Data detection method and system, detection device, computer equipment and storage medium

Abstract

The application provides a data detection method and system, a detection device, computer equipment and a storage medium. The application obtains detection signals of surfaces of each wafer in a planetary reaction cavity, cuts the detection signals based on a trigger signal to obtain a first signal sequence and a second signal sequence; processes the first signal sequence by using a classification model to obtain a material type on each wafer; queries a corresponding optical model based on the material type; and iteratively optimizes the optical model, the first signal sequence and the second signal sequence by using a decoupling algorithm to obtain a temperature detection value and a reflectivity detection value. The application divides the original detection signal into separate signal sequences based on the trigger signal, then determines the material type according to the separate signal sequences, selects the corresponding optical model according to the material type, and finally decouples the optical model to obtain the temperature detection value and the reflectivity detection value with high accuracy, thereby realizing accurate analysis of a single wafer signal and greatly improving process repeatability and stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor detection, and in particular to a data detection method and system, a detection device, a computer device and a storage medium. BACKGROUND

[0002] Metal organic chemical vapor deposition (MOCVD) equipment is the core equipment for preparing compound semiconductor (such as GaAs, GaN) epitaxial wafers. In a typical planetary reaction cavity, a main graphite disc revolves under the drive of a motor, and a plurality of satellite discs carried on the main graphite disc also rotate under the push of gas flow. After the cavity is heated to a certain temperature and chemical gas is introduced, a chemical reaction occurs, and a thin film layer required by a semiconductor device is deposited on the wafer surface, thereby growing the thin film layer. The composite motion of revolution and rotation aims to realize the uniformity of temperature and reaction gas distribution in the reaction cavity, thereby ensuring that the thickness and composition of the thin film grown on each wafer carrier disc are consistent.

[0003] In the adjustment process before epitaxial growth begins, a standard wafer and a companion wafer are usually placed in the equipment to determine whether the equipment cavity is operating normally. When different materials such as germanium Ge, gallium arsenide GaAs and sapphire Sapphire substrates are placed on the plurality of satellite discs, the emissivity and reflectivity of the different material wafers differ greatly.

[0004] Therefore, the design of the planetary turntable brings both the benefit of good process uniformity and an inherent problem for optical detection: the coupling of the measurement field of view and the motion trajectory. The measurement field of view of a fixedly installed optical detector (such as an infrared pyrometer or a thermal imager) is fixed. When revolution and rotation are performed simultaneously, the optical signal received by the detector at any time is the sum of the radiant energy of all wafer surfaces within the field of view of the detector. In the case of coexistence of different substrates, this signal is a dynamically changing mixture of radiation from substrates with different emissivities and temperatures. Traditional techniques usually simply regard this mixed signal as a whole and inversely calculate the temperature based on an "average" or "equivalent" emissivity. This method completely ignores the physical nature of different materials, and the inversely calculated temperature is a value without clear physical meaning; or simply averages, the average value of the signal within a period of time is obtained, but the average value is still a mixed result weighted by the proportions of various materials, and the information of a single material cannot be stripped out, resulting in inaccurate inversely calculated temperature and reflectivity values that cannot truly represent the process state of any specific wafer. SUMMARY

[0005] The present application aims to provide a data detection method and system, a detection device, a computer device and a storage medium to overcome the problem that the detection signal obtained in the traditional technology is a mixed signal, resulting in inaccurate inversely calculated temperature and reflectivity values.

[0006] In a first aspect, the application provides a data detection method, the method comprising: obtaining a detection signal of a wafer surface; wherein each wafer is arranged in a planetary reaction cavity, and the detection signal comprises a spectral reflection signal and a thermal radiation signal; cutting the spectral reflection signal and the thermal radiation signal based on a trigger signal to obtain a first signal sequence and a second signal sequence; wherein one trigger signal represents one rotation of a graphite disc in the planetary reaction cavity; processing the first signal sequence using a pre-trained classification model to obtain a material type on each wafer; based on the material type, querying a pre-set database to obtain an optical model corresponding to the material type; and iteratively optimizing the optical model, the first signal sequence, and the second signal sequence using a decoupling algorithm to obtain a temperature detection value and a reflectivity detection value.

[0007] In one embodiment, the cutting of the spectral reflection signal and the thermal radiation signal based on the trigger signal to obtain the first signal sequence and the second signal sequence comprises: for each trigger signal, determining a time interval for each satellite disc to reach a preset position, taking the trigger signal as a time starting point; taking a plurality of time points within the time interval, and setting a time window centered on each time point to cut a signal segment corresponding to each time window from the spectral reflection signal and the thermal radiation signal; the width of the time window satisfies that the detection signals of all wafers on the satellite disc are completely captured; and splicing the signal segments belonging to the same satellite disc in time sequence to obtain the first signal sequence and the second signal sequence of each satellite disc.

[0008] In one embodiment, the classification model comprises a convolution network, a support vector machine network, and a random forest network; and the processing of the first signal sequence using the pre-trained classification model to obtain the material type on each wafer comprises: extracting a feature vector representing the optical characteristics of different materials from the first signal sequence; inputting the feature vector into the convolution network, the support vector machine network, and the random forest network to obtain a first classification result, a second classification result, and a third classification result; and performing weighted summation on the first classification result, the second classification result, and the third classification result to determine the material type according to the summation result.

[0009] In one of the embodiments, the feature vector includes a time domain feature vector and a frequency domain feature vector; the feature vector representing the optical characteristics of different materials is extracted from the first signal sequence, including: cutting the first signal sequence to obtain a sub-signal sequence of each wafer on the satellite disk; based on each of the sub-signal sequences, a time domain feature vector is extracted, including at least one of the average value, the signal intensity, and the number; performing Fourier transform on each of the sub-signal sequences to extract the main frequency component to obtain a frequency domain feature vector, including the growth rate and / or the film thickness.

[0010] In one of the embodiments, the cutting of the first signal sequence to obtain a sub-signal sequence of each wafer on the satellite disk includes: detecting a pulse peak in the first signal sequence; wherein one of the pulse peaks corresponds to one wafer; determining the starting point and the ending point of each pulse peak; based on the starting point and the ending point of each pulse peak, the first signal sequence is cut to obtain a plurality of signal segments; the signal intensity of each signal segment is calculated, and each signal segment is classified into different categories according to the signal intensity; the signal segments belonging to the same category and obtained by continuous acquisition are spliced in time sequence to obtain the sub-signal sequence of each wafer.

[0011] In one of the embodiments, the feature vector further includes a model feature vector; the feature vector representing the optical characteristics of different materials is extracted from the first signal sequence, including: calculating the similarity between the curve of the first signal sequence changing with time and the theoretically predicted curve of each candidate material in the preset database to obtain a plurality of initial correlation coefficients, each of the initial correlation coefficients constituting the model feature vector.

[0012] In one of the embodiments, the optical model is used to characterize an optical constant function of an optical constant with respect to wavelength, temperature, the optical constant including refractive index and extinction coefficient; the iterative optimization of the optical model, the first signal sequence and the second signal sequence by using the decoupling algorithm to obtain a temperature detection value and a reflectivity detection value includes: generating predicted data under a given temperature and surface state parameter based on the optical constant function of the material by using a pre-constructed physical forward model; wherein the predicted data includes theoretical reflectivity and theoretical radiation intensity; the surface state parameter includes at least one of an emissivity correction coefficient, a surface roughness parameter and an oxidation layer influence parameter; calculating a residual error of the predicted data and measured data step by step, the measured data including measured reflectivity and measured radiation intensity, and the measured reflectivity being obtained based on the sub-signal sequence, and the measured radiation intensity being obtained based on the second signal sequence; using a decoupled alternating optimization algorithm to alternately and iteratively optimize the temperature and the surface state parameter, and using a nonlinear least squares algorithm to minimize a residual error function until convergence; taking the temperature at the model convergence as the temperature detection value, and correcting the measured reflectivity based on the surface state parameter at the convergence to obtain the reflectivity detection value.

[0013] In one of the embodiments, the method further includes: displaying a curve of the temperature value and the reflectivity value of each wafer over time, and generating an alarm signal when the curve is out of a process range.

[0014] In a second aspect, the present application provides a data detection system, the system comprising: an acquisition module configured to acquire detection signals of wafer surfaces; wherein each wafer is arranged in a planetary reaction cavity, and the detection signals include spectral reflectance signals and thermal radiation signals; a signal segmentation module configured to cut the spectral reflectance signals and the thermal radiation signals based on trigger signals to obtain first signal sequences and second signal sequences; wherein one trigger signal represents one rotation of a graphite disc in the planetary reaction cavity; a processing module configured to process each first signal sequence by using a pre-trained classification model to obtain a material type on each wafer; query an optical model corresponding to the material type from a pre-set database based on the material type; and perform iterative optimization of the optical model, the first signal sequence and the second signal sequence by using a decoupling algorithm to obtain a temperature detection value and a reflectivity detection value.

[0015] In a third aspect, the present application also provides a detection device, comprising: an optical measurement module, configured to emit a light beam to the surface of each wafer in the planetary reaction cavity and collect a spectral reflection signal of the surface of each wafer; a temperature measurement module, configured to detect a thermal radiation signal of the surface of each wafer; a signal line, connected with a spindle encoder or a motor controller of a planetary turntable machine, configured to transmit a trigger signal; wherein one trigger signal represents one rotation of the graphite disc in the planetary reaction cavity; the data detection system as described in the second aspect, configured to periodically process the spectral reflection signal and the thermal radiation signal according to the trigger signal, to obtain a temperature detection value and a reflectivity detection value.

[0016] In a fourth aspect, the present application also provides a computer device. The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method steps in the first aspect when executing the computer program.

[0017] In a fifth aspect, the present application also provides a computer readable storage medium. The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method steps in the first aspect.

[0018] The above data detection method and system, detection device, computer device and storage medium have at least the following advantages:

[0019] The present application obtains detection signals of the surfaces of each wafer in the planetary reaction cavity, the detection signals comprising a spectral reflection signal and a thermal radiation signal; the spectral reflection signal and the thermal radiation signal are cut based on a trigger signal to obtain a first signal sequence and a second signal sequence; wherein one trigger signal represents one rotation of the graphite disc in the planetary reaction cavity; a pre-trained classification model is used to process the first signal sequence to obtain the material type on each wafer; based on the material type, an optical model corresponding to the material type is queried from a preset database; a decoupling algorithm is used to iteratively optimize the optical model, the first signal sequence and the second signal sequence to obtain a temperature detection value and a reflectivity detection value. By using the above scheme, the present application divides the original detection signal into separate signal sequences based on the trigger signal, then determines the material type according to the separate signal sequences, selects the corresponding optical model according to the material type, and finally decouples the optical model to obtain the temperature detection value and the reflectivity detection value with high accuracy, which realizes accurate analysis of a single wafer signal and greatly improves the process repeatability and stability. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 It is a structural block diagram of the detection device in one embodiment;

[0021] Figure 2 It is a flowchart of the data detection method in one embodiment;

[0022] Figure 3 Flowchart of the process of cutting detection data step in one embodiment;

[0023] Figure 4 Flowchart of the process of obtaining material type step in one embodiment;

[0024] Figure 5 Flowchart of the process of extracting feature vector step in one embodiment;

[0025] Figure 6 Flowchart of the process of decoupling algorithm step in one embodiment;

[0026] Figure 7 Flowchart of the process of data detection method detailed description in one embodiment;

[0027] Figure 8 Structure block diagram of data detection system in one embodiment;

[0028] Figure 9 Internal structure diagram of computer device in one embodiment. DETAILED DESCRIPTION

[0029] The present application is described by specific embodiments, and those skilled in the art can easily understand other advantages and effects of the present application from the disclosure of the specification. The present application can also be implemented or applied by other different specific embodiments, and the details in the specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0030] Some exemplary embodiments of the present application are described for illustrative purposes, and it should be understood that the present application can be implemented in other ways not specifically shown in the drawings.

[0031] Please refer to Figure 1Optionally, in one embodiment, the application provides a detection device arranged above the planetary reaction cavity 1, wherein the planetary reaction cavity 1 is provided with a rotatable graphite disc 11, and the graphite disc 11 is provided with a plurality of self-rotating star discs 12, and each star disc 12 is provided with a plurality of wafer carrier discs 13 for placing wafers, and the graphite disc 11 is controlled to rotate by a motor, and each star disc 12 is self-rotated by air flow. After the cavity is heated to a certain temperature, chemical gas is introduced, chemical reaction occurs, and thin film layer required by semiconductor devices is grown on the wafer surface. In this embodiment, at least one standard piece and a plurality of accompanying pieces are placed on the wafer carrier disc 13, and the temperature and reflectivity of the surface of the standard piece are detected during the adjustment process before epitaxial growth starts, so as to determine whether the planetary reaction cavity 1 is operating normally. For the convenience of description, the standard piece and the accompanying piece are referred to as wafers hereinafter.

[0032] The detection device comprises an optical measurement module 2, a temperature measurement module 3, a signal line 4 and a data detection system 5.

[0033] The optical measurement module 2 is arranged above the planetary reaction cavity 1, and is used for emitting light beams to the surfaces of the wafers in the planetary reaction cavity 1 and collecting the spectral reflection signals of the surfaces of the wafers.

[0034] The optical measurement module 2 comprises a transmitting unit and a receiving unit.

[0035] The transmitting unit comprises at least one LED light source for generating a measurement light beam, and a transmitting optical fiber for guiding the measurement light beam to a probe, and the light beam collimated and converged by the probe is projected to the surfaces of the rotating wafers through the observation window of the planetary reaction cavity 1.

[0036] The receiving unit comprises a receiving optical fiber for collecting the light beams reflected from the surfaces of the wafers, and at least one APD detector for converting the received reflected light signals into analog electrical signals.

[0037] The temperature measurement module 3 is used for detecting thermal radiation signals of the surfaces of the wafers. Exemplarily, the temperature measurement module 3 can adopt a pyrometer.

[0038] The signal line 4 is connected at one end to a main shaft encoder or a motor controller of the planetary turntable machine, and is connected at the other end to the data detection system 5, and is used for transmitting trigger signals and transmitting to the data detection system 5; wherein one trigger signal represents that the graphite disc 11 in the planetary reaction cavity 1 rotates one circle. The trigger signal is strictly synchronous with the rotation angle of the graphite disc 11, which ensures that the data acquisition is not deviated from the locking of the physical position of the graphite disc 11.

[0039] The data detection system 5 is used for periodically processing the spectral reflection signals and the thermal radiation signals according to the trigger signals, so as to obtain temperature detection values and reflectivity detection values.

[0040] Specifically, the data detection system 5 receives the detection signal of the wafer surface; wherein the detection signal includes a spectral reflection signal and a thermal radiation signal; the spectral reflection signal and the thermal radiation signal are cut based on the trigger signal respectively, to obtain a first signal sequence and a second signal sequence; a pre-trained classification model is used to process the first signal sequence to obtain the material type on each wafer; based on the material type, the optical model corresponding to the material type is queried from the preset database; and the optical model, the first signal sequence and the second signal sequence are iteratively optimized by using a decoupling algorithm to obtain the temperature detection value and the reflectivity detection value.

[0041] The above detection device, the optical measurement module is used for collecting the spectral reflection signal, the temperature measurement module is used for collecting the thermal radiation signal, the signal line is used for transmitting the trigger signal in real time, and the data detection system is used for periodically processing the spectral reflection signal and the thermal radiation signal based on the trigger signal to obtain the temperature detection value and the reflectivity detection value. By using the above scheme, the accurate analysis of a single wafer signal is realized based on hardware triggering and software processing, and the process repeatability and stability are greatly improved.

[0042] Please refer to Figure 2 In one embodiment, the present application provides a data detection method, the method comprising:

[0043] Step 202, obtaining the detection signal of the wafer surface; wherein each wafer is arranged in a planetary reaction cavity, and the detection signal includes a spectral reflection signal and a thermal radiation signal.

[0044] Specifically, each wafer is arranged on a wafer carrier plate, and is arranged according to a set position according to test needs.

[0045] It should be noted that the above detection signal has been converted into a digital electrical signal.

[0046] Further, since the detection signal is the sum of the radiant energy of all wafer surfaces in the field of view of the optical detector, regarding this mixed signal as a whole will lead to inaccuracy in monitoring, which is specifically manifested in:

[0047] Temperature inversion inaccuracy: the basic formula of radiation temperature measurement (Planck blackbody radiation law) requires the emissivity of the object to be known to calculate the true temperature from the measured radiation intensity. The emissivity of different materials differs greatly, for example, the emissivity of Ge at high temperature may be higher than 0.6, while the emissivity of sapphire may be lower than 0.3, which means that for the same true temperature, they show completely different radiation intensities. Treating the mixed signal with one emissivity will cause the inverted temperature to deviate seriously from the true temperature of all substrates.

[0048] Reflectance / emittance measurement distortion: When measuring film thickness and reflectance with a laser reflectometer or interferometry, the interference of background thermal radiation and the confusion of signal sources can distort the measurement values. Without distinguishing the signal sources, it is impossible to accurately determine the real reflectance change of each material surface, so as to accurately monitor the growth rate and surface morphology of the thin film.

[0049] Process state misjudgment: inaccurate temperature and reflectance data cannot truly represent the process state of any specific wafer. Equipment engineers may make misjudgments based on these false data, considering that the equipment cavity is normal or abnormal, which may lead to starting formal production when the equipment state is poor, causing large quantities of products to be scrapped, and unnecessary maintenance and adjustment when the equipment state is normal, reducing equipment utilization.

[0050] Step 204, based on the trigger signal, cutting the spectral reflectance signal and the thermal radiation signal respectively to obtain the first signal sequence and the second signal sequence; wherein, one trigger signal represents that the graphite disc in the planetary reaction cavity rotates one circle.

[0051] Specifically, in the original detection signal, the spectral reflectance signal and the thermal radiation signal are mixed detection signals of all wafers. In the case of different wafer surface materials, the emissivity and reflectivity of different material wafers differ greatly. If the collected optical signal is treated as a mixed whole or simply averaged, the temperature and reflectance obtained by inversion will have serious deviations.

[0052] Based on this, the present application associates the timing of data processing with the rotation angle of the graphite disc by setting a hardware trigger signal. One trigger signal is generated when the graphite disc rotates one circle. Taking the trigger signal as the time starting point, the timing relationship is determined according to the position of each star-shaped disc, so as to realize the cutting of the original detection data.

[0053] Step 206, using a pre-trained classification model to process the first signal sequence to obtain the material type on each wafer; based on the material type, querying the optical model corresponding to the material type from the preset database.

[0054] Specifically, the classification model is used to infer the possibility of the material type on the wafer by analyzing the reflectance spectrum signal. It should be understood that the classification model is trained based on historical detection data, historical temperature and historical reflectance.

[0055] The preset database is set in the process recipe in advance based on the generated plan. Each satellite disc corresponds to a candidate material list, which lists all material types that can be placed on the satellite disc. For example, the list of satellite discs is Ge, GaAs and Sappire. Further, the preset database includes material name, wavelength range, temperature range, refractive index (n) data set, extinction coefficient (k) data set, surface state parameter initial value or value range, etc., and is pre-set with theoretical reflectivity and thermal radiation model. Among them, the surface state parameters include at least one of emissivity correction coefficient, surface roughness parameter and oxidation layer influence parameter.

[0056] In step 208, a decoupling algorithm is used to iteratively optimize the optical model, the first signal sequence and the second signal sequence to obtain the temperature detection value and the reflectivity detection value.

[0057] Specifically, the decoupling algorithm refers to separating two or more physical quantities that are physically coupled together and jointly affect the final measurement signal, so that the value of each quantity can be independently and accurately obtained. In the embodiment, the spectral reflectance signal and the thermal radiation signal are processed simultaneously, and the optimal temperature (T) and real reflectivity (R) values are solved through an iterative optimization algorithm, thereby eliminating the errors caused by material differences and thin film interference effects.

[0058] The above data detection method acquires detection signals of surfaces of each wafer in the planetary reaction chamber, and the detection signals include spectral reflectance signals and thermal radiation signals. The spectral reflectance signals and the thermal radiation signals are cut based on trigger signals to obtain a first signal sequence and a second signal sequence. One trigger signal represents one rotation of a graphite disc in the planetary reaction chamber. A pre-trained classification model is used to process the first signal sequence to obtain material types on each wafer. An optical model corresponding to the material type is obtained by querying a preset database based on the material type. A decoupling algorithm is used to iteratively optimize the optical model, the first signal sequence and the second signal sequence to obtain a temperature detection value and a reflectivity detection value. According to the above scheme, the original detection signals are segmented into separate signal sequences based on the trigger signals, and then the material types are determined according to the separate signal sequences. The corresponding optical model is selected according to the material type, and finally the temperature detection value and the reflectivity detection value with high accuracy are obtained by decoupling according to the optical model. The single wafer signal is accurately analyzed, and the process repeatability and stability are greatly improved.

[0059] Please refer to Figure 3 Optionally, the spectral reflectance signals and the thermal radiation signals are cut based on the trigger signals to obtain the first signal sequence and the second signal sequence, including:

[0060] In step 302, for each trigger signal, the time interval at which each satellite disc reaches the preset position is determined with the trigger signal as the time starting point.

[0061] Step 304, a plurality of time points are set in the time interval, a time window is set around each time point, and a signal segment corresponding to each time window is cut from the spectral reflection signal and the thermal radiation signal; the width of the time window satisfies that the detection signals of all wafers on the satellite disc are completely captured.

[0062] Step 306, the signal segments belonging to the same satellite disc and obtained by continuous acquisition are spliced in time sequence to obtain a first signal sequence and a second signal sequence corresponding to each satellite disc.

[0063] Specifically, the position of each satellite disc on the graphite disc is fixed, and the angular offset of each satellite disc and the main shaft can be obtained according to the mechanical structure parameters of the disc. When the graphite disc rotates 360°, that is, an impulse signal is generated, which is the trigger signal. Taking the trigger signal as the time starting point, the time interval of each satellite disc reaching the preset position can be calculated according to the angular offset of each satellite disc and the real-time rotation speed. Exemplarily, the probe in the embodiment of the present application is arranged at the physical center position of the reaction cavity, and the above-mentioned preset position refers to the position that the center of the satellite disc reaches the position directly below the probe. It should be understood that when each satellite disc reaches the preset position with the rotation of the graphite disc, the wafer on the satellite disc will be rotated to the position below the probe in turn with the rotation of the satellite disc. By setting a time point in the time interval and setting a time window around the time point, a signal segment corresponding to each time window can be cut from the original detection signal; and then the signal segments of each satellite disc are spliced in time sequence to obtain a first signal sequence and a second signal sequence of each satellite disc. Exemplarily, the width of the time window should be greater than the diameter of the satellite disc divided by the linear velocity.

[0064] Exemplarily, the original spectral reflection signal S(t), the original thermal radiation signal X(t) and the number of satellite discs are 5, then five independent long-time first signal sequences S1(t), S2(t), S3(t), S4(t) and S5(t) and second signal sequences X1(t), X2(t), X3(t), X4(t) and X5(t) are finally cut out.

[0065] By using the above scheme, the original mixed signal is separated into independent signals of each satellite disc, which is convenient for subsequent material analysis and coupling algorithm processing.

[0066] Please refer to Figure 4 Optionally, in the case that the classification model includes a convolution network, a support vector machine network and a random forest network, a pre-trained classification model is used to process each signal sequence to obtain the material type on each wafer, including:

[0067] Step 402, extracting a feature vector representing the optical characteristics of different materials from the first signal sequence.

[0068] Step 404, inputting the feature vector into a convolution network, a support vector machine network and a random forest network respectively to obtain a first classification result, a second classification result and a third classification result.

[0069] Step 406, performing weighted summation on the first classification result, the second classification result and the third classification result, and determining the material type according to the summation result.

[0070] Specifically, the classification model in the embodiment of the application adopts a parallel model. The convolution network adopts a CNN (Convolutional Neural Network), which adopts a one-dimensional convolutional neural network structure, contains 3 convolutional layers (the convolution kernel sizes are 16, 8 and 4 respectively), each convolutional layer is followed by a ReLU activation function and a maximum pooling layer, and finally outputs the initial probability distribution P_cnn of each category through a fully connected layer. The support vector machine network adopts a SVM (Support Vector Machine), which uses a support vector machine with a radial basis function (RBF) kernel, the input is a multi-dimensional feature vector extracted from the original signal, and the output is a probability estimate P_svm. The random forest network constructs a random forest containing 100 decision trees, the input features are the same as those of the SVM branch, and the output probability estimate is P_rf.

[0071] Further, the input data of the classification model is the preprocessed optical signal time sequence of at least one wavelength, forming a multi-channel input data format, each sample contains multiple data, and the number of data is the number of rotational speeds / number of star-shaped discs* sampling frequency.

[0072] Further, the above probabilities are weighted and summed to obtain a summation result: P_final = α×P_cnn + β×P_svm + γ×P_rf, wherein α, β and γ are weight coefficients. And the above weight coefficients can be determined by grid search optimization, and exemplarily, the values in the embodiment of the application are α = 0.5, β = 0.3 and γ = 0.2.

[0073] Optionally, in the case of multiple summation results, the material type corresponding to the summation result with the maximum value is selected as the matching material of the current satellite disc.

[0074] Optionally, the application also sets a preset threshold, and outputs an alarm signal in the case that the summation result is less than the preset threshold, prompting “material cannot be identified”, and allowing software intervention for inspection, preventing mis-matching from causing all subsequent calculations to fail.

[0075] It should be noted that the classification model is trained based on historical detection data and historical process recipes.

[0076] By adopting the above scheme, the CNN is good at extracting local features, the SVM is suitable for high-dimensional space classification, and the random forest has good anti-overfitting ability. By weightedly fusing the above multiple models, the comprehensive performance of the model is optimal, and accurate identification of different material wafers is realized.

[0077] Please refer to Figure 5 Optionally, in the case that the feature vector includes a time domain feature vector and a frequency domain feature vector, the feature vector representing the optical characteristics of different materials is extracted from the first signal sequence, including:

[0078] Step 502, cutting the first signal sequence to obtain a sub-signal sequence of each wafer on a satellite disc.

[0079] Step 504, based on each sub-signal sequence, a time domain feature vector is extracted, and the time domain feature vector includes at least one of the average value, signal intensity, and quantity.

[0080] Step 506, performing Fourier transform on each sub-signal sequence to extract its main frequency component to obtain a frequency domain feature vector, and the frequency domain feature vector includes growth rate and / or film thickness.

[0081] Specifically, the present application performs secondary cutting on the spectral reflection signal, that is, after first cutting to obtain the optical signal time sequence of each satellite disc, the optical signal time sequence of each wafer on the satellite disc is further cut.

[0082] The time domain feature represents the law of the amplitude of the signal changing with time, so that the model can learn the intuitive statistical characteristics and time dynamic behavior of the signal. The frequency domain feature represents the frequency component of the signal and the energy distribution of each frequency component, so that the model can identify the periodic component and resonance frequency that is difficult to perceive in the time domain.

[0083] Through the time domain feature vector and the frequency domain feature vector, the above CNN can automatically learn the local dependency between features, discover deeper feature combinations; the SVM can ensure the optimality of the classification boundary when the feature dimension is high, prevent overfitting, and improve the stability and generalization ability of the model; the random forest can determine which time domain or frequency domain features contribute most to classification by constructing multiple decision trees and voting, providing guidance for feature optimization, and the above integrated characteristics make the final decision more robust.

[0084] Optionally, cutting the first signal sequence to obtain a sub-signal sequence of each wafer on a satellite disc includes:

[0085] detecting pulse peaks in the first signal sequence, wherein one pulse peak corresponds to one wafer; determining the start point and the end point of each pulse peak; cutting the first signal sequence based on the start point and the end point of each pulse peak to obtain a plurality of signal segments; calculating the signal intensity of each signal segment, and classifying each signal segment according to the signal intensity; and splicing the signal segments belonging to the same class in time sequence to obtain a sub-signal sequence of each wafer.

[0086] For example, three wafers are arranged on one satellite disc, and the first signal sequence corresponding to the satellite disc is S1(t). In the time window of the satellite disc, the three wafers will pass under the probe in turn due to the rotation of the satellite disc, and therefore, at least three pulse peaks will appear in the S1(t) signal, one pulse peak representing that the probe scans one wafer, and the number of pulse peaks is related to the rotation speed of the satellite disc, that is, the faster the rotation speed, the more the pulse peaks.

[0087] Due to the difference in reflectivity of different materials, the pulse peak heights of different wafers are different. For example, the pulse peak height of a sapphire material is smaller than that of a GaAs material.

[0088] The start and end points of the above three pulse peaks are found by a peak detection algorithm or by setting a dynamic threshold, so as to cut S1(t) into at least three shorter signal segments, each of which corresponds to the entire process of one wafer being scanned by the probe.

[0089] It should be understood that even on the same satellite disc, the signal intensities of different wafers can be different due to the surface state or slight difference in position. By calculating the signal intensity of each sub-signal sequence, the signal segments from the same physical wafer can be classified and spliced, thereby generating an independent sub-signal sequence for each wafer. By using the above scheme, the time continuity and statistical redundancy of the optical timing curve can be utilized to effectively improve the accuracy, stability and robustness of the subsequent coupling algorithm.

[0090] Further, the present application also pre-stores the fixed physical position sequence of each wafer on each satellite disc. After obtaining the pulse peaks, the time sequence of the pulse peaks is compared with the rotation sequence of the wafers to avoid false matching.

[0091] Optionally, in the case where the feature vector further includes a model feature vector, the feature vector representing the optical characteristics of different materials is extracted from the first signal sequence, including:

[0092] The similarity between the curve of the first signal sequence changing with time and the theoretically predicted curves of each candidate material in the preset database is calculated to obtain a plurality of initial correlation coefficients, and each initial correlation coefficient constitutes a model feature vector.

[0093] Specifically, the initial correlation coefficient represents the similarity of the two, and by converting the time series data into a similarity score with the theoretical model of each material, an input feature with clear physical meaning can be provided for the subsequent classification algorithm.

[0094] Referring to Figure 6 Optionally, the optical model is used to represent an optical constant function of an optical constant with respect to a wavelength and a temperature, the optical constant including a refractive index and an extinction coefficient; and the optical model, the first signal sequence and the second signal sequence are iteratively optimized by using a decoupling algorithm to obtain a temperature detection value and a reflectivity detection value, including:

[0095] In step 602, a pre-constructed physical forward model is used to generate predicted data under a given temperature and surface state parameter based on an optical constant function of the material. The predicted data includes theoretical reflectivity and theoretical radiation intensity. The surface state parameter includes at least one of an emissivity correction coefficient, a surface roughness parameter, and an oxidation layer influence parameter.

[0096] In step 604, the residual error between the predicted data and the measured data is calculated step by step. The measured data includes measured reflectivity and measured radiation intensity. The measured reflectivity is obtained based on the sub-signal sequence, and the measured radiation intensity is obtained based on the second signal sequence.

[0097] In step 606, a decoupled alternating optimization algorithm is used to iteratively optimize the temperature and the surface state parameter alternately, and a nonlinear least squares algorithm is used to minimize the residual error function until convergence.

[0098] In step 608, the temperature at the time of model convergence is taken as the temperature detection value, and the measured reflectivity is corrected based on the surface state parameter at the time of convergence to obtain the reflectivity detection value.

[0099] Specifically, after determining the material type, the optical model corresponding to the material type can be called from the pre-set database. The expression of the optical model is n, k = f(λ, T), where n represents the refractive index, k represents the extinction coefficient, λ represents the wavelength, and T represents the temperature. The optical model is used to represent an optical constant function of an optical constant with respect to a wavelength and a temperature, i.e., the refractive index n and the extinction coefficient k of different substrates and thin film materials at different wavelengths and temperatures.

[0100] It should be noted that the temperature is one of the core physical quantities to be solved in this scheme, which directly affects the optical constants (n, k) and the thermal radiation intensity of the material. The goal of optimizing the temperature is to find the physical true value that best meets the spectral reflectivity signal and the thermal radiation signal.

[0101] The surface state parameter is a compensation term or correction factor, which is used to describe and quantify the model parameter of the difference between the actual optical characteristics of the wafer surface and the ideal smooth and clean surface, and is used to compensate the influence of the wafer surface optical characteristics on the temperature measurement and reflectivity measurement. The purpose of optimizing the surface state parameter is to compensate the system error caused by the non-ideal surface state to the temperature and reflectivity measurement, so as to make the solution of the core physical quantity (temperature, reflectivity) more accurate.

[0102] Optionally, in order to improve the data accuracy, the above-mentioned sub-signal sequence and the second signal sequence can be obtained based on a plurality of trigger signals, that is, after each trigger signal arrives, the spectral reflectance signal is cut twice to obtain the sub-signal sequence of each wafer, and the thermal radiation signal is cut once to obtain the second signal sequence of each satellite disc. After a plurality of trigger signals, the sub-signal sequences of the wafers are spliced in time sequence to obtain a first time sequence combination; the second signal sequences of the satellite discs are spliced in time sequence to obtain a second time sequence combination; and the measured reflectivity and the measured radiation intensity are calculated based on the first time sequence combination and the second time sequence combination.

[0103] Optionally, the application also processes the sub-signal sequences of the wafers into a first time sequence curve and processes the second time sequence combination of the satellite discs into a second time sequence curve. The above-mentioned first time sequence curve and the second time sequence curve are used as input data of the decoupling algorithm. The goal of the decoupling algorithm is to find a unique temperature value that can simultaneously and optimally explain the two curves. Instead of directly distinguishing, it is through a "decoupling iterative optimization" process driven by a physical model. The system simultaneously analyzes the reflectivity curve unique to each wafer and the thermal radiation curve of each satellite disc, and through continuous hypothesis, correction and re-hypothesis, finally finds a temperature value and a surface state parameter that can simultaneously and perfectly explain the two curves.

[0104] Optionally, in order to further improve the data accuracy, the application also performs secondary segmentation on the second signal sequence to obtain a thermal radiation sub-sequence of each wafer, and then processes the thermal radiation sub-sequence of each wafer into a second time sequence curve. In this way, the data input into the coupling algorithm is the reflectivity curve unique to each wafer and the thermal radiation curve of each wafer. Exemplarily, the way of secondary segmentation of the second signal sequence includes: determining the passing time of each wafer in advance according to the wafer size and the rotating speed; performing denoising and baseline correction on the second signal sequence, calculating the envelope or integral energy, and according to the pre-set passing time constraint and intensity threshold, subdividing the second signal sequence into radiation sub-windows corresponding to each wafer to obtain the thermal radiation sub-sequence of each wafer.

[0105] The physical forward model is based on an optical constant function, and introduces a surface state parameter to compensate for the actual surface effect. Given the temperature and the surface state parameter, the theoretical reflectivity and the theoretical radiation intensity consistent with the detection link can be generated.

[0106] Further, in the alternating iterative optimization process, the reflectivity-related surface state parameters are fixed for temperature optimization, and then the temperature is optimized for surface state parameters, and such alternating iteration is used to minimize the residual function using a nonlinear least squares algorithm until convergence.

[0107] The above steps are repeated, and each wafer sub-signal sequence is processed in turn using the classification model to obtain the material type, and then the decoupling algorithm is used for iterative optimization to finally obtain the detection value and reflectivity detection value of each wafer.

[0108] The above data detection method realizes independent and accurate real-time analysis of the temperature and reflectivity of each wafer when different material wafers are placed on the planetary turntable through the signal separation algorithm and the machine learning model, and solves the problem that mixed signals cannot be distinguished in traditional technologies. Further, the application adopts a weighted fusion ensemble learning framework combined with an online learning mechanism, so that the system can adapt to process changes and the introduction of new materials, continuously optimize the recognition accuracy, and improve the long-term use accuracy. Further, through the combination of the multivariate decoupling algorithm and the optical constant database, the in-situ monitoring accuracy of mixed material wafers is improved. The above intelligent material automatic identification and parameter analysis reduces the dependence on the professional experience of the operator, improves the process repeatability and stability, and is conducive to standardized production.

[0109] Optionally, the above data detection method further comprises:

[0110] The temperature value and reflectivity value of each wafer over time are displayed, and an alarm signal is generated when the change curve exceeds the process range.

[0111] Specifically, the application also provides the user with the temperature value and reflectivity value of each wafer over time through the human-machine interface, so as to facilitate the user to quickly identify abnormalities, trigger correction or interlocking in advance, and shorten the processing delay.

[0112] Please refer to Figure 7 , the following will be combined Figure 7 to describe the data detection method of the application in detail:

[0113] The application first obtains the detection signal of the wafer surface, and the detection signal includes a spectral reflectance signal and a thermal radiation signal.

[0114] Further, the application determines, in response to the received trigger signal, a time interval for each satellite disc to reach a preset position, takes a plurality of time points within the time interval, sets a time window centered on each time point, and cuts a signal segment corresponding to each time window from the spectral reflection signal and the thermal radiation signal; the width of the time window satisfies that the detection signals of all wafers on the satellite disc are completely captured.

[0115] Further, the application detects pulse peaks in the first signal sequence; wherein one pulse peak corresponds to one wafer; determines the start point and the end point of each pulse peak; cuts the first signal sequence based on the start point and the end point of each pulse peak to obtain a plurality of signal segments; calculates the signal intensity of each signal segment, and divides the signal segments into different categories according to the signal intensity; splices the signal segments belonging to the same category in time sequence to obtain a sub-signal sequence of each wafer. Based on the sub-signal sequence, a feature vector is extracted, which includes a time domain feature vector, a frequency domain feature vector and a model feature vector. The feature vector is input into a convolution network, a support vector machine network and a random forest network respectively, and weighted summation is performed, and material matching is performed according to the summation result. If the optimal matching is obtained, the material type is determined, otherwise the classification result is marked and an alarm is given.

[0116] Further, the application calls an optical model from a preset database according to the material type, adopts a physical forward model, generates prediction data under given temperature and surface state parameters based on an optical constant function of the material, gradually calculates the residual error of the prediction data and the measured data, adopts a decoupled alternating optimization algorithm to alternately and iteratively optimize the temperature and the surface state parameter, and uses a nonlinear least squares algorithm to minimize the residual error function until convergence; the temperature at the time of model convergence is taken as the temperature detection value, and the measured reflectivity is corrected based on the surface state parameter at the time of convergence to obtain the reflectivity detection value.

[0117] Further, after the temperature detection value and the reflectivity detection value of one wafer are obtained, the above feature extraction, material classification and decoupling algorithm are repeated to finally obtain the temperature detection value and the reflectivity detection value of all wafers.

[0118] It should be understood that, although the steps in the flowcharts involved in the embodiments described above are shown in sequence according to the arrows, the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of the steps is not strictly limited in sequence, and the steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of the steps or stages is not necessarily sequential, but can be alternately or alternately executed with at least part of other steps or steps or stages in other steps.

[0119] Based on the same inventive concept, the embodiments of the present application also provide a data detection system, which is suitable for the data detection method described above, and the implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method. Therefore, the specific limitations in one or more device embodiments provided below can refer to the limitations of the method described above, and will not be repeated here.

[0120] Please refer to Figure 8 In one embodiment, the present application provides a data detection system, comprising: an acquisition module, a signal segmentation module, and a processing module.

[0121] The acquisition module is configured to acquire detection signals of wafer surfaces, wherein each wafer is arranged in a planetary reaction cavity, and the detection signals include spectral reflection signals and thermal radiation signals.

[0122] The signal segmentation module is configured to cut the spectral reflection signals and the thermal radiation signals based on trigger signals to obtain first signal sequences and second signal sequences, wherein one trigger signal represents one rotation of a graphite disc in the planetary reaction cavity.

[0123] The processing module is configured to process the first signal sequences using a pre-trained classification model to obtain material types on each wafer, query a pre-set database to obtain an optical model corresponding to the material type based on the material type, and perform iterative optimization on the optical model, the first signal sequences, and the second signal sequences using a decoupling algorithm to obtain temperature detection values and reflectivity detection values.

[0124] Optionally, the signal segmentation module cuts the spectral reflectance signal and the thermal radiation signal based on the trigger signals to obtain a first signal sequence and a second signal sequence, including: for each trigger signal, determining a time interval in which each satellite disc reaches a preset position, taking the trigger signal as a time starting point; taking a plurality of time points in the time interval, setting a time window centered on each time point, and cutting a signal segment corresponding to each time window from the spectral reflectance signal and the thermal radiation signal; the width of the time window satisfies that the detection signals of all wafers on the satellite disc are completely captured; and splicing the signal segments belonging to the same satellite disc in time sequence to obtain the first signal sequence and the second signal sequence corresponding to each satellite disc.

[0125] Optionally, in the case where the classification model includes a convolution network, a support vector machine network, and a random forest network, the processing module processes each signal sequence using a pre-trained classification model to obtain the material type on each wafer, including: extracting a feature vector representing the optical characteristics of different materials from the first signal sequence; inputting the feature vector into the convolution network, the support vector machine network, and the random forest network to obtain a first classification result, a second classification result, and a third classification result; and performing weighted summation on the first classification result, the second classification result, and the third classification result to determine the material type according to the summation result.

[0126] Optionally, in the case where the feature vector includes a time domain feature vector and a frequency domain feature vector, the processing module extracts the feature vector representing the optical characteristics of different materials from the first signal sequence, including: cutting the first signal sequence to obtain a sub-signal sequence of each wafer on a satellite disc; extracting a time domain feature vector based on each sub-signal sequence, the time domain feature vector including at least one of an average value, a signal intensity, and a quantity; performing Fourier transform on each sub-signal sequence to extract its main frequency component to obtain a frequency domain feature vector, the frequency domain feature vector including a growth rate and / or a film thickness.

[0127] Optionally, the processing module cuts the first signal sequence to obtain a sub-signal sequence of each wafer on a satellite disc, including: detecting a pulse peak in the first signal sequence; wherein one pulse peak corresponds to one wafer; determining the starting point and the ending point of each pulse peak; cutting the first signal sequence based on the starting point and the ending point of each pulse peak to obtain a plurality of signal segments; calculating the signal intensity of each signal segment, and classifying the signal segments into different classes according to the signal intensity; and splicing the signal segments belonging to the same class in time sequence to obtain the sub-signal sequence of each wafer.

[0128] Optionally, the processing module extracts the feature vector representing the optical characteristics of different materials from the first signal sequence in the case that the feature vector further comprises a model feature vector, including: calculating the similarity between the curve of the first signal sequence changing over time and the theoretically predicted curve of each candidate material in the preset database to obtain a plurality of initial correlation coefficients, and each initial correlation coefficient constitutes the model feature vector.

[0129] Optionally, the optical model is used to represent an optical constant function of the optical constant with respect to wavelength, temperature, and the optical constant includes refractive index and extinction coefficient; the processing module uses a decoupling algorithm to iteratively optimize the optical model, the first signal sequence and the second signal sequence to obtain the temperature detection value and the reflectivity detection value, including: using a pre-constructed physical forward model to generate predicted data under a given temperature and surface state parameter based on the optical constant function of the material. Wherein, the predicted data includes theoretical reflectivity and theoretical radiation intensity; the surface state parameter includes at least one of the emissivity correction coefficient, the surface roughness parameter and the oxide layer influence parameter; the residual error between the predicted data and the measured data is calculated step by step, the measured data includes the measured reflectivity and the measured radiation intensity, and the measured reflectivity is obtained based on the sub-signal sequence, and the measured radiation intensity is obtained based on the second signal sequence; using a decoupled alternating optimization algorithm, alternatingly iteratively optimizing the temperature and the surface state parameter, and using a nonlinear least squares algorithm to minimize the residual error function until convergence; the temperature at the time of model convergence is taken as the temperature detection value, and the measured reflectivity is corrected based on the surface state parameter at the time of convergence to obtain the reflectivity detection value.

[0130] Optionally, the data detection system further comprises a display module.

[0131] The display module is configured to display the temperature value and the reflectivity value of each wafer over time, and generate an alarm signal when the change curve exceeds the process range.

[0132] The data detection system realizes independent and accurate real-time analysis of the temperature and reflectivity of each wafer when the planet-type turntable is placed with different material wafers through the signal separation algorithm and the machine learning model, and solves the problem that the mixed signal cannot be distinguished in the traditional technology. Further, the application adopts a weighted fusion ensemble learning framework combined with an online learning mechanism, so that the system can adapt to process changes and the introduction of new materials, continuously optimize the recognition accuracy, and improve the long-term use accuracy. Further, through the combination of the multivariate decoupling algorithm and the optical constant database, the in-situ monitoring accuracy of mixed material wafers is improved. The above intelligent material automatic identification and parameter analysis reduces the dependence on the professional experience of the operator, improves the process repeatability and stability, and is conducive to standardized production.

[0133] The various modules in the data detection system described above can be implemented in whole or in part by software, hardware, and combinations thereof. The various modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in a computer device in software form, so as to be called and executed by a processor to perform the operations corresponding to the various modules.

[0134] In a feasible embodiment, a computer device, which can be a terminal, has an internal structure diagram as shown in Figure 9 The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to perform wired or wireless communication with external terminals. The wireless communication can be achieved through WIFI, mobile cellular network, NFC (Near Field Communication), or other technologies. The computer program is executed by the processor to implement the data detection method described above. The display unit of the computer device is configured to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball, or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0135] Those skilled in the art can understand that Figure 9 The structure shown in the above embodiment is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. Specifically, the computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0136] In a feasible embodiment, a computer device is provided, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method steps in the data detection method described above.

[0137] In a feasible embodiment, a computer readable storage medium is provided, which stores a computer program. The computer program is executed by a processor to implement the method steps in the data detection method described above.

[0138] In one possible implementation, a computer program product is provided, which includes a computer program that, when executed by a processor, implements the method steps of the above data detection method.

[0139] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0140] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present disclosure.

[0141] The above-described embodiments are merely illustrative of several embodiments of the present application, which are described in more detail and in a specific manner, but should not be construed as limiting the scope of the patent of the present application. It should be noted that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A data detection method characterized by, The method comprises: acquiring detection signals of wafer surfaces; wherein each wafer is arranged in a planetary reaction cavity, and the detection signals include spectral reflection signals and thermal radiation signals; cutting the spectral reflection signals and the thermal radiation signals based on trigger signals to obtain first signal sequences and second signal sequences; wherein one trigger signal represents one rotation of a graphite disc in the planetary reaction cavity; processing the first signal sequences by using a pre-trained classification model to obtain material types on each wafer; based on the material types, querying an optical model corresponding to the material type from a pre-set database; the optical model is used to represent an optical constant function of optical constants with respect to wavelength and temperature, and the optical constants include refractive index and extinction coefficient; iteratively optimizing the optical model, the first signal sequences and the second signal sequences by using a decoupling algorithm to obtain temperature detection values and reflectivity detection values, including: generating prediction data under given temperature and surface state parameters based on the optical constant function of the material by using a pre-constructed physical forward model; wherein the prediction data includes theoretical reflectivity and theoretical radiation intensity; and the surface state parameters include at least one of emissivity correction coefficient, surface roughness parameter and oxidation layer influence parameter; gradually calculating residuals of the prediction data and measured data; the measured data includes measured reflectivity and measured radiation intensity, and the measured reflectivity is obtained based on a sub-signal sequence, and the measured radiation intensity is obtained based on the second signal sequence; the sub-signal sequence is obtained by cutting the first signal sequence; alternately iteratively optimizing temperature and surface state parameters by using a decoupled alternating optimization algorithm, and minimizing a residual function by using a nonlinear least squares algorithm until convergence; taking the temperature at the time of model convergence as the temperature detection value, and correcting the measured reflectivity based on the surface state parameters at the time of convergence to obtain the reflectivity detection value.

2. The method of claim 1, wherein, The method comprises: for each trigger signal, determining a time interval for each satellite disc to reach a pre-set position, taking the trigger signal as a time starting point; taking a plurality of time points in the time interval, setting a time window centered on each time point, and cutting a signal segment corresponding to each time window from the spectral reflection signals and the thermal radiation signals; the width of the time window satisfies that the detection signals of all wafers on the satellite disc are completely captured; splicing the signal segments belonging to the same satellite disc in time sequence to obtain the first signal sequences and the second signal sequences of each satellite disc.

3. The method of claim 1, wherein, The classification model includes a convolution network, a support vector machine network and a random forest network. The method comprises: extracting feature vectors representing optical characteristics of different materials from the first signal sequences; inputting the feature vectors into the convolution network, the support vector machine network and the random forest network respectively to obtain first classification results, second classification results and third classification results; performing weighted summation on the first classification results, the second classification results and the third classification results, and determining the material type according to the summation result.

4. The method of claim 3, wherein, The feature vectors include time domain feature vectors and frequency domain feature vectors. The feature vectors representing the optical characteristics of different materials are extracted from the first signal sequence, including: cutting the first signal sequence to obtain a sub-signal sequence of each wafer on the satellite disc; based on each sub-signal sequence, extracting a time domain feature vector, the time domain feature vector including at least one of the average value, signal intensity, and quantity; performing Fourier transform on each sub-signal sequence to extract its main frequency component to obtain a frequency domain feature vector, the frequency domain feature vector including growth rate and / or film thickness.

5. The method of claim 4, wherein, The cutting of the first signal sequence to obtain a sub-signal sequence of each wafer on the satellite disc includes: detecting pulse peaks in the first signal sequence; wherein one pulse peak corresponds to one wafer; determining the start point and end point of each pulse peak; cutting the first signal sequence based on the start point and end point of each pulse peak to obtain a plurality of signal segments; calculating the signal intensity of each signal segment, and classifying each signal segment according to the signal intensity; splicing the signal segments belonging to the same class in time sequence to obtain a sub-signal sequence of each wafer.

6. The method of claim 4, wherein, The feature vectors also include model feature vectors; The feature vectors representing the optical characteristics of different materials are extracted from the first signal sequence, further including: calculating the similarity between the curve of the first signal sequence changing with time and the theoretical prediction curve of each candidate material in the preset database to obtain a plurality of initial correlation coefficients, each initial correlation coefficient constituting the model feature vector.

7. The method of claim 1, wherein, The method further includes: displaying the temperature value and reflectivity value of each wafer changing with time, and generating an alarm signal when the change curve exceeds the process range.

8. A data detection system characterized by, The system includes: an acquisition module for acquiring detection signals of wafer surfaces; wherein each wafer is arranged in a planetary reaction cavity, and the detection signals include spectral reflection signals and thermal radiation signals; a signal segmentation module for cutting the spectral reflection signals and the thermal radiation signals based on trigger signals to obtain first signal sequences and second signal sequences; wherein one trigger signal represents one rotation of the graphite disc in the planetary reaction cavity. The processing module is configured to process each of the first signal sequences by using a pre-trained classification model to obtain a material type on each wafer, and query a preset database based on the material type to obtain an optical model corresponding to the material type, wherein the optical model is used to represent an optical constant function of an optical constant with respect to a wavelength and a temperature, and the optical constant includes a refractive index and an extinction coefficient; and iteratively optimize the optical model, the first signal sequence, and the second signal sequence by using a decoupling algorithm to obtain a temperature detection value and a reflectivity detection value, including: generating predicted data under a given temperature and a surface state parameter based on an optical constant function of a material by using a pre-constructed physical forward model, wherein the predicted data includes a theoretical reflectivity and a theoretical radiation intensity, and the surface state parameter includes at least one of a radiation correction coefficient, a surface roughness parameter, and an oxidation layer influence parameter; The residual error of the predicted data and measured data is calculated step by step, wherein the measured data includes a measured reflectivity and a measured radiation intensity, the measured reflectivity is obtained based on a sub-signal sequence, and the measured radiation intensity is obtained based on the second signal sequence, and the sub-signal sequence is obtained by cutting the first signal sequence; The temperature and the surface state parameter are alternately iteratively optimized by using a decoupled alternating optimization algorithm, and a nonlinear least squares algorithm is used to minimize a residual function until convergence is achieved; The temperature at the time of model convergence is taken as the temperature detection value, and the measured reflectivity is corrected based on the surface state parameter at the time of convergence to obtain the reflectivity detection value.

9. A detection device, characterized in that The optical measurement module is configured to emit a light beam to a surface of each wafer in a planetary reaction chamber and collect a spectral reflectance signal of the surface of each wafer; The temperature measurement module is configured to detect a thermal radiation signal of the surface of each wafer; The signal line is connected with a main shaft encoder or a motor controller of a planetary turntable machine table, and is used to transmit a trigger signal, wherein one trigger signal represents one rotation of a graphite disc in the planetary reaction chamber. The data detection system as claimed in claim 8 is configured to periodically process the spectral reflectance signal and the thermal radiation signal based on the trigger signal to obtain the temperature detection value and the reflectivity detection value. The processor executes the computer program to implement the steps of the method of any one of claims 1-7. 10.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-9. The computer program is executed by the processor to implement the steps of the method of any one of claims 1-7.

11. A computer readable storage medium having stored thereon a computer program, characterized in that, ​

Images