Q of the corresponding semiconductor process. lim , will [0,Q lim The interval is divided into three consecutive and non-overlapping intervals, namely interval 1, interval 2, and interval 3, corresponding to [0, Q]. lim / 3]、(Q lim / 3,2Q lim / 3]、(2Q lim / 3,Q lim ]; The predicted pollutant infiltration rates were matched with three intervals, and different warnings were issued for each interval: If it falls within the first interval, a Level 1 warning will be issued. The specific measures are as follows: activate the positive pressure air curtain device around the door, with the air curtain airflow direction tilting outward along the door plane and the airflow rate maintained within the preset range. At the same time, open the micro negative pressure adsorption tank below the door to adsorb any trace pollutants that may have seeped in. The adsorption air volume is a times the area's normal exhaust air volume. Under low-risk conditions, pollutants are only located around the door and have not yet spread into the area. At this time, a positive pressure air curtain can be used to form an airflow barrier to prevent external pollutants from rushing in with the airflow when the door is opened. Meanwhile, a micro negative pressure adsorption tank can help remove residual trace pollutants. The two work together to accurately cut off the infiltration path. If it falls within interval two, a level two warning will be issued. The specific measures are as follows: increase the air volume of the regional exhaust system to b times the benchmark air volume, switch the exhaust path to the high-efficiency filter branch, activate the high-efficiency filter module at the exhaust end, and simultaneously adjust the angle of the airflow equalization plate in the area to concentrate the airflow direction towards the exhaust vent and shorten the residence time of pollutants in the area. At medium risk, some pollutants have already seeped into the area. Simply blocking the passage cannot eliminate the existing pollution. At this time, increasing the exhaust volume can increase the negative pressure gradient in the area, promote the convergence of pollutants towards the air outlet, and high-efficiency filtration can prevent pollutants from leaking out with the exhaust air and polluting other areas. If it falls within interval three, a level three warning will be issued. The specific measures are as follows: immediately switch the air intake system of the core process equipment from conventional fresh air intake to built-in inert gas reserve source supply, and close all exposed air intake ports of the equipment. At the same time, upgrade the filter module at the air intake end of the equipment from conventional high-efficiency level to ultra-high-efficiency level, and start the internal circulation purification device of the equipment to purify the inside of the equipment cavity in real time. In high-risk situations, pollutants have already seeped into the area in large quantities and may spread to the vicinity of core equipment. Therefore, using an inert gas reserve source can completely isolate the intake of external pollutants, while ultra-high efficiency filtration and internal circulation purification can remove pollutants that have seeped into the equipment, preventing pollutants from affecting semiconductor products from the source.

[0021] As a third embodiment of this application, this embodiment further discloses a method for calculating a window risk score based on embodiments one and two, such as... Figure 2 As shown, the specific content includes: For a single window, the maximum opening angle, maximum gap opening degree, and event interval difference are extracted for each door opening event. If the current door opening event is the last event within the window, the difference is taken as the dedicated attenuation period itself. Simultaneously, the disturbance impact duration for each door opening event is recorded. The disturbance impact duration is selected from the dedicated differential pressure attenuation period and the event interval difference. Specifically, when the dedicated differential pressure attenuation period ≤ the event interval difference, the disturbance impact duration = the dedicated differential pressure attenuation period; when the dedicated differential pressure attenuation period > the event interval difference, the disturbance impact duration = the event interval difference. The event interval difference is the difference between the dedicated differential pressure attenuation period of this door opening event and the time interval of the next adjacent door opening event. The opening angle refers to the angle formed between the door body and the plane of the door frame after the door body rotates around the hinge. It reflects the range of the door body opening. The larger the range, the more intense the exchange of air pressure between the area and the outside world. It is the core influencing factor of the pressure difference disturbance intensity. For each door opening event, the maximum opening angle is multiplied by the maximum gap opening degree to obtain the basic disturbance value, and then the basic disturbance value is divided by the disturbance influence duration to obtain the dynamic disturbance intensity value. The summation of the dynamic disturbance intensity values ​​of all door opening events within the window, multiplied by the window superposition coefficient, yields the final window risk score. The window overlay coefficient specifically includes: Extract the timing data of all door opening events within the window and label them as E1, E2, ..., E in the order of occurrence. n And simultaneously retrieve the dedicated differential pressure decay periods T1, T2, ..., T for the previous n-1 door opening events. n-1 ; Calculate the superposition density factor f for each group of adjacent door opening events. i Its specific expression is: f i =T i / △t i , where △t i For the door opening event E i With E i+1 The time interval, i∈[1,n-1]; f i The larger the value, the higher the degree to which the previous disturbance has not dissipated, and the stronger the superposition effect. When the number of door opening events n=1 within the window, it indicates no superposition effect; at this time, the window superposition coefficient is... When the number of door opening events n≥2 within the window, the window stacking coefficient is... .

[0022] like Figure 3 This invention provides a method for intelligent monitoring and early warning of gaseous pollutants in semiconductor manufacturing; As a fourth embodiment of this application, the specific steps include the following: Step 1: For each door opening event of the same door in the cleanroom, collect the real-time differential pressure curve and gap opening degree of the door, calculate the exclusive differential pressure decay period of the door opening event, and set a fixed monitoring period and opening number threshold. When the number of door opening events in the fixed monitoring period exceeds the corresponding opening number threshold, divide all door opening events in the fixed monitoring period into dynamic time windows, calculate the risk score of each window, and take the window with the highest risk score as the target window. Step 2: Extract the independent differential pressure curves of all door opening events within the target window, use differential pressure disturbance phase alignment to correct the timing deviation of each independent differential pressure curve, and divide the effective disturbance segment from the corresponding differential pressure curve through the exclusive differential pressure decay period of each door opening event. Then, combine the effective disturbance segments of all door opening events in the same phase to synthesize the total differential pressure curve of the window. Step 3: Calculate the cumulative impulse of the disturbance of the total pressure difference curve of the window, and predict the amount of pollutant penetration by combining the pollutant penetration coefficient of the semiconductor process in the corresponding region. Execute early warning linkage according to the pollutant penetration level.

[0023] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0024] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A semiconductor manufacturing gaseous pollutant intelligent monitoring and early warning system, characterized in that, The application relates to a window screening module, which collects a real-time differential pressure curve and a slit opening degree of a door of a clean room for each door opening event of the door, calculates an exclusive differential pressure decay period of the door opening event, sets a fixed monitoring period and an opening number threshold value, divides all door opening events in the fixed monitoring period into dynamic time windows when the number of door opening events in the fixed monitoring period exceeds the opening number threshold value, calculates a risk score of each window, and takes a window with the highest risk score as a target window. A curve construction module extracts independent differential pressure curves of all door opening events in the target window, corrects time sequence deviations of the independent differential pressure curves by adopting differential pressure disturbance phase alignment, divides effective disturbance sections from the corresponding differential pressure curves according to the exclusive differential pressure decay period of each door opening event, and synthesizes a window total differential pressure curve by superimposing the effective disturbance sections of all door opening events according to the same phase fluctuation. A real-time early warning module calculates a disturbance cumulative impulse of the window total differential pressure curve, predicts a pollution penetration amount by combining a pollution penetration coefficient of a corresponding area semiconductor process, and executes early warning linkage according to the pollution penetration amount classification. The specific operation of calculating the exclusive differential pressure decay period of the door opening event is as follows: locating a differential pressure peak point of the door opening event from the collected real-time differential pressure curve, taking a differential pressure baseline of a region without door opening and equipment disturbance as a steady-state differential pressure, recording a time consumption of the differential pressure from the peak point to the steady-state differential pressure, and taking the time consumption as a basic differential pressure decay period of the door opening event. 2.The semiconductor manufacturing gaseous pollutant intelligent monitoring and early warning system according to claim 1, wherein, The specific steps of dividing all door opening events in the fixed monitoring period into dynamic time windows are as follows: extracting all door opening events in the fixed monitoring period according to time sequence, calling the exclusive differential pressure decay period of each door opening event, and establishing a correlation list corresponding to the event occurrence time sequence and the exclusive differential pressure decay period. The specific operation of calculating the risk score of the window is as follows: ​ ​ ​ 3.The semiconductor manufacturing gaseous pollutant intelligent monitoring and early warning system according to claim 1, wherein, ​ ​ ​ ​ 4.The semiconductor manufacturing gaseous pollutant intelligent monitoring and early warning system according to claim 1, wherein, ​ Extract the maximum opening angle, gap opening degree maximum value, event interval difference of each door opening event, if the current door opening event is the last door opening event in the window, the difference is taken as the exclusive decay period itself, and the disturbance influence duration of each door opening event is recorded, the disturbance influence duration is selected from the exclusive pressure difference decay period and the event interval difference, specifically, when the exclusive pressure difference decay period ≤ event interval difference, the disturbance influence duration = exclusive pressure difference decay period, when the exclusive pressure difference decay period > event interval difference, the disturbance influence duration = event interval difference; The event interval difference is the difference between the exclusive pressure difference decay period of the door opening event and the time interval of the adjacent next door opening event; For each door opening event, multiply the maximum opening angle by the maximum gap opening degree to get the basic disturbance value, and then divide the basic disturbance value by the disturbance influence duration to get the dynamic disturbance intensity value; Sum the dynamic disturbance intensity values of all door opening events in the window, and then multiply by the window superposition coefficient to get the window risk score. 5.The semiconductor manufacturing gaseous pollutant intelligent monitoring and early warning system according to claim 4, wherein, The specific method for obtaining the window superposition coefficient is: Extract the timing data of all door opening events within the window, and label them as E in the order of occurrence j And at the same time, call the exclusive pressure difference decay period T of the previous n-1 door opening events i Where j ∈ [1, n], i ∈ [1, n-1], n is the number of door opening events within the window; calculating a superimposed density factor f for each group of adjacent door opening events i with the specific expression: f i = T i / △t i wherein△t i is the time interval between the door opening event E i and E i+1 ​ When the number of door opening events n in the window = 1, the window overlay coefficient ; when the number of door opening events n in the window ≥ 2, at this time the window overlay coefficient . 6.The semiconductor manufacturing gaseous pollutant intelligent monitoring and early warning system according to claim 1, wherein, The time sequence deviation of each independent pressure difference curve is corrected by pressure difference disturbance phase alignment, which specifically includes: Based on the steady-state pressure difference Ps of each independent pressure difference curve, locate the disturbance starting feature point of the curve; Translate the time axis of each independent pressure difference curve, so that the disturbance starting feature points of all pressure difference curves correspond to the same time zero point; The specific operation of locating the disturbance starting feature point of the independent pressure difference curve is: On this independent pressure difference curve, select the first sampling point that meets the following two conditions as the disturbance starting feature point: Condition one: the pressure difference sampling value of this sampling point is greater than Ps, and the next three pressure difference sampling points are all greater than Ps; Condition two: calculate the pressure difference difference between this sampling point and the previous adjacent sampling point, and divide by the time interval of the two sampling points to get the first derivative of this point, which is required to be greater than 0. 7.The semiconductor manufacturing gaseous pollutant intelligent monitoring and early warning system according to claim 1, wherein, The specific operation of dividing the effective disturbance section is: retrieve the exclusive pressure difference decay period of each independent pressure difference curve corresponding to the door opening event, and set the disturbance starting feature point of the pressure difference curve as the starting point. The curve section with a duration equal to the exclusive pressure difference decay period is the effective disturbance section. 8.The semiconductor manufacturing gaseous pollutant intelligent monitoring and early warning system according to claim 1, wherein, The specific steps for calculating the disturbance cumulative impulse of the window total pressure difference curve are: The window total pressure difference curve is divided into two stages of disturbance rising section and disturbance attenuation section, and the pressure difference deviation value at each sampling time in the two stages is calculated, that is, the difference between the pressure difference value at this time and Ps, the disturbance rising section is the curve section in which the total pressure difference rises from the steady state pressure difference Ps to the total peak value P max , and the disturbance attenuation section is the curve section in which the total pressure difference falls back from P max to Ps; Integrate the pressure difference deviation of the disturbance rising section and the disturbance decay section with respect to time to get the rising section impulse and the decay section impulse, and add them to get the disturbance cumulative impulse of the window total pressure difference curve. 9.The semiconductor manufacturing gaseous pollutant intelligent monitoring and early warning system according to claim 1, wherein, The specific steps for predicting the pollutant penetration amount are: Retrieve the basic pollutant penetration coefficient corresponding to the clean room area, and extract the maximum gap opening degree of all door opening events in the target window, and take its average as the window gap feature parameter. The basic pollutant penetration coefficient refers to the penetration amount of pollutants through a unit area gap under a unit pressure difference impulse; Multiply the disturbance cumulative impulse by the basic pollutant penetration coefficient to get the theoretical penetration amount under the standard gap state, and then multiply the theoretical penetration amount by the window gap feature parameter to predict the pollutant penetration amount.

10. A method for semiconductor manufacturing gaseous contamination intelligent monitoring and early warning, used for executing the semiconductor manufacturing gaseous contamination intelligent monitoring and early warning system of any one of claims 1-9, characterized in that, It includes: Step1、For each door opening event of a clean room, the real-time differential pressure curve and the gap opening degree of the door are collected, the exclusive differential pressure decay period of the door opening event is calculated, and at the same time, a fixed monitoring period and an opening frequency threshold are set. When the number of door opening events in the fixed monitoring period exceeds the corresponding opening frequency threshold, all door opening events in the fixed monitoring period are divided into dynamic time windows, and the risk score of each window is calculated. The window with the highest risk score is taken as the target window; Step2、Extract the independent differential pressure curve of all door opening events in the target window, align the time sequence deviation of each independent differential pressure curve by using the differential pressure disturbance phase, divide the effective disturbance section from the corresponding differential pressure curve through the exclusive differential pressure decay period of each door opening event, and superimpose the effective disturbance sections of all door opening events according to the same phase to synthesize the window total differential pressure curve; Step3、Calculate the disturbance cumulative impulse of the window total differential pressure curve, and predict the pollutant penetration amount combined with the pollutant penetration coefficient of the corresponding regional semiconductor process, and execute the early warning linkage according to the pollutant penetration amount grading.

Images