Process data monitoring method and semiconductor process equipment

Abstract

The embodiment of the invention provides a process data monitoring method, and the method comprises the steps: collecting first real-time process data in each process step, the first real-time process data comprising real-time parameter values of a target parameter at a plurality of first time points; respectively inputting the first real-time process data into a Gaussian component model corresponding to each process step to obtain a plurality of probability values; the Gaussian component model is obtained by training according to historical process data, and the Gaussian component model corresponding to the maximum value in the plurality of probability values is determined as a target Gaussian component model; inputting the first real-time process data into the target Gaussian component model to obtain a process data interval; and when the second real-time process data is not in the process data interval, outputting prompt information.

Description

Technical Field

[0001] The present application relates to the field of semiconductor technology, and particularly to a process data monitoring method and a semiconductor process equipment. Background Art

[0002] At present, with the development of automation technology, the real-time process data corresponding to the target parameters in the process steps is an important factor affecting the process effect. The consistency between the actual parameter value and the target parameter value is directly related to the quality of the process. At the same time, during the heating process in the process, the power supply output is high, which is likely to cause equipment safety problems. Therefore, monitoring the real-time changes of process data and sending an alarm in a timely manner in case of an abnormality is beneficial to ensuring the process index quality and safety of the equipment.

[0003] During the current process data monitoring process, it is necessary to monitor the real-time process data corresponding to the target parameters according to the threshold values given by human experience. However, the threshold values given by human experience have limitations, and false negatives and false positives are likely to occur, and problems in the processing steps cannot be found in a timely and accurate manner. Summary of the Invention

[0004] The technical problem to be solved by the embodiments of the present application is that problems in the processing steps cannot be found in a timely and accurate manner.

[0005] To solve the above problems, the embodiments of the present application disclose a process data monitoring method, and the process data monitoring method includes:

[0006] Collect the first real-time process data in each process step, and the first real-time process data includes: the real-time parameter values of the target parameter at a plurality of first time points;

[0007] Input the first real-time process data into the Gaussian component model corresponding to each process step respectively to obtain a plurality of probability values; the Gaussian component model is trained according to historical process data, and the historical process data includes: the historical parameter values of the target parameter at a plurality of training time points and the process step corresponding to each training time point; one Gaussian component model is associated with one process step;

[0008] Determine the Gaussian component model corresponding to the maximum value among the plurality of probability values as the target Gaussian component model;

[0009] Input the first real-time process data into the target Gaussian component model to obtain a process data interval;

[0010] In the case where the second real-time process data is not within the process data interval, output a prompt message, and the second real-time process data includes: the real-time parameter values of the target parameter at a plurality of second time points, and the plurality of second times are after the plurality of first time points.

[0011] Embodiments of the present application disclose a semiconductor process equipment, which includes:

[0012] A controller, configured to collect first real-time process data in each of the process steps, where the first real-time process data includes: real-time parameter values of a target parameter at a plurality of first time points;

[0013] Input the first real-time process data into Gaussian component models corresponding to each of the process steps respectively to obtain a plurality of probability values; the Gaussian component models are trained according to historical process data, and the historical process data includes: historical parameter values of the target parameter at a plurality of training time points and the process steps corresponding to each of the training time points; one Gaussian component model is associated with one process step;

[0014] Determine the Gaussian component model corresponding to the maximum value among the plurality of probability values as the target Gaussian component model;

[0015] Input the first real-time process data into the target Gaussian component model to obtain a process data interval;

[0016] In the case where second real-time process data is not within the process data interval, output a prompt message, where the second real-time process data includes: real-time parameter values of the target parameter at a plurality of second time points, and the plurality of second times are after the plurality of first time points.

[0017] According to the embodiments of the present application, by collecting first real-time process data in each process step, the first real-time process data includes: real-time parameter values of a target parameter at a plurality of first time points; inputting the first real-time process data into Gaussian component models corresponding to each process step respectively to obtain a plurality of probability values; the Gaussian component models are trained according to historical process data, and the historical process data includes: historical parameter values of the target parameter at a plurality of training time points and the process steps corresponding to each training time point; one Gaussian component model is associated with one process step; determining the Gaussian component model corresponding to the maximum value among the plurality of probability values as the target Gaussian component model; thus, the first real-time process data corresponding to each first time point can be automatically divided into Gaussian component models corresponding to each process step, so as to input the first real-time process data into the target Gaussian component model to obtain a process data interval corresponding to the real-time parameter value of each first time point, which solves the limitation of designing the process data interval only based on manual experience and improves the accuracy of the process data interval. Thus, in the case where second real-time process data collected at time points after the first time point, i.e., a plurality of second time points, is not within the process data interval, outputting a prompt message can accurately monitor the second real-time process data and timely and accurately discover problems in the processing steps. Description of the Drawings

[0018] Figure 1 A schematic diagram of process data provided by an embodiment of the present application is shown;

[0019] Figure 2 A flowchart of a Gaussian component model training method provided by an embodiment of the present application is shown;

[0020] Figure 3 A flowchart of a process data monitoring method provided by an embodiment of the present application is shown;

[0021] Figure 4 A schematic diagram of a Gaussian component model provided by an embodiment of the present application is shown;

[0022] Figure 5 A schematic diagram of a Gaussian component model provided by an embodiment of the present application is shown:

[0023] Figure 6 A schematic diagram of the structure of a semiconductor process device provided by an embodiment of the present application is shown. Detailed implementation manners

[0024] The features and exemplary embodiments of various aspects of the present application will be described in detail below. To make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only configured to explain the present application and are not configured to limit the present application. For those skilled in the art, the present application can be implemented without some of these specific details. The following description of the embodiments is only intended to provide a better understanding of the present application by showing examples of the present application.

[0025] It should be noted that, in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, the elements defined by the statement "including..." do not exclude the presence of additional identical elements in the process, method, article or device including the elements.

[0026] First, in combination with Figure 1 A brief description of the semiconductor process equipment involved in the invention is given.

[0027] Semiconductor process equipment can be chemical vapor deposition (CVD). In CVD equipment, temperature is one of the important parameter indicators affecting the process effect. The consistency between the actual temperature of the temperature zone and the target temperature is directly related to the quality of the process. At the same time, due to the high power output of the power supply during the temperature rise process, it is easy to cause equipment safety problems. Therefore, monitoring the real-time changes of the key variables in the temperature zone and issuing alarms in a timely manner in case of abnormalities is beneficial to ensuring the process index quality and safety of the equipment.

[0028] As Figure 1 shown, during the entire temperature control process, it is mainly divided into 4 stages: room temperature stage - heating stage - high temperature maintenance process stage - cooling stage.

[0029] Based on the above application scenarios, the process data monitoring method provided by the embodiments of the present application will be described in detail below. As Figure 2 shown, it includes step 210 - step 240.

[0030] Step 210, obtain the historical process data in each process step;

[0031] Obtain the historical process data X∈R in each process step N×m ;

[0032] Among them, X is the target parameter data collected at any sampling point, R represents the data set of historical process data, N is the number of sampling points, and m is the number of target parameters; both N and m are positive integers. This formula means that the historical process data is the m target parameter data collected at N sampling points, and the target parameter data is the data corresponding to the target parameter. For example, if the target parameter is temperature, the target parameter data is the temperature value;

[0033] The historical process data includes: the historical parameter values of the target parameter at multiple training time points and the process steps corresponding to each training time point.

[0034] The multiple training time points include the time sequence sampling points of 4 stages: room temperature stage - heating stage - high temperature maintenance process stage - cooling stage.

[0035] Exemplarily, m = 2, that is, the parameter values corresponding to the training time points include:

[0036] As Figure 1 shown, the target set temperature is 65°C, and the change trends of two variables, TC-C and Power, include: from the room temperature stage (sampling times 1 - 52) - heating stage (sampling times 52 - 65) - high temperature maintenance process stage (sampling times 65 - 223) - cooling stage (sampling times 223 - 287).

[0037] At the training time point n = 1, T = 24.1 and P = 0, that is, the historical process data at the sampling point n is (24.1, 0).

[0038] The target parameters include: the temperature of the heating belt TC-C, simply referred to as T; the working output power of the heating belt Power, simply referred to as P. The unit of TC-C is °C, and the unit of Power is %.

[0039] m represents the number of target parameters. When the target parameters include the temperature of the heating belt and the working output power of the heating belt, there are two target parameters, that is, m = 2.

[0040] The sampling point is n = 1. The target parameter data corresponding to the temperature of the heating belt is: 24.1; the target parameter data corresponding to the working output power of the heating belt is 0. Therefore, at the training time point n = 1, the historical process data is (24.1, 0).

[0041] Step 220: Establish an initial Gaussian component model corresponding to each process step; among them, the model parameters to be estimated corresponding to each initial Gaussian component model include: mean vector, covariance matrix, and weight;

[0042] Exemplarily, the parameters θ1 = {μ1, Σ1} of the Gaussian component model corresponding to the room temperature stage 1, where the mean vector of stage 1 represents the clustering center of each data point in this stage, including the means of variables T and P;

[0043] The covariance matrix represents the degree of dispersion between each point, including the autocovariance of variables T and P and the covariance between variables.

[0044] In statistics, the autocovariance of a specific time series or continuous signal is the covariance between the signal and the signal after time translation.

[0045] Step 230: Establish a Gaussian mixture model based on the initial Gaussian component model;

[0046] Establish a Gaussian mixture model (Gaussian Mixture Modle, GMM) based on the initial Gaussian component model.

[0047] Obtain the parameters θ of the Gaussian mixture model characterizing the distribution of each stage i ={μ i ,Σ i}, divide into 4 stages according to the clustering result of the model, and establish Gaussian component models for each stage from the parameters of the Gaussian mixture model.

[0048] The clustering results include: room temperature stage, heating-up stage, high-temperature maintenance process stage, and cooling-down stage.

[0049] Among them, during the entire temperature control process, it is mainly divided into 4 stages: room temperature stage - heating stage - high temperature maintenance process stage - cooling stage. The following is an explanation for each stage respectively:

[0050] In the room temperature stage, the equipment temperature zone is in a non - working state, the output of Power is 0, and TC - C fluctuates within the room temperature range;

[0051] In the heating stage, before the process starts, the heating preparation for each temperature zone will be carried out. The engineer issues the target set value for each temperature zone. Through the heating closed - loop control, the heating Power output approaches 100%, achieving the purpose of rapidly heating TC - C;

[0052] In the high temperature maintenance process stage, when the target set value is reached, it enters the high temperature maintenance process stage. Due to reasons such as the heat dissipation of the heating belt, the heating Power maintains a non - zero low - power output working state to ensure that TC - C maintains a constant temperature state of the set temperature;

[0053] In the cooling stage, the heating Power output is 0, and the equipment is cooled through natural heat dissipation.

[0054] The historical process data is classified through the mixture Gaussian model to obtain the historical process data of the clustering result. The historical process data of one clustering result corresponds to one stage.

[0055] As Figure 4 shown, for the Gaussian component model N(μ k ,Σ k ), k can be 1, 2, 3, and 4.

[0056] Step 240: Train the mixture Gaussian model according to the historical process data until the mixture Gaussian model meets the preset convergence condition, and obtain the Gaussian component model corresponding to each process step.

[0057] Among them, multiple process steps include: process steps in the room temperature stage, process steps in the heating stage, process steps in the high temperature maintenance process stage, and process steps in the cooling stage.

[0058] Among them, in the step of training the mixture Gaussian model according to the historical process data until the mixture Gaussian model meets the preset convergence condition and obtaining the Gaussian component model corresponding to each process step, it can specifically include the following steps:

[0059] Determine the probability density function according to the historical process data corresponding to each process step; an initial Gaussian component model is associated with one process step; the probability density function is used to characterize the probability that the historical process data belongs to each process step;

[0060] Weight the multiple probability density functions based on multiple initial weight values to obtain the mixture model probability density function;

[0061] Determine the posterior probability value according to the probability density function, the model parameters to be estimated, and the initial weight values;

[0062] According to the posterior probability value, perform multiple iterative trainings on the Gaussian mixture model until the probability density function of the mixture model meets the preset convergence condition, and obtain the model parameter values corresponding to each Gaussian component model, so as to obtain the Gaussian component model corresponding to each process step.

[0063] The characteristics of model convergence include: the loss function is less than the preset threshold, the parameters are stable, or the gradient approaches 0.

[0064] When the model converges, the value of the loss function will gradually decrease, indicating that the prediction result of the model is getting closer and closer to the actual value. When the model converges, the parameter values of the model will gradually stabilize. When the model converges, the gradient of the model approaches 0.

[0065] The principle of the Gaussian mixture model is to smoothly approximate any shape density distribution in the weighted form of a finite number of single Gaussian probability density functions. Assume that x ∈ R m is an m-dimensional sample from a multi-stage process, where m represents the number of target parameters, the number of stages K = 4, and determine the probability density function according to the historical process data corresponding to each process step:

[0066]

[0067] Weight multiple probability density functions based on multiple initial weight values to obtain the probability density function of the mixture model:

[0068]

[0069] where K is the number of Gaussian components in the GMM; g(x|θ i ) is the Gaussian density function of the i-th component C i ;

[0070]

[0071] θ i ={μ i ,Σ i} is the parameter set of the i-th Gaussian sub-model, and μ i and Σ i represent the mean vector and covariance matrix of the i-th component respectively;

[0072] θ={θ1,…,θ K}={μ1,Σ1,…,μ K ,Σ K} is the global parameter set composed of K Gaussian sub-models.

[0073] Among them, the parameters to be estimated include: the mean vector μ, the covariance matrix Σ, and the weight ω.

[0074] The parameters to be estimated are obtained through the Expectation Maximization (EM) algorithm.

[0075] The Expectation-maximization algorithm (also translated as the Expectation Maximization algorithm) is an algorithm for finding the maximum likelihood estimate or maximum a posteriori estimate of parameters in a probability model, where the probability model depends on unobservable latent variables.

[0076] The Expectation-maximization algorithm calculates through two steps alternately. The first step is to calculate the expectation (E), using the existing estimated values of the hidden variables to calculate their maximum likelihood estimate values. The second step is to maximize (M), maximizing the maximum likelihood value obtained in the E step to calculate the parameter values. The parameter estimate values found in the M step are used in the next E step calculation, and this process alternates continuously.

[0077] The number of Gaussian components K is the order of the model and also represents the number of stages, which is usually determined by the minimum information length criterion. However, in the process of variable operation in this article, there are clearly 4 stages: the room temperature stage - the heating stage - the high temperature maintenance process stage - the cooling stage. Therefore, K = 4.

[0078] The EM algorithm substitutes each sample point into the model and repeats the E-Step and M-Step to make formula (1) converge to obtain the values of the parameters to be estimated and the posterior probability of the sample point corresponding to a certain Gaussian component.

[0079] The posterior probability refers to the uncertainty of each parameter to be estimated on the basis of having obtained the data of each sampling point.

[0080] For the given training data X ∈ R N×m and the initial parameter values The EM algorithm iterates according to the following steps:

[0081] Among them, according to the probability density function, the model parameters to be estimated, and the initial weight values, the posterior probability value is determined, that is, the E-Step in the EM algorithm: the posterior probability is obtained by the Bayesian rule:

[0082]

[0083] Among them, represents the posterior probability that the j-th training data belongs to the k-th Gaussian component after the s-th iteration.

[0084]

[0085] Among them, respectively represent the mean vector, covariance matrix, and weight of the k-th Gaussian component after the (s + 1)-th iteration.

[0086] Thus, the model parameter values corresponding to each Gaussian component model are obtained to obtain the Gaussian component model corresponding to each process step.

[0087] In the embodiments of the present application, based on the process steps in different stages, the Gaussian mixture model method that can more accurately fit the parameter value fluctuations of the target parameters can be used to divide the temperature zone working stage, and the monitoring model unique to each stage can be obtained. The distributed model established in this way is closer to the actual temperature control process, so as to better master the temperature change trend and discover unknown anomalies.

[0088] Figure 3 It is a flowchart of a process data monitoring method provided by an embodiment of the present application.

[0089] As Figure 3 shown, the process data monitoring method may include step 310 - step 350, which are specifically as follows:

[0090] Step 310, collect the first real-time process data in each process step. The first real-time process data includes: the real-time parameter values of the target parameter at multiple first time points;

[0091] Among them, the training time points are historical time points, and the historical process data corresponding to the training time points is used to train the Gaussian mixture model until the Gaussian mixture model meets the preset convergence condition, and the Gaussian component model corresponding to each process step is obtained.

[0092] The Gaussian component model corresponding to each process step is used to predict the first real-time process data corresponding to the first time point to obtain a process data interval.

[0093] The multiple process steps include: the process steps in the room temperature stage, the process steps in the heating stage, the process steps in the high-temperature maintenance process stage, and the process steps in the cooling stage.

[0094] Among them, the first real-time process data x Online (1×m), where m is a positive integer, and m represents the number of target parameters.

[0095] Step 320, input the first real-time process data into the Gaussian component model corresponding to each process step respectively to obtain multiple probability values; the Gaussian component model is trained according to the historical process data, and the historical process data includes: the historical parameter values of the target parameter at multiple training time points and the process step corresponding to each training time point; one Gaussian component model is associated with one process step;

[0096] Among them, the probability value is used to describe the probability that the first real-time process data belongs to the process step corresponding to the Gaussian component model.

[0097] By calculating a sample point x of the first real-time process data On li ne The posterior probability of each corresponding Gaussian component model, that is, the probability value:

[0098]

[0099] Step 330, determine the Gaussian component model corresponding to the maximum value among multiple probability values as the target Gaussian component model;

[0100] Improve the accuracy of alarm and reduce the maintenance cost caused by a high false alarm rate. When performing offline modeling, fully exploit the information in the historical data under normal operation of the temperature zone to obtain the Gaussian component model corresponding to each process step.

[0101] By inputting the first real-time process data into the Gaussian component model corresponding to each process step respectively, multiple probability values are obtained, and the probability that the first real-time process data belongs to a certain process stage can be determined. By determining the maximum value among multiple probability values, the stage to which the data belongs can be determined. Since one process stage is associated with one Gaussian component model, determining the Gaussian component model corresponding to the maximum value among multiple probability values as the target Gaussian component model can accurately determine the target Gaussian component model applicable to the first real-time process data. Thus, the parameters of each provided Gaussian component model are more accurate, providing a basis for the accuracy of subsequent threshold design and the effectiveness of alarm.

[0102] Determine the Gaussian component model corresponding to the maximum value among multiple probability values as the target Gaussian component model, and the process step associated with the target Gaussian component model is the process step to which the first real-time process data belongs.

[0103] Specifically, it can be determined through formula (6):

[0104]

[0105] Among them, P(x Online ∈C k ), is the probability value, representing the probability that the first real-time process data belongs to a certain process stage. C1 represents the room temperature stage, C2 represents the heating stage, C3 represents the high-temperature maintenance process stage, and C4 represents the cooling stage.

[0106] Formula (6) is used to determine the maximum value among multiple probability values, that is, which process stage the first real-time process data belongs to with the highest probability.

[0107] k can be 1, 2, 3, and 4. K is 4.

[0108] Exemplarily, P(x Online ∈ C1) = 0.7;

[0109] P(x Online ∈ C2) = 0.5;

[0110] P(x Online ∈ C3) = 0.9;

[0111] P(x Online ∈ C4) = 0.6;

[0112] P(x Online ∈ C3) is the maximum value among multiple probability values, then the first real-time process data belongs to the high-temperature maintenance process stage.

[0113] Thus, by establishing a Gaussian mixture model, each historical data point is clustered to obtain the cluster centers, that is, the first real-time process data corresponding to each first time point is automatically divided into each process step. By mining the historical data information in the normal working state, the limitation of designing the alarm threshold only based on expert experience is solved, and the accuracy of the alarm is improved.

[0114] Step 340, input the first real-time process data into the target Gaussian component model to obtain a process data interval;

[0115] As Figure 5 shown, after determining the target Gaussian component model corresponding to the first real-time process data, input the first real-time process data into the target Gaussian component model to obtain a process data interval, and the first real-time process data of one first time point corresponds to one process data interval.

[0116] In a possible embodiment, it is characterized in that the target parameters at least include: the temperature of the heating belt and the working output power of the heating belt; Step 340 includes:

[0117] Input the first real-time process data into the target Gaussian component model to determine the conditional mean and conditional variance corresponding to any one of the target parameters;

[0118] According to the conditional mean and conditional variance, determine the process data interval corresponding to any one of the parameters.

[0119] The more the number of target parameters, the more accurate the output process data interval, which can reserve space for the alarm monitoring of the subsequent electrical addition sensor scheme. The more the number of target parameters, the richer the mined information, and the higher the accuracy of the alarm.

[0120] Embodiments of the present application can dynamically calculate the process data interval for each sampling point. The process data interval is designed to be dynamically variable considering the state of each new sample, which can improve the timeliness of alarms and provide sufficient time for subsequent safety actions.

[0121] Thus, the first real-time process data corresponding to each first time point can be automatically divided into Gaussian component models corresponding to each process step, so as to input the first real-time process data into the target Gaussian component model to obtain the process data interval corresponding to the real-time parameter value of each first time point, solving the limitation of designing the process data interval only based on manual experience and improving the accuracy of the process data interval.

[0122] Among them, the target parameters may further include: the central temperature of the heating belt and the edge temperature of the heating belt.

[0123] The CVD equipment is a preparation equipment that uses gaseous compounds or mixtures of compounds to undergo chemical reactions on the heated surface of the substrate, thereby forming a non-volatile coating on the surface of the substrate. The heating belt refers to the component used for heating in the CVD equipment.

[0124] Among them, in the step of inputting the first real-time process data into the target Gaussian component model to determine the conditional mean and conditional variance of any one of the target parameters, it may specifically include the following steps:

[0125] Divide the first real-time process data into a first parameter value and a second parameter value. The second parameter value is the parameter value in the first real-time process data other than the first parameter value, and the first parameter value is the parameter value corresponding to any one of the parameters;

[0126] Input the first parameter value and the second parameter value into the target Gaussian component model to determine the mean vector and the covariance matrix;

[0127] According to the mean vector and the covariance matrix, determine the conditional mean and conditional variance of any one of the parameters.

[0128] First, divide the first real-time process data into a first parameter value and a second parameter value, that is, x (1) and all other variables x (1) except x (2) , that is, where x (1) ∈R 1 , x (2) ∈R m-1 ,

[0129] Among them, the Gaussian component model N(μ k ,Σ k ).

[0130] Then, input the first parameter value and the second parameter value into the target Gaussian component model to determine the mean vector and the covariance matrix.

[0131] Wherein, the mean vector is:

[0132]

[0133] The covariance matrix is:

[0134]

[0135] Wherein, Σ ij represents the covariance matrix of variable x (i) and variable x (j) , i, j = 1, 2.

[0136] Since variable x (1) follows a conditional Gaussian distribution when the values of other variable groups x (2) are determined:

[0137] x (1) |x (2) ~N(μ 1|2 , Σ 1|2 ) (9)

[0138] According to the mean vector and the covariance matrix, determine the conditional mean and the conditional variance corresponding to any parameter. Among them, the conditional mean is:

[0139]

[0140] The conditional variance is:

[0141]

[0142] Among them, in the step of determining the process data interval corresponding to any parameter according to the conditional mean and the conditional variance, it may specifically include the following steps:

[0143] Calculate the first threshold and the second threshold according to the preset confidence level, the conditional mean and the conditional variance;

[0144] Determine the process data interval corresponding to any parameter according to the first threshold and the second threshold. The process data interval includes the first threshold and the second threshold, and the first threshold is greater than the second threshold.

[0145] The confidence level represents the probability that normal data sample points are in the corresponding confidence interval. If it is too high, the confidence interval range is small and false alarms are likely to occur. If it is too low, the confidence interval range is large and alarms cannot be thrown.

[0146] Since a probability less than 0.05 is considered a small probability event, in the embodiments of the present application, a confidence level of 95% can be taken while satisfying the alarm accuracy and the tolerance rate.

[0147] According to the preset confidence level, conditional mean, and conditional variance, calculate the first threshold and the second threshold. According to the first threshold and the second threshold, determine that the process data interval corresponding to any parameter is (μ 1|2 - 2Σ 1|2 , μ 1|2 + 2Σ 1|2 ).

[0148] Set the x Online in the sample point x (1) at the second threshold as μ 1|2 - 2Σ 1|2 , and set the first threshold as μ 1|2 + 2Σ 1|2 . Similarly, the second threshold Online of the x (i) corresponding to the first real-time process data of any other target parameter of the sample point x and the first threshold

[0149] Step 350, in the case where the second real-time process data is not within the process data interval, output a prompt message. The second real-time process data includes: the real-time parameter values of the target parameter at multiple second time points, and the multiple second times are after multiple first time points.

[0150] In the case where the second real-time process data is not within the process data interval, output a prompt message to provide a more accurate alarm for the abnormality during the process step switching.

[0151] By designing dynamic alarm thresholds, the false alarm rate and the missed alarm rate can be reduced, and the accurate alarm requirements under the on-line monitoring of the temperature zone can be met, providing guarantee for the process parameter quality and equipment safety.

[0152] In a possible embodiment, it is characterized in that step 350 includes:

[0153] In the case where the real-time parameter values of the target parameter at multiple second time points are greater than the first threshold, output a first prompt message;

[0154] In the case where the real-time parameter values of the target parameter at multiple second time points are less than the second threshold, output a second prompt message.

[0155] The moment of the alarm and the corresponding variable sensor are beneficial to timely determine the location where the fault occurs during the process.

[0156] Among them, considering that the risk of ultra-high alarm is higher, the real-time parameter values of the target parameter at multiple second time points and the first threshold are preferentially determined. If the real-time parameter values x of the target parameter at multiple second time points (i) are greater than the first threshold then an alarm is directly issued; if not, then the real-time parameter values of the target parameter at multiple second time points and the second threshold are then determined. If x (i) is less than or equal to the second threshold then an alarm is issued, and the other is in a normal state.

[0157] Thus, dynamic high alarm thresholds and low alarm thresholds are designed according to the characteristics of each different stage. When the real-time parameter value is greater than the first threshold, a first prompt message is output; when the real-time parameter value is less than the second threshold, a second prompt message is output, providing a more comprehensive abnormal temperature zone monitoring.

[0158] According to the embodiments of the present application, by collecting the first real-time process data in each process step, the first real-time process data includes: the real-time parameter values of the target parameter at multiple first time points; the first real-time process data is respectively input into the Gaussian component models corresponding to each process step to obtain multiple probability values; the Gaussian component models are trained according to historical process data, and the historical process data includes: the historical parameter values of the target parameter at multiple training time points and the process step corresponding to each training time point; one Gaussian component model is associated with one process step; the Gaussian component model corresponding to the maximum value among the multiple probability values is determined as the target Gaussian component model; thus, the first real-time process data corresponding to each first time point can be automatically divided into the Gaussian component models corresponding to each process step, so as to input the first real-time process data into the target Gaussian component model to obtain the process data interval corresponding to the real-time parameter value of each first time point, solving the limitation of designing the process data interval only based on manual experience and improving the accuracy of the process data interval. Thus, when the second real-time process data collected at time points after the first time point, that is, multiple second time points, is not within the process data interval, a prompt message is output, which can accurately monitor the second real-time process data and timely and accurately discover problems in the processing steps.

[0159] It should be noted that for the method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the embodiments of the present application are not limited by the described action sequence, because according to the embodiments of the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0160] Refer to Figure 6, showing a structural block diagram of a semiconductor processing apparatus according to an embodiment of the present application. The semiconductor processing apparatus 610 includes: a process chamber, a gas inlet assembly, a heater, a temperature measurement component, and a controller 611; the semiconductor processing apparatus 610 may be a CVD apparatus.

[0161] Among them, the controller 611 is configured to collect first real-time process data in each process step. The first real-time process data includes: real-time parameter values of the target parameter at a plurality of first time points.

[0162] Input the first real-time process data into the Gaussian component model corresponding to each process step respectively to obtain a plurality of probability values; the Gaussian component model is trained according to historical process data. The historical process data includes: historical parameter values of the target parameter at a plurality of training time points and the process step corresponding to each training time point; one Gaussian component model is associated with one process step.

[0163] Determine the Gaussian component model corresponding to the maximum value among the plurality of probability values as the target Gaussian component model.

[0164] Input the first real-time process data into the target Gaussian component model to obtain a process data interval.

[0165] In the case where the second real-time process data is not within the process data interval, output a prompt message. The second real-time process data includes: real-time parameter values of the target parameter at a plurality of second time points, and the plurality of second times are after the plurality of first time points.

[0166] In an optional embodiment of the present application, the controller 611 is further configured to

[0167] Obtain historical process data in each process step.

[0168] Establish an initial Gaussian component model corresponding to each process step; among them, the model parameters to be estimated corresponding to each initial Gaussian component model include: mean vector, covariance matrix, and weight.

[0169] Establish a mixture Gaussian model based on the initial Gaussian component model.

[0170] Train the mixture Gaussian model according to the historical process data until the mixture Gaussian model meets the preset convergence condition to obtain the Gaussian component model corresponding to each process step.

[0171] In an optional embodiment of the present application, the controller 611 is further configured to determine a probability density function according to the historical process data corresponding to each process step; one initial Gaussian component model is associated with one process step; the probability density function is used to characterize the probability that the historical process data belongs to each process step.

[0172] Weight multiple probability density functions based on multiple initial weight values to obtain a mixture model probability density function;

[0173] Determine posterior probability values according to the probability density function, the model parameters to be estimated, and the initial weight values;

[0174] Perform multiple iterative trainings on the mixture Gaussian model according to the posterior probability values until the mixture model probability density function meets the preset convergence condition, and obtain the model parameter values corresponding to each Gaussian component model, so as to obtain the Gaussian component model corresponding to each process step.

[0175] In an optional embodiment of the present application, the controller 611 is further configured to input the first real-time process data into the target Gaussian component model to determine the conditional mean and conditional variance corresponding to any one of the target parameters;

[0176] Determine the process data interval corresponding to any one of the parameters according to the conditional mean and conditional variance.

[0177] In an optional embodiment of the present application, the controller 611 is further configured to divide the first real-time process data into a first parameter value and a second parameter value, where the second parameter value is the parameter value other than the first parameter value in the first real-time process data, and the first parameter value is the parameter value corresponding to any one of the parameters;

[0178] Input the first parameter value and the second parameter value into the target Gaussian component model to determine the mean vector and the covariance matrix;

[0179] Determine the conditional mean and conditional variance corresponding to any one of the parameters according to the mean vector and the covariance matrix.

[0180] In an optional embodiment of the present application, the controller 611 is further configured to calculate a first threshold and a second threshold according to the preset confidence level, the conditional mean, and the conditional variance;

[0181] Determine the process data interval corresponding to any one of the parameters according to the first threshold and the second threshold. The process data interval includes the first threshold and the second threshold, and the first threshold is greater than the second threshold.

[0182] In an optional embodiment of the present application, the controller 611 is further configured to output a first prompt message when the real-time parameter values of the target parameter at multiple second time points are greater than the first threshold;

[0183] Output a second prompt message when the real-time parameter values of the target parameter at multiple second time points are less than the second threshold.

[0184] According to an embodiment of the present application, by collecting first real-time process data in each process step, the first real-time process data includes: real-time parameter values of a target parameter at multiple first time points; inputting the first real-time process data into Gaussian component models corresponding to each process step respectively to obtain multiple probability values; the Gaussian component models are trained according to historical process data, and the historical process data includes: historical parameter values of the target parameter at multiple training time points and the corresponding process step for each training time point; one Gaussian component model is associated with one process step; determining the Gaussian component model corresponding to the maximum value among the multiple probability values as the target Gaussian component model; thereby, the first real-time process data corresponding to each first time point can be automatically divided into Gaussian component models corresponding to each process step, so as to input the first real-time process data into the target Gaussian component model to obtain a process data interval corresponding to the real-time parameter value of each first time point, solving the limitation of designing the process data interval only based on manual experience and improving the accuracy of the process data interval. Thus, in the case that the second real-time process data collected at time points after the first time point, i.e., multiple second time points, is not within the process data interval, a prompt message is output, which can accurately monitor the second real-time process data and timely and accurately discover problems in the processing steps.

[0185] For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For related parts, refer to the partial description of the method embodiment.

[0186] The embodiment of the present application further provides an electronic device, including: a processor, a memory, and a computer program stored on the memory and capable of running on the processor. When the computer program is executed by the processor, it realizes each process of the above-mentioned method embodiment for monitoring process data and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0187] The embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it realizes each process of the above-mentioned method embodiment for monitoring process data and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0188] For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For related parts, refer to the partial description of the method embodiment.

[0189] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other.

[0190] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, apparatuses, or computer program products. Therefore, the embodiments of the present application can take the form of completely hardware embodiments, completely software embodiments, or embodiments combining software and hardware aspects. Moreover, the embodiments of the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0191] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing terminal devices generate a device for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0192] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0193] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, such that a series of operation steps are executed on the computer or other programmable terminal device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0194] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to these embodiments once they know the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0195] Finally, it should also be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or terminal device comprising the said element.

[0196] The above has introduced in detail a process data monitoring method and a semiconductor process device provided by the present application. Specific examples are used in this text to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A process data monitoring method, applied to semiconductor process equipment, wherein the process of the semiconductor process equipment includes a plurality of sequentially performed process steps, characterized in that, The described process data monitoring method includes: Collecting first real-time process data in each of the described process steps, where the first real-time process data includes: real-time parameter values of a target parameter at multiple first time points; Inputting the first real-time process data into the Gaussian component model corresponding to each of the process steps respectively to obtain multiple probability values; the Gaussian component model is trained based on historical process data, and the historical process data includes: historical parameter values of the target parameter at multiple training time points and the process step corresponding to each training time point; one Gaussian component model is associated with one process step; Determining the Gaussian component model corresponding to the maximum value among the multiple probability values as the target Gaussian component model; Inputting the first real-time process data into the target Gaussian component model to obtain a process data interval; In the case where the second real-time process data is not within the process data interval, outputting a prompt message, where the second real-time process data includes: real-time parameter values of the target parameter at multiple second time points, and the multiple second time points are after the multiple first time points.

2. The method according to claim 1, characterized in that, Before the step of inputting the first real-time process data into the Gaussian component model corresponding to each of the process steps respectively to obtain multiple probability values, the method further includes: Obtaining the historical process data in each of the process steps; Establishing an initial Gaussian component model corresponding to each of the process steps; where the model parameters to be estimated corresponding to each initial Gaussian component model include: mean vector, covariance matrix, and weight; Establishing a mixture Gaussian model based on the initial Gaussian component model; Training the mixture Gaussian model based on the historical process data until the mixture Gaussian model meets a preset convergence condition to obtain the Gaussian component model corresponding to each process step.

3. The method according to claim 2, wherein The step of training the mixture Gaussian model based on the historical process data until the mixture Gaussian model meets a preset convergence condition to obtain the Gaussian component model corresponding to each process step includes: Determining a probability density function according to the historical process data corresponding to each process step; one initial Gaussian component model is associated with one process step; the probability density function is used to characterize the probability that the historical process data belongs to each process step; Weighting the multiple probability density functions based on multiple initial weight values to obtain a mixture model probability density function; Determining a posterior probability value according to the probability density function, the model parameters to be estimated, and the initial weight values; Performing multiple iterative trainings on the mixture Gaussian model according to the posterior probability value until the mixture model probability density function meets the preset convergence condition to obtain the model parameter values of the model parameters corresponding to each Gaussian component model, so as to obtain the Gaussian component model corresponding to each process step.

4. The method according to claim 1, wherein The target parameter at least includes: heating belt temperature and heating belt working output power; the step of inputting the first real-time process data into the target Gaussian component model to obtain a process data interval includes: Input the first real-time process data into the target Gaussian component model to determine the conditional mean and conditional variance corresponding to any one of the target parameters; Determine the process data interval corresponding to any one of the parameters according to the conditional mean and the conditional variance.

5. The method according to claim 4, wherein The step of inputting the first real-time process data into the target Gaussian component model to determine the conditional mean and conditional variance corresponding to any one of the target parameters includes: Divide the first real-time process data into a first parameter value and a second parameter value, where the second parameter value is the parameter value in the real-time process data other than the first parameter value, and the first parameter value is the parameter value corresponding to any one of the parameters; Input the first parameter value and the second parameter value into the target Gaussian component model to determine the mean vector and the covariance matrix; Determine the conditional mean and conditional variance corresponding to any one of the parameters according to the mean vector and the covariance matrix.

6. The method according to claim 4, wherein The step of determining the process data interval corresponding to any one of the parameters according to the conditional mean and the conditional variance includes: Calculate a first threshold and a second threshold according to a preset confidence level, the conditional mean, and the conditional variance; Determine the process data interval corresponding to any one of the parameters according to the first threshold and the second threshold. The process data interval includes the first threshold and the second threshold, and the first threshold is greater than the second threshold.

7. The method according to claim 6, wherein The step of outputting a prompt message when the second real-time process data is not within the process data interval includes: Output a first prompt message when the real-time parameter values of the target parameter at multiple second time points are greater than the first threshold; Output a second prompt message when the real-time parameter values of the target parameter at multiple second time points are less than the second threshold.

8. The method according to claim 1, wherein The multiple process steps include: process steps in the room temperature stage, process steps in the heating stage, process steps in the high-temperature maintenance process stage, and process steps in the cooling stage.

9. A semiconductor process equipment, characterized in that, The process of the semiconductor process equipment includes multiple sequentially performed process steps. The semiconductor process equipment includes: A controller for collecting the first real-time process data in each process step. The first real-time process data includes: the real-time parameter values of the target parameter at multiple first time points; Input the first real-time process data into the Gaussian component model corresponding to each process step respectively to obtain multiple probability values; the Gaussian component model is trained according to historical process data. The historical process data includes: the historical parameter values of the target parameter at multiple training time points and the process step corresponding to each training time point; one Gaussian component model is associated with one process step; Determine the Gaussian component model corresponding to the maximum value among the multiple probability values as the target Gaussian component model; Input the first real-time process data into the target Gaussian component model to obtain a process data interval; In the case where the second real-time process data is not within the process data range, a prompt message is output, where the second real-time process data includes: real-time parameter values of the target parameter at a plurality of second time points, and the plurality of second times are after the plurality of first time points.

10. A semiconductor process equipment, comprising a process chamber, an intake component, a heater, a temperature measuring component and a controller, characterized in that, The controller includes at least one processor and at least one memory, and a computer program is stored in the memory. When the computer program is executed by the processor, the method described in any one of claims 1-8 is implemented.

Images