CN118248253B - Semiconductor gaseous molecular pollutant anomaly prediction and emergency response system and method - Google Patents

Semiconductor gaseous molecular pollutant anomaly prediction and emergency response system and method Download PDF

Info

Publication number
CN118248253B
CN118248253B CN202410627456.4A CN202410627456A CN118248253B CN 118248253 B CN118248253 B CN 118248253B CN 202410627456 A CN202410627456 A CN 202410627456A CN 118248253 B CN118248253 B CN 118248253B
Authority
CN
China
Prior art keywords
image
pollutant
concentration
firefly
prediction
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202410627456.4A
Other languages
Chinese (zh)
Other versions
CN118248253A (en
Inventor
江大白
胡增
王枫
汪刚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Applied Technology Co Ltd
Original Assignee
China Applied Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Applied Technology Co Ltd filed Critical China Applied Technology Co Ltd
Priority to CN202410627456.4A priority Critical patent/CN118248253B/en
Publication of CN118248253A publication Critical patent/CN118248253A/en
Application granted granted Critical
Publication of CN118248253B publication Critical patent/CN118248253B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/70—Machine learning, data mining or chemometrics
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/0004—Gaseous mixtures, e.g. polluted air
    • G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
    • G01N33/0062—General constructional details of gas analysers, e.g. portable test equipment concerning the measuring method or the display, e.g. intermittent measurement or digital display
    • G01N33/0063—General constructional details of gas analysers, e.g. portable test equipment concerning the measuring method or the display, e.g. intermittent measurement or digital display using a threshold to release an alarm or displaying means
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Chemical & Material Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Physics (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • Biomedical Technology (AREA)
  • Food Science & Technology (AREA)
  • Pathology (AREA)
  • Multimedia (AREA)
  • Medicinal Chemistry (AREA)
  • Combustion & Propulsion (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses an anomaly prediction and emergency response system and method for semiconductor gaseous molecular pollutants, which relate to the field of electric digital data processing, acquire historical monitoring data of the semiconductor gaseous molecular pollutants, generate images of the changes of the concentrations of various semiconductor gaseous molecular pollutants along with time, analyze various images in a concentration change image set of the pollutants by adopting an image analysis algorithm, and then carry out fitting prediction on the concentrations of the current semiconductor gaseous molecular pollutants by adopting a corresponding prediction algorithm; according to the method, the pollutant types in the pollutant type set are determined according to the historical monitoring data of the semiconductor gaseous molecular pollutants, so that the semiconductor gaseous molecular pollutants are comprehensively simulated and predicted, and the safety is improved; and a proper algorithm model is selected according to the magnitude of the image complexity value to simulate and predict the pollutant data, so that the prediction efficiency is improved.

Description

Semiconductor gaseous molecular pollutant anomaly prediction and emergency response system and method
Technical Field
The invention belongs to the field of electric digital data processing, and particularly relates to a semiconductor gaseous molecular pollutant abnormality prediction and emergency response system and method.
Background
Semiconductor gaseous molecular contaminants are gaseous contaminants that may be present during semiconductor fabrication that negatively impact device performance and reliability. These contaminants can come from a variety of sources, including factory environments, raw materials, equipment gas emissions, and the like, and some common gaseous molecular contaminants of semiconductors include: hydrofluoric acid, hydrochloric acid, phosphoric acid, sulfuric acid, nitric acid, sulfur dioxide, hydrogen sulfide, acetic acid, ethyl lactate, trimethylbenzene, triethyl phosphate, diethyl phthalate, etc., the existence of these pollutants may cause problems of transistor leakage, metal corrosion, unstable oxide layer, etc., and adversely affect the performance and reliability of semiconductor devices, so that in the semiconductor manufacturing process, it is very important to control and monitor these gaseous molecular pollutants, and appropriate measures such as air purification, gas filtration and monitoring instruments, etc. are taken, which can help reduce the influence of these pollutants and ensure the quality and reliability of semiconductor devices.
Chinese patent application CN117473706a discloses a method and a system for pre-evaluating gaseous pollutant emission of interior decoration materials, which combines a dry building material emission model, a wet building material emission model and a ventilation model with adsorption information and purification information to construct a pre-evaluation model, and evaluates gaseous pollutant emission of interior decoration materials through the pre-evaluation model to predict pollutant concentration at each stage in an actual building.
Sources of AMC in a semiconductor manufacturing plant are divided into external and internal sources, the external source mainly comes from the air cleaning and filtration system supplied to the interior of the clean room, and the interior may come from multiple sources including process residues, FOUP contamination, cleaning solvent use, personnel correlation, material escape in the clean room, etc., which if the AMC cannot be predicted in concentration and pre-processed, can cause significant damage to wafers stored in the clean room and cause loss to the enterprise.
Disclosure of Invention
Aiming at the problems in the related art, the invention provides a semiconductor gaseous molecular pollutant anomaly prediction and emergency response system and method, so as to overcome the technical problems in the prior art.
In order to solve the technical problems, the invention is realized by the following technical scheme:
The invention relates to a semiconductor gaseous molecular pollutant abnormality prediction and emergency response method, which comprises the following steps:
s1, acquiring historical monitoring data of gaseous molecular pollutants to obtain a first data set; obtaining a pollutant class set according to the first data set; obtaining corresponding abnormal concentration threshold values according to the characteristics of various gaseous molecular pollutants in the pollutant class set to obtain an abnormal concentration threshold value set;
S2, generating images of the concentration of various gaseous molecular pollutants with time according to a first data set, displaying the generated images on an image display screen, adopting image shooting equipment to shoot the image display screen to obtain a pollutant concentration change image set, setting an image complexity threshold set, adopting an image analysis algorithm to analyze various images in the pollutant concentration change image set to obtain a pollutant concentration change image complexity set, setting a concentration prediction algorithm set, and adopting a prediction algorithm corresponding to the concentration prediction algorithm set when the value of the pollutant concentration change image complexity set reaches a threshold value in the image complexity threshold value set to perform fitting prediction on the concentration of the current gaseous molecular pollutants;
S3, when the concentration of the gaseous molecular pollutants is larger than or equal to an abnormal concentration threshold value of the corresponding pollutants in the abnormal concentration threshold set in the fitting prediction process of the concentration of the gaseous molecular pollutants in the S2, sending out an electric signal to start the corresponding emergency treatment equipment to treat the gaseous molecular pollutants;
contaminant classes the concentrated gaseous molecular contaminant classes include, for example, acidic molecules, basic molecules, condensable molecules, doping molecules, poorly soluble organic species, and volatile organic species, etc., wherein the acidic molecules include, for example, hydrofluoric acid, hydrochloric acid, phosphoric acid, sulfuric acid, nitric acid, sulfur dioxide, hydrogen sulfide, and acetic acid, etc.; basic molecules include, for example, amines and ammonia; the condensable molecules comprise ethyl lactate, trimethylbenzene, triethyl phosphate, diethyl phthalate and the like; doping molecules include, for example, triethyl phosphate, boron trifluoride, boric acid, and the like; poorly soluble organic species include, for example, trimethylsilanol, methylene chloride, siloxanes, and the like; volatile organic compounds such as isopropanol, benzene, toluene, xylene and PGMEA are used for determining the pollutant types in the pollutant type set according to the historical monitoring data of the gaseous molecular pollutants, so that the gaseous molecular pollutants are comprehensively simulated and predicted without omission, and the safety is improved; in addition, the prediction efficiency is improved by analyzing the image generated by the historical detection data to determine the complexity of the image and then selecting a proper algorithm model according to the magnitude of the complexity value of the image to perform simulation prediction on the pollutant data, wherein the image analysis algorithm comprises a convolutional neural network algorithm.
Preferably, the step S1 includes the steps of:
s11, acquiring historical monitoring data of gaseous molecular pollutants to obtain a first data set WhereinIs the firstHistorical monitoring data of gaseous-like molecular contaminants,The total number of the gaseous molecular pollutant types;
S12, according to the first data set Obtaining a pollutant class setWhereinFor the first data setMiddle (f)The contaminant species corresponding to the historical monitoring data of the gaseous molecular contaminant species;
S13, obtaining corresponding abnormal concentration threshold values according to the characteristics of various gaseous molecular pollutants in the pollutant class set to obtain an abnormal concentration threshold value set WhereinAs a set of pollutant classesMiddle (f)An abnormal concentration threshold of contaminant-like material;
by generating the first data set, the pollutant class set and the abnormal concentration threshold set, a basis is provided for subsequent prediction and analysis of a prediction result.
Preferably, the step S2 includes the steps of:
s21, according to the first data set Generating images of the concentration of various gaseous molecular pollutants changing along with time, and displaying the generated images on an image display screen;
S22, shooting the image display screen by adopting image shooting equipment to obtain a pollutant concentration change image set WhereinIs the firstA concentration variation image of the contaminant-like substance;
s23, setting an image complexity threshold set Image complexity level setWhereinA first complexity threshold and a second complexity threshold,The first image complexity level, the second image complexity level and the third image complexity level are respectively;
Set concentration prediction algorithm set WhereinRespectively representing a linear regression algorithm, a Levenberg-Marquardt algorithm and a BP neural network algorithm;
The first image complexity level corresponds to a linear regression algorithm, the second image complexity level corresponds to a Levenberg-Marquardt algorithm, and the third image complexity level corresponds to a BP neural network algorithm;
s24, adopting an image analysis algorithm to change the pollutant concentration image set Processing each image in the image, and calculating the corresponding image complexity to obtain a pollutant concentration change image complexity setWhereinIs the firstConcentration of the contaminant-like changes image complexity;
s25, carrying out image complexity collection on the pollutant concentration change Image complexity value and image complexity threshold set for each type of contaminant concentration variation in the imageThe first complexity threshold value and the second complexity threshold value are compared, and the image complexity level set is matched according to the comparison resultDetermining the complexity level of a pollutant concentration change image, and selecting a corresponding concentration prediction algorithm according to the complexity level of the pollutant concentration change image so as to fit and predict historical monitoring data of the gaseous molecular pollutants;
The method comprises the steps of generating a pollutant concentration change image set and an image complexity threshold set, analyzing the pollutant concentration change image set by adopting an image analysis algorithm to obtain the number of features of the image, and taking the number of features as the complexity of the image, so that a basis is provided for classifying the image subsequently.
Preferably, the step S24 includes the steps of:
S241, adopting a convolutional neural network to change the concentration change image set of the pollutants Extracting features of each image to obtain an image feature setWhereinIs the firstA concentration-varied image feature set of the contaminant-like,，Is the firstContaminant-like concentration variation image NoThe characteristics of the device are that,Is the firstTotal number of features in the concentration variation image of the contaminant-like;
S242, setting the total number of features in the concentration change image of the pollutant as the complexity of the concentration change image of the pollutant to obtain a concentration change image complexity set of the pollutant ；
By adopting the convolutional neural network to extract the characteristics of the image, the image can be efficiently analyzed, so that the extracted characteristics are comprehensive and accurate, and more accurate data is provided for classifying the image subsequently.
Preferably, the step S241 includes the steps of:
s2411, in the contaminant concentration variation image set Randomly select% Data as training set;
S2412, setting the size of the convolution kernel in the convolution neural network as Setting the convolution step length asSetting the characteristic dimension asThe input and output calculation formulas among the layers in the convolutional neural network are as follows:
wherein: the characteristic diagram is input into a convolutional neural network; for the feature map output in the convolutional neural network, Is the convolution kernel size;
Setting an objective function of the convolutional neural network as follows:
wherein: for the objective function of the convolutional neural network, AndA generator and a discriminator respectively,AndRespectively input expected values of the original pollutant concentration change images and network expected values after the original pollutant concentration change images are processed,As an image of the change in the original contaminant concentration,An image after processing the original pollutant concentration change image;
Setting a convolutional neural network encryption layer, a pooling layer and a decryption layer, wherein the encryption layer comprises A first convolution kernelA second convolution kernel; each convolution layer in the decryption layer is provided with a corresponding pooling layer;
the discrimination loss function of the discriminator in the decryption layer is as follows:
wherein: In order to achieve a value of the loss function, For the original noisy image,To at the firstTraining loss at layer;
S2413, inputting the training set into a convolutional neural network for training, and setting a network convergence threshold value When the loss function value of the convolutional neural network and the network convergence threshold valueWhen the absolute value of the difference value is smaller than 0.001, training is completed, and a converged network model is obtained;
S2414 to be All images in the image are input into the converged network model for network extraction to obtain the image feature set；
The encryption layer, the pooling layer and the decryption layer are set in the convolutional neural network, so that the convolutional neural network is finer in the process of extracting the features, and the result is more accurate.
Preferably, the step S25 includes the steps of:
S251, when When it willIs set to a complexity level ofWhen (when)When it willIs set to a complexity level ofWhen (when)When it willIs set to a complexity level of。
S252, setting a prediction time period setWhereinTo aim at the firstThe time period for which the class of contaminants needs to be predicted, whenIs of the complexity level ofWhen adopting linear regression algorithm pairFitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change condition in the period to obtain a first partial prediction result data set; When (when)Is of the complexity level ofWhen the Levenberg-Marquardt algorithm pair is adoptedFitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change condition in the period to obtain a second partial prediction result data set; When (when)Is of the complexity level ofWhen the BP neural network algorithm is adoptedFitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change condition in the period to obtain a third partial prediction result data set；
S253, the first partial prediction result data setSecond partial prediction result data setAnd third partial prediction result data setMerging to obtain a final prediction result data setWhereinIs the firstContaminants of the class in the futurePredictive data of the concentration change conditions over a period of time,；
When the distribution of data points is linear, the complexity of the image is smaller than that of the imageWhen each data point distribution is in curve type distribution, the complexity of the image is greater than or equal toAnd is smaller thanWhen the distribution of each data point is not linear and is not curvilinear, the complexity of the image is greater than or equal toBy adopting the scheme, the images are classified according to the complexity, the images in linear distribution are simulated and predicted by adopting a linear regression algorithm, the images in curve distribution are simulated and predicted by adopting a Levenberg-Marquardt algorithm, and the rest images are simulated and predicted by adopting a BP neural network algorithm, so that the characteristics of the algorithm are fully combined, and the prediction process is more efficient.
Preferably, the BP neural network algorithm pair is adopted in S252Fitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change in the time period comprises the following steps:
s2521, constructing a BP neural network model, and setting an initial weight value of the BP neural network model Bias valueAnd adopting a firefly algorithm to perform the initial weight valueBias valueOptimizing; obtaining the optimal weight valueOptimum bias value；
S2522, calculating an error value of the BP neural network model to obtain a network model error value；
S2523, setting the network convergence error precision thresholdWhen the network model error valueFor the optimal weight valueOptimum bias valueUpdate and return to S2522, when network model error valueOutputting the predicted data to obtain the third partial predicted result data set；
By improving the BP neural network model by adopting a firefly algorithm, the performance of the BP neural network model is improved, so that the BP neural network model has a better prediction effect.
Preferably, the optimization process in S2521 includes the following steps:
s25211, set Dimensional space and saidThe dimension space is randomly distributed withFireflies only and constitute firefly populationsWhereinIs the first of the firefly populationsOnly the firefly is used for the production of the feed,The sum of the initial weight value and the offset value number is obtained; setting the firstFirefly onlyThe position in the firefly population isObtaining a position set of all fireflies in the firefly populationIn the following50% Of the data are randomly selected as initial weight values in the BP neural network model,The rest 50% of the data in the firefly population is used as the bias value in the BP neural network model, and the initial fluorescein values of all fireflies in the firefly population are set as followsA movement radius ofAnd set upThe objective function of the location isWill beAs an error function of the BP neural network model;
s25212, updating the luciferin values of all fireflies in the firefly population, according to S25211 Objective function of locationSetting the firstFirefly onlyThe fluorescein update formula of (2) is as follows:
wherein: Is that The luciferin weight of the firefly at the moment,；AndRespectively the firstFirefly onlyTime of day and time of dayA fluorescein value at a time; Update rate for fluorescein;
s25213, when Firefly onlyAround which there is a movement range and a firstWhen only the moving range of fireflies is contacted or there is a firefly with a superposition part, the method is as followsFirefly onlyIs updated according to the position of the first part, and after the position updating is completed, the first partFirefly onlyIs updated and the first is setThe firefly-only location update formula is as follows:
wherein: Is the moving step length; Is the first in firefly population Firefly only and the firstOnly the distance between fireflies; And Respectively the firstFirefly onlyTime of day and time of dayThe position in the firefly population at the moment; Is the first The location of fireflies only in the firefly population;
When the first is Firefly onlyNo movement range and no second surroundingWhen only the moving range of fireflies is contacted or there is a firefly with a superposition part, the method is as followsFirefly onlyUpdating the moving radius of the moving frame, and entering the next step after the moving radius is updated; setting the firstThe updated formula for firefly movement radius only is as follows:
wherein: And Respectively the firstFirefly onlyTime of day and time of dayA radius of movement at the moment; is a radius of movement threshold; is the adjacent domain change rate; a threshold value for the number of fireflies in adjacent domains; Is the first Only the size of firefly adjacent domains;
s25214, setting a first iteration number threshold When the iteration number of the firefly algorithm is greater than or equal to the first iteration number thresholdWhen the algorithm stops iterating, the optimal weight value of the BP neural network model is obtainedOptimum bias value；
According to the BP neural network model, the problem of local optimum or larger error is likely to exist in the iterative process of predicting the pollutant concentration data, and the initial weight and the bias value of the BP neural network model can be better optimized by adopting a firefly algorithm, so that the BP neural network model can obtain the optimal weight value and the optimal bias value, the problem of the BP neural network model that the local optimum or larger error is likely to exist in the iterative process is solved, the accuracy of a prediction result is improved, and an accurate basis is provided for subsequent emergency treatment.
Preferably, the step S3 includes the steps of:
S31, will And abnormal concentration threshold setComparing and settingWhereinCentralizing for predicted time periodsIn the time period ofThe first time nodeA predicted value of the concentration of the quasi-contaminant;
S32, when When start the firstEmergency disposal device for pollutantsPollutant-like concentration is treated whenWhen it is not needed to be toEmergency treatment of the pollutant;
whether emergency treatment is needed is determined according to the prediction data, so that the abnormal condition mode of the concentration of the pollutants can be prevented in advance, and the safety is improved.
Preferably, the gaseous molecular contaminant comprises a semiconductor gaseous molecular contaminant.
The invention also discloses a system for realizing the semiconductor gaseous molecular pollutant anomaly prediction and emergency response method, which comprises a pollutant history monitoring data acquisition processing module, a pollutant image generation module, a pollutant concentration change image processing module, a pollutant concentration prediction module and a pollutant emergency treatment module;
the pollutant history monitoring data acquisition processing module is used for acquiring history monitoring data of semiconductor gaseous molecular pollutants and generating a pollutant class set and an abnormal concentration threshold set;
The pollutant image generation module is used for generating images of the concentration of various semiconductor gaseous molecular pollutants with time according to the first data set, and the generated images are displayed on the image display screen;
The pollutant concentration change image processing module is used for analyzing various images in the pollutant concentration change image set by adopting an image analysis algorithm;
the pollutant concentration prediction module is used for carrying out fitting prediction on the concentration of the semiconductor gaseous molecular pollutants;
The pollutant emergency treatment module is used for feeding back according to the prediction result of the pollutant concentration prediction module so as to start corresponding emergency treatment equipment to treat the semiconductor gaseous molecular pollutants.
The invention has the following beneficial effects:
The invention sets a pollutant history monitoring data acquisition and processing module, a pollutant image generation module, a pollutant concentration change image processing module, a pollutant concentration prediction module and a pollutant emergency treatment module, wherein the pollutant category concentrated semiconductor gaseous molecular pollutant category generated in the pollutant history monitoring data acquisition and processing module comprises acid molecules, alkaline molecules, condensable molecules, doping molecules, insoluble organic substances, volatile organic substances and the like, and the acid molecules comprise hydrofluoric acid, hydrochloric acid, phosphoric acid, sulfuric acid, nitric acid, sulfur dioxide, hydrogen sulfide, acetic acid and the like; basic molecules include, for example, amines and ammonia; the condensable molecules comprise ethyl lactate, trimethylbenzene, triethyl phosphate, diethyl phthalate and the like; doping molecules include, for example, triethyl phosphate, boron trifluoride, boric acid, and the like; poorly soluble organic species include, for example, trimethylsilanol, methylene chloride, siloxanes, and the like; volatile organic compounds such as isopropanol, benzene, toluene, xylene and PGMEA are used for determining the pollutant types in the pollutant type set according to the historical monitoring data of the semiconductor gaseous molecular pollutants, so that the comprehensive simulation and prediction of various semiconductor gaseous molecular pollutants are realized, omission is avoided, and the safety is improved; in addition, the pollutant image generation module, the pollutant concentration change image processing module and the pollutant concentration prediction module analyze images generated by historical detection data to determine the complexity of the images, and then select a proper algorithm model to simulate and predict the pollutant data according to the magnitude of the image complexity value, so that the prediction efficiency is improved, AMC can be preprocessed, the possibility of damage to wafers stored in a clean room caused by AMC is reduced, and the damage to enterprises is avoided.
In the invention, when the distribution of each data point is linear, the complexity of the image is smaller than that of the imageWhen each data point distribution is in curve type distribution, the complexity of the image is greater than or equal toAnd is smaller thanWhen the distribution of each data point is not linear and is not curvilinear, the complexity of the image is greater than or equal toBy adopting the scheme, the images are classified according to the complexity, the images in linear distribution are simulated and predicted by adopting a linear regression algorithm, the images in curve distribution are simulated and predicted by adopting a Levenberg-Marquardt algorithm, and the rest images are simulated and predicted by adopting a BP neural network algorithm, so that the characteristics of the algorithm are fully combined, and the prediction process is more efficient.
According to the method, the problem of local optimum or larger error is likely to exist in the iterative process of predicting the pollutant concentration data according to the BP neural network model, and the initial weight and the bias value of the BP neural network model can be better optimized by adopting the firefly algorithm, so that the BP neural network model can obtain the optimal weight value and the optimal bias value, the problem of the BP neural network model that the local optimum or larger error is likely to exist in the iterative process is solved, the accuracy of a prediction result is improved, and an accurate basis is provided for subsequent emergency treatment.
Of course, it is not necessary for any one product to practice the invention to achieve all of the advantages set forth above at the same time.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the invention, the drawings that are needed for the description of the embodiments will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the invention, and that it is also possible for a person skilled in the art to obtain the drawings from these drawings without inventive effort.
FIG. 1 is a schematic flow chart of the semiconductor gaseous molecular contaminant anomaly prediction and emergency response system for predicting the concentration of semiconductor gaseous molecular contaminant.
Detailed Description
The following description of the technical solutions in the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, based on the embodiments in the invention, which a person of ordinary skill in the art would obtain without inventive faculty, are within the scope of the invention.
In the description of the present invention, it should be understood that the terms "open," "upper," "lower," "top," "middle," "inner," and the like indicate an orientation or positional relationship, merely for convenience of description and to simplify the description, and do not indicate or imply that the components or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus should not be construed as limiting the invention.
Referring to fig. 1, the present invention is a semiconductor gaseous molecular contaminant anomaly prediction and emergency response method, comprising the steps of:
s1, acquiring historical monitoring data of gaseous molecular pollutants to obtain a first data set; obtaining a pollutant class set according to the first data set; obtaining corresponding abnormal concentration threshold values according to the characteristics of various gaseous molecular pollutants in the pollutant class set to obtain an abnormal concentration threshold value set;
the step S1 comprises the following steps:
s11, acquiring historical monitoring data of gaseous molecular pollutants to obtain a first data set WhereinIs the firstHistorical monitoring data of gaseous-like molecular contaminants,The total number of the gaseous molecular pollutant types;
S12, according to the first data set Obtaining a pollutant class setWhereinFor the first data setMiddle (f)The contaminant species corresponding to the historical monitoring data of the gaseous molecular contaminant species;
S13, obtaining corresponding abnormal concentration threshold values according to the characteristics of various gaseous molecular pollutants in the pollutant class set to obtain an abnormal concentration threshold value set WhereinAs a set of pollutant classesMiddle (f)An abnormal concentration threshold of contaminant-like material;
S2, generating images of the concentration of various gaseous molecular pollutants with time according to a first data set, displaying the generated images on an image display screen, adopting image shooting equipment to shoot the image display screen to obtain a pollutant concentration change image set, setting an image complexity threshold set, adopting an image analysis algorithm to analyze various images in the pollutant concentration change image set to obtain a pollutant concentration change image complexity set, setting a concentration prediction algorithm set, and adopting a prediction algorithm corresponding to the concentration prediction algorithm set when the value of the pollutant concentration change image complexity set reaches a threshold value in the image complexity threshold value set to perform fitting prediction on the concentration of the current gaseous molecular pollutants;
The step S2 comprises the following steps:
s21, according to the first data set Generating images of the concentration of various gaseous molecular pollutants changing along with time, and displaying the generated images on an image display screen;
S22, shooting the image display screen by adopting image shooting equipment to obtain a pollutant concentration change image set WhereinIs the firstA concentration variation image of the contaminant-like substance;
s23, setting an image complexity threshold set Image complexity level setWhereinA first complexity threshold and a second complexity threshold,The first image complexity level, the second image complexity level and the third image complexity level are respectively;
Set concentration prediction algorithm set WhereinRespectively representing a linear regression algorithm, a Levenberg-Marquardt algorithm and a BP neural network algorithm;
The first image complexity level corresponds to a linear regression algorithm, the second image complexity level corresponds to a Levenberg-Marquardt algorithm, and the third image complexity level corresponds to a BP neural network algorithm;
s24, adopting an image analysis algorithm to change the pollutant concentration image set Processing each image in the image, and calculating the corresponding image complexity to obtain a pollutant concentration change image complexity setWhereinIs the firstConcentration of the contaminant-like changes image complexity;
the step S24 includes the steps of:
S241, adopting a convolutional neural network to change the concentration change image set of the pollutants Extracting features of each image to obtain an image feature setWhereinIs the firstA concentration-varied image feature set of the contaminant-like,，Is the firstContaminant-like concentration variation image NoThe characteristics of the device are that,Is the firstTotal number of features in the concentration variation image of the contaminant-like;
the step S241 includes the steps of:
s2411, in the contaminant concentration variation image set Randomly select% Data as training set;
S2412, setting the size of the convolution kernel in the convolution neural network as Setting the convolution step length asSetting the characteristic dimension asThe input and output calculation formulas among the layers in the convolutional neural network are as follows:
wherein: the characteristic diagram is input into a convolutional neural network; for the feature map output in the convolutional neural network, Is the convolution kernel size;
Setting an objective function of the convolutional neural network as follows:
wherein: for the objective function of the convolutional neural network, AndA generator and a discriminator respectively,AndRespectively input expected values of the original pollutant concentration change images and network expected values after the original pollutant concentration change images are processed,As an image of the change in the original contaminant concentration,An image after processing the original pollutant concentration change image;
Setting a convolutional neural network encryption layer, a pooling layer and a decryption layer, wherein the encryption layer comprises A first convolution kernelA second convolution kernel; each convolution layer in the decryption layer is provided with a corresponding pooling layer;
the discrimination loss function of the discriminator in the decryption layer is as follows:
wherein: In order to achieve a value of the loss function, For the original noisy image,To at the firstTraining loss at layer;
S2413, inputting the training set into a convolutional neural network for training, and setting a network convergence threshold value When the loss function value of the convolutional neural network and the network convergence threshold valueWhen the absolute value of the difference value is smaller than 0.001, training is completed, and a converged network model is obtained;
S2414 to be All images in the image are input into the converged network model for network extraction to obtain the image feature set；
S242, setting the total number of features in the concentration change image of the pollutant as the complexity of the concentration change image of the pollutant to obtain a concentration change image complexity set of the pollutant；
S25, carrying out image complexity collection on the pollutant concentration changeImage complexity value and image complexity threshold set for each type of contaminant concentration variation in the imageThe first complexity threshold value and the second complexity threshold value are compared, and the image complexity level set is matched according to the comparison resultDetermining the complexity level of a pollutant concentration change image, and selecting a corresponding concentration prediction algorithm according to the complexity level of the pollutant concentration change image so as to fit and predict historical monitoring data of the gaseous molecular pollutants;
The step S25 includes the steps of:
S251, when When it willIs set to a complexity level ofWhen (when)When it willIs set to a complexity level ofWhen (when)When it willIs set to a complexity level of。
S252, setting a prediction time period setWhereinTo aim at the firstThe time period for which the class of contaminants needs to be predicted, whenIs of the complexity level ofWhen adopting linear regression algorithm pairFitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change condition in the period to obtain a first partial prediction result data set; When (when)Is of the complexity level ofWhen the Levenberg-Marquardt algorithm pair is adoptedFitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change condition in the period to obtain a second partial prediction result data set; When (when)Is of the complexity level ofWhen the BP neural network algorithm is adoptedFitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change condition in the period to obtain a third partial prediction result data set；
The BP neural network algorithm pair is adopted in S252Fitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change in the time period comprises the following steps:
s2521, constructing a BP neural network model, and setting an initial weight value of the BP neural network model Bias valueAnd adopting a firefly algorithm to perform the initial weight valueBias valueOptimizing; obtaining the optimal weight valueOptimum bias value；
The optimization process in S2521 includes the steps of:
s25211, set Dimensional space and saidThe dimension space is randomly distributed withFireflies only and constitute firefly populationsWhereinIs the first of the firefly populationsOnly the firefly is used for the production of the feed,The sum of the initial weight value and the offset value number is obtained; setting the firstFirefly onlyThe position in the firefly population isObtaining a position set of all fireflies in the firefly populationIn the following50% Of the data are randomly selected as initial weight values in the BP neural network model,The rest 50% of the data in the firefly population is used as the bias value in the BP neural network model, and the initial fluorescein values of all fireflies in the firefly population are set as followsA movement radius ofAnd set upThe objective function of the location isWill beAs an error function of the BP neural network model;
s25212, updating the luciferin values of all fireflies in the firefly population, according to S25211 Objective function of locationSetting the firstFirefly onlyThe fluorescein update formula of (2) is as follows:
wherein: Is that The luciferin weight of the firefly at the moment,；AndRespectively the firstFirefly onlyTime of day and time of dayA fluorescein value at a time; Update rate for fluorescein;
s25213, when Firefly onlyAround which there is a movement range and a firstWhen only the moving range of fireflies is contacted or there is a firefly with a superposition part, the method is as followsFirefly onlyIs updated according to the position of the first part, and after the position updating is completed, the first partFirefly onlyIs updated and the first is setThe firefly-only location update formula is as follows:
wherein: Is the moving step length; Is the first in firefly population Firefly only and the firstOnly the distance between fireflies; And Respectively the firstFirefly onlyTime of day and time of dayThe position in the firefly population at the moment; Is the first The location of fireflies only in the firefly population;
When the first is Firefly onlyNo movement range and no second surroundingWhen only the moving range of fireflies is contacted or there is a firefly with a superposition part, the method is as followsFirefly onlyUpdating the moving radius of the moving frame, and entering the next step after the moving radius is updated; setting the firstThe updated formula for firefly movement radius only is as follows:
wherein: And Respectively the firstFirefly onlyTime of day and time of dayA radius of movement at the moment; is a radius of movement threshold; is the adjacent domain change rate; a threshold value for the number of fireflies in adjacent domains; Is the first Only the size of firefly adjacent domains;
s25214, setting a first iteration number threshold When the iteration number of the firefly algorithm is greater than or equal to the first iteration number thresholdWhen the algorithm stops iterating, the optimal weight value of the BP neural network model is obtainedOptimum bias value；
S2522, calculating an error value of the BP neural network model to obtain a network model error value；
S2523, setting the network convergence error precision thresholdWhen the network model error valueFor the optimal weight valueOptimum bias valueUpdate and return to S2522, when network model error valueOutputting the predicted data to obtain the third partial predicted result data set；
S253, the first partial prediction result data setSecond partial prediction result data setAnd third partial prediction result data setMerging to obtain a final prediction result data setWhereinIs the firstContaminants of the class in the futurePredictive data of the concentration change conditions over a period of time,；
S3, when the concentration of the gaseous molecular pollutants is larger than or equal to an abnormal concentration threshold value of the corresponding pollutants in the abnormal concentration threshold set in the fitting prediction process of the concentration of the gaseous molecular pollutants in the S2, sending out an electric signal to start the corresponding emergency treatment equipment to treat the gaseous molecular pollutants;
the step S3 comprises the following steps:
S31, will And abnormal concentration threshold setComparing and settingWhereinCentralizing for predicted time periodsIn the time period ofThe first time nodeA predicted value of the concentration of the quasi-contaminant;
S32, when When start the firstEmergency disposal device for pollutantsPollutant-like concentration is treated whenWhen it is not needed to be toAnd carrying out emergency treatment on the pollutant.
The semiconductor gaseous molecular pollutant anomaly prediction and emergency response system comprises a pollutant history monitoring data acquisition processing module, a pollutant image generation module, a pollutant concentration change image processing module, a pollutant concentration prediction module and a pollutant emergency treatment module;
the pollutant history monitoring data acquisition processing module is used for acquiring history monitoring data of semiconductor gaseous molecular pollutants and generating a pollutant class set and an abnormal concentration threshold set;
The pollutant image generation module is used for generating images of the concentration of various semiconductor gaseous molecular pollutants with time according to the first data set, and the generated images are displayed on the image display screen;
The pollutant concentration change image processing module is used for analyzing various images in the pollutant concentration change image set by adopting an image analysis algorithm;
the pollutant concentration prediction module is used for carrying out fitting prediction on the concentration of the semiconductor gaseous molecular pollutants;
The pollutant emergency treatment module is used for feeding back according to the prediction result of the pollutant concentration prediction module so as to start corresponding emergency treatment equipment to treat the semiconductor gaseous molecular pollutants.
In the description of the present specification, the descriptions of the terms "one embodiment," "example," "specific example," and the like, mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The above disclosed preferred embodiments of the invention are merely intended to help illustrate the invention. The preferred embodiments are not exhaustive or to limit the invention to the precise form disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best understand and utilize the invention.

Claims (6)

1. A semiconductor gaseous molecular pollutant abnormality prediction and emergency response method is characterized in that: the method comprises the following steps:
s1, acquiring historical monitoring data of gaseous molecular pollutants to obtain a first data set; obtaining a pollutant class set according to the first data set; obtaining corresponding abnormal concentration threshold values according to the characteristics of various gaseous molecular pollutants in the pollutant class set to obtain an abnormal concentration threshold value set;
S2, generating images of the concentration of various gaseous molecular pollutants with time according to a first data set to obtain a pollutant concentration change image set, setting an image complexity threshold set, analyzing various images in the pollutant concentration change image set by adopting an image analysis algorithm to obtain a pollutant concentration change image complexity set, setting a concentration prediction algorithm set, and when the value of the pollutant concentration change image complexity set reaches a threshold value in the image complexity threshold set, adopting a prediction algorithm corresponding to the concentration prediction algorithm set corresponding to the threshold value to carry out fitting prediction on the concentration of the current gaseous molecular pollutants;
S3, when the concentration of the gaseous molecular pollutants is larger than or equal to an abnormal concentration threshold value of the corresponding pollutants in the abnormal concentration threshold set in the fitting prediction process of the concentration of the gaseous molecular pollutants in the S2, sending out an electric signal to start the corresponding emergency treatment equipment to treat the gaseous molecular pollutants;
the step S1 comprises the following steps:
s11, acquiring historical monitoring data of gaseous molecular pollutants to obtain a first data set WhereinIs the firstHistorical monitoring data of gaseous-like molecular contaminants,The total number of the gaseous molecular pollutant types;
S12, according to the first data set Obtaining a pollutant class setWhereinFor the first data setMiddle (f)The contaminant species corresponding to the historical monitoring data of the gaseous molecular contaminant species;
S13, obtaining corresponding abnormal concentration threshold values according to the characteristics of various gaseous molecular pollutants in the pollutant class set to obtain an abnormal concentration threshold value set WhereinAs a set of pollutant classesMiddle (f)An abnormal concentration threshold of contaminant-like material;
The step S2 comprises the following steps:
s21, according to the first data set Generating images of the concentration of various gaseous molecular pollutants changing along with time, and displaying the generated images on an image display screen;
S22, shooting the image display screen by adopting image shooting equipment to obtain a pollutant concentration change image set WhereinIs the firstA concentration variation image of the contaminant-like substance;
s23, setting an image complexity threshold set Image complexity level setWhereinA first complexity threshold and a second complexity threshold,The first image complexity level, the second image complexity level and the third image complexity level are respectively;
Set concentration prediction algorithm set WhereinRespectively representing a linear regression algorithm, a Levenberg-Marquardt algorithm and a BP neural network algorithm;
The first image complexity level corresponds to a linear regression algorithm, the second image complexity level corresponds to a Levenberg-Marquardt algorithm, and the third image complexity level corresponds to a BP neural network algorithm;
s24, adopting an image analysis algorithm to change the pollutant concentration image set Processing each image in the image, and calculating the corresponding image complexity to obtain a pollutant concentration change image complexity setWhereinIs the firstConcentration of the contaminant-like changes image complexity;
s25, carrying out image complexity collection on the pollutant concentration change Image complexity value and image complexity threshold set for each type of contaminant concentration variation in the imageThe first complexity threshold value and the second complexity threshold value are compared, and the image complexity level set is matched according to the comparison resultDetermining the complexity level of a pollutant concentration change image, and selecting a corresponding concentration prediction algorithm according to the complexity level of the pollutant concentration change image so as to fit and predict historical monitoring data of the gaseous molecular pollutants;
the step S24 includes the steps of:
S241, adopting a convolutional neural network to change the concentration change image set of the pollutants Extracting features of each image to obtain an image feature setWhereinIs the firstA concentration-varied image feature set of the contaminant-like,，Is the firstContaminant-like concentration variation image NoThe characteristics of the device are that,Is the firstTotal number of features in the concentration variation image of the contaminant-like;
S242, setting the total number of features in the concentration change image of the pollutant as the complexity of the concentration change image of the pollutant to obtain a concentration change image complexity set of the pollutant ；
The step S241 includes the steps of:
s2411, in the contaminant concentration variation image set Randomly select% Data as training set;
S2412, setting the size of the convolution kernel in the convolution neural network as Setting the convolution step length asSetting the characteristic dimension asThe input and output calculation formulas among the layers in the convolutional neural network are as follows:
；
wherein: the characteristic diagram is input into a convolutional neural network; for the feature map output in the convolutional neural network, Is the convolution kernel size;
Setting an objective function of the convolutional neural network as follows:
；
wherein: for the objective function of the convolutional neural network, AndA generator and a discriminator respectively,AndRespectively input expected values of the original pollutant concentration change images and network expected values after the original pollutant concentration change images are processed,As an image of the change in the original contaminant concentration,An image after processing the original pollutant concentration change image;
Setting a convolutional neural network encryption layer, a pooling layer and a decryption layer, wherein the encryption layer comprises A first convolution kernelA second convolution kernel; each convolution layer in the decryption layer is provided with a corresponding pooling layer;
the discrimination loss function of the discriminator in the decryption layer is as follows:
；
wherein: In order to achieve a value of the loss function, For the original noisy image,To at the firstTraining loss at layer;
S2413, inputting the training set into a convolutional neural network for training, and setting a network convergence threshold value When the loss function value of the convolutional neural network and the network convergence threshold valueWhen the absolute value of the difference value is smaller than 0.001, training is completed, and a converged network model is obtained;
S2414 to be All images in the image are input into the converged network model for network extraction to obtain the image feature set。
2. The method for anomaly prediction and emergency response of gaseous molecular pollutants of a semiconductor according to claim 1, wherein the method comprises the following steps: the step S25 includes the steps of:
S251, when When it willIs set to a complexity level ofWhen (when)When it willIs set to a complexity level ofWhen (when)When it willIs set to a complexity level of；
S252, setting a prediction time period setWhereinTo aim at the firstThe time period for which the class of contaminants needs to be predicted, whenIs of the complexity level ofWhen adopting linear regression algorithm pairFitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change condition in the period to obtain a first partial prediction result data set; When (when)Is of the complexity level ofWhen the Levenberg-Marquardt algorithm pair is adoptedFitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change condition in the period to obtain a second partial prediction result data set; When (when)Is of the complexity level ofWhen the BP neural network algorithm is adoptedFitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change condition in the period to obtain a third partial prediction result data set；
S253, the first partial prediction result data setSecond partial prediction result data setAnd third partial prediction result data setMerging to obtain a final prediction result data setWhereinIs the firstContaminants of the class in the futurePredictive data of the concentration change conditions over a period of time,。
3. The method for anomaly prediction and emergency response of gaseous molecular pollutants of a semiconductor according to claim 2, wherein the method comprises the following steps: the BP neural network algorithm pair is adopted in S252Fitting the data, and matching according to the fitting resultFuture ofPredicting the concentration change in the time period comprises the following steps:
s2521, constructing a BP neural network model, and setting an initial weight value of the BP neural network model Bias valueAnd adopting a firefly algorithm to perform the initial weight valueBias valueOptimizing to obtain an optimal weight valueOptimum bias value；
S2522, calculating an error value of the BP neural network model to obtain a network model error value；
S2523, setting the network convergence error precision thresholdWhen the network model error valueFor the optimal weight valueOptimum bias valueUpdate and return to S2522, when network model error valueOutputting the predicted data to obtain the third partial predicted result data set。
4. A method of semiconductor gaseous molecular contaminant anomaly prediction and emergency response according to claim 3, wherein: the optimization process in S2521 includes the steps of:
s25211, set Dimensional space and saidThe dimension space is randomly distributed withFireflies only and constitute firefly populationsWhereinIs the first of the firefly populationsOnly the firefly is used for the production of the feed,The sum of the initial weight value and the offset value number is obtained; setting the firstFirefly onlyThe position in the firefly population isObtaining a position set of all fireflies in the firefly populationIn the following50% Of the data are randomly selected as initial weight values in the BP neural network model,The rest 50% of the data in the firefly population is used as the bias value in the BP neural network model, and the initial fluorescein values of all fireflies in the firefly population are set as followsA movement radius ofAnd set upThe objective function of the location isWill beAs an error function of the BP neural network model;
s25212, updating the luciferin values of all fireflies in the firefly population, according to S25211 Objective function of locationSetting the firstFirefly onlyThe fluorescein update formula of (2) is as follows:
；
wherein: Is that The luciferin weight of the firefly at the moment,；AndRespectively the firstFirefly onlyTime of day and time of dayA fluorescein value at a time; Update rate for fluorescein;
s25213, when Firefly onlyAround which there is a movement range and a firstWhen only the moving range of fireflies is contacted or there is a firefly with a superposition part, the method is as followsFirefly onlyIs updated according to the position of the first part, and after the position updating is completed, the first partFirefly onlyIs updated and the first is setThe firefly-only location update formula is as follows:
；
wherein: Is the moving step length; Is the first in firefly population Firefly only and the firstOnly the distance between fireflies; And Respectively the firstFirefly onlyTime of day and time of dayThe position in the firefly population at the moment; Is the first The location of fireflies only in the firefly population;
When the first is Firefly onlyNo movement range and no second surroundingWhen only the moving range of fireflies is contacted or there is a firefly with a superposition part, the method is as followsFirefly onlyUpdating the moving radius of the moving frame, and entering the next step after the moving radius is updated; setting the firstThe updated formula for firefly movement radius only is as follows:
；
wherein: And Respectively the firstFirefly onlyTime of day and time of dayA radius of movement at the moment; is a radius of movement threshold; is the adjacent domain change rate; a threshold value for the number of fireflies in adjacent domains; Is the first Only the size of firefly adjacent domains;
s25214, setting a first iteration number threshold When the iteration number of the firefly algorithm is greater than or equal to the first iteration number thresholdWhen the algorithm stops iterating, the optimal weight value of the BP neural network model is obtainedOptimum bias value。
5. The method for anomaly prediction and emergency response of gaseous molecular pollutants of a semiconductor according to claim 1, wherein the method comprises the following steps: the step S3 comprises the following steps:
S31, will And abnormal concentration threshold setComparing and settingWhereinCentralizing for predicted time periodsIn the time period ofThe first time nodeA predicted value of the concentration of the quasi-contaminant;
S32, when When start the firstEmergency disposal device for pollutantsPollutant-like concentration is treated whenWhen it is not needed to be toAnd carrying out emergency treatment on the pollutant.
6. A system for implementing the semiconductor gaseous molecular contaminant anomaly prediction and emergency response method of any one of claims 1-5, wherein: the device comprises a pollutant history monitoring data acquisition and processing module, a pollutant image generation module, a pollutant concentration change image processing module, a pollutant concentration prediction module and a pollutant emergency treatment module;
the pollutant history monitoring data acquisition processing module is used for acquiring history monitoring data of semiconductor gaseous molecular pollutants and generating a pollutant class set and an abnormal concentration threshold set;
The pollutant image generation module is used for generating images of the concentration of various semiconductor gaseous molecular pollutants with time according to the first data set, and the generated images are displayed on the image display screen;
The pollutant concentration change image processing module is used for analyzing various images in the pollutant concentration change image set by adopting an image analysis algorithm;
the pollutant concentration prediction module is used for carrying out fitting prediction on the concentration of the semiconductor gaseous molecular pollutants;
The pollutant emergency treatment module is used for feeding back according to the prediction result of the pollutant concentration prediction module so as to start corresponding emergency treatment equipment to treat the semiconductor gaseous molecular pollutants.
CN202410627456.4A 2024-05-21 2024-05-21 Semiconductor gaseous molecular pollutant anomaly prediction and emergency response system and method Active CN118248253B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202410627456.4A CN118248253B (en) 2024-05-21 2024-05-21 Semiconductor gaseous molecular pollutant anomaly prediction and emergency response system and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202410627456.4A CN118248253B (en) 2024-05-21 2024-05-21 Semiconductor gaseous molecular pollutant anomaly prediction and emergency response system and method

Publications (2)

Publication Number Publication Date
CN118248253A CN118248253A (en) 2024-06-25
CN118248253B true CN118248253B (en) 2024-08-06

Family

ID=91552800

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202410627456.4A Active CN118248253B (en) 2024-05-21 2024-05-21 Semiconductor gaseous molecular pollutant anomaly prediction and emergency response system and method

Country Status (1)

Country Link
CN (1) CN118248253B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113588884A (en) * 2021-08-03 2021-11-02 亚翔系统集成科技（苏州）股份有限公司 Pollution source searching method based on AMC (advanced mezzanine technology) online monitoring system
CN115238807A (en) * 2022-07-29 2022-10-25 中用科技有限公司 AMC detection method based on artificial intelligence

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007322108A (en) * 2006-06-05 2007-12-13 Ohbayashi Corp Prediction method for gaseous contaminant concentration, designing method based on prediction method, and construction method for building designed by designing method
CN117741048A (en) * 2023-10-09 2024-03-22 中用科技(南通)有限公司 Analysis method of gaseous molecular pollutants in clean factory
CN117129636B (en) * 2023-10-23 2024-01-26 中用科技有限公司 Gaseous molecular pollutant on-line monitoring system for semiconductor manufacturing
CN117352088A (en) * 2023-10-26 2024-01-05 广州大学 Prediction method of spatial pollutant distribution based on convolutional neural network

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113588884A (en) * 2021-08-03 2021-11-02 亚翔系统集成科技（苏州）股份有限公司 Pollution source searching method based on AMC (advanced mezzanine technology) online monitoring system
CN115238807A (en) * 2022-07-29 2022-10-25 中用科技有限公司 AMC detection method based on artificial intelligence

Also Published As

Publication number Publication date
CN118248253A (en) 2024-06-25

Similar Documents

Publication Publication Date Title
US11940740B2 (en) Methods and apparatus for obtaining diagnostic information relating to an industrial process
Cui et al. Research on remaining useful life prediction of rolling element bearings based on time-varying Kalman filter
TWI524189B (en) Method and system for detection of tool performance degradation and mismatch and related computer program product and apparatus
KR20230141930A (en) Advanced semiconductor process optimization and adaptive control during manufacturing
Harkat et al. Enhanced data validation strategy of air quality monitoring network
CN113330463A (en) Chamber matching in semiconductor device tools using neural networks
Pan et al. A virtual metrology system for predicting end-of-line electrical properties using a MANCOVA model with tools clustering
CN110889260B (en) Method and device for detecting process parameters, electronic equipment and computer readable medium
CN117741048A (en) Analysis method of gaseous molecular pollutants in clean factory
TW202145304A (en) Semiconductor device manufacturing system and semiconductor device manufacturing method
CN118248253B (en) Semiconductor gaseous molecular pollutant anomaly prediction and emergency response system and method
CN118015795B (en) Ppb-level pollutant monitoring and early warning method and system based on convolutional neural network
TW202420003A (en) Sustainability monitoring platform with sensor support
CN114252582A (en) Soil pollutant monitoring method based on Kalman filtering
CN114861522A (en) Precision manufacturing quality monitoring method and device based on artificial intelligence meta-learning technology
CN113222884B (en) Method and system for inspecting a sample
CN117436532A (en) Root cause analysis method for gaseous molecular pollutants in clean room
CN113588884B (en) Pollution source searching method based on AMC on-line monitoring system
CN118211942B (en) Semiconductor gaseous molecular pollutant space distribution management system and method
CN117672928B (en) Box opening method
US20240210916A1 (en) Machine and deep learning techniques for predicting ecological efficiency in substrate processing
CN118465877B (en) Method and apparatus for distinguishing the concentration of ozone contribution from local and external sources of pollution
CN116228749B (en) Wafer defect detection method and system based on inverse fact interpretation
Tsao et al. Real-Time Fab-Wise Airborne Molecular Contaminant (AMC) Monitoring System Using Multiple Fourier Transform Infrared (FTIR) Spectrometers in a Semiconductor Plant
US20240329622A1 (en) Time series anomaly detection in integrated circuit (ic) fabrication environment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant