Polishing liquid monitoring method and monitoring system

Abstract

The application discloses a polishing liquid monitoring method and a monitoring system, and belongs to the technical field of polishing liquid. The polishing liquid monitoring method comprises the following steps: step 1: obtaining component data of the polishing liquid along with a use process and mark data corresponding to the component data, wherein the component data comprises the concentration of each component, and the mark data comprises PH value, spectral information and temperature data; and step 2: obtaining time temperature data, time PH data and time spectral information of the polishing liquid to be measured as prediction information. In the technical scheme, when the effective components in the polishing liquid are monitored, the polishing liquid is not sampled, but the prediction information of the polishing liquid in the use process is collected, and then a prediction model is used to predict the component data, so that the monitoring efficiency is higher, and the component change in the polishing liquid can be monitored more quickly.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of polishing liquid, in particular to a polishing liquid monitoring method. BACKGROUND

[0002] Polishing liquid is widely used in various manufacturing and processing fields, such as semiconductor manufacturing, automobile manufacturing, CNC machining, etc. The composition in the polishing liquid will change as the number of uses increases, thereby reducing the effect of the polishing liquid. Therefore, the polishing liquid needs to be replaced in time according to the actual production requirements of the polishing liquid. Before replacing the polishing liquid, the composition in the polishing liquid needs to be accurately detected.

[0003] The traditional polishing liquid detection method generally analyzes the composition of the polishing liquid through chemical quantitative experiments to obtain the composition content of the polishing liquid. This polishing liquid composition analysis method needs to filter, separate, extract, and quantify the collected polishing liquid sample to obtain the content of each component in the polishing liquid. Therefore, the composition detection of the polishing liquid requires a large amount of time cost.

[0004] In practice, it has been a difficult problem in the industry to monitor the composition of the polishing liquid in the use process of the polishing liquid. For the replacement of the polishing liquid, the commonly used technical means is to detect the composition of the polishing liquid once the number of uses of the polishing liquid reaches the expected number or reaches the expected time, if the composition of the polishing liquid meets the requirements, continue to use, if not, stop using, and replace the polishing liquid. This method cannot continuously monitor the composition change of the polishing liquid in the use process of the polishing liquid. Therefore, it is easy to continue to use the polishing liquid that cannot meet the processing requirements for polishing operation when the composition of the polishing liquid does not meet the polishing requirements, thereby resulting in poor polishing effect and causing defective products. SUMMARY

[0005] The summary part of the present application is used to introduce the concept in a brief form, which will be described in detail in the specific embodiment part. The summary part of the present application is not intended to identify the key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0006] As a first aspect of the present application, in order to solve the technical problems mentioned in the background part, some embodiments of the present application provide a polishing liquid monitoring method, comprising the following steps:

[0007] Step 1: obtaining composition data of the polishing liquid along with the use process and mark data corresponding to the composition data, the composition data including the concentration of each component, and the mark data including PH value, spectral information and temperature data;

[0008] Step 2: Obtain the time-temperature data, time-PH data and time-spectrum information of the polishing liquid to be tested as prediction information;

[0009] Step 3: Construct a component prediction model, take the prediction information as the input of the prediction model, and take the component data as the output, and cross-train the prediction model;

[0010] The prediction model includes an LSTM neural network, a convolutional neural network, a fully connected layer and a cycle optimization layer.

[0011] The time-temperature data and the time-PH data are input into the LSTM neural network to obtain first high-level features;

[0012] The spectrum information is input into the convolutional neural network to obtain second high-level features;

[0013] The first high-level features and the second high-level features are input into the fully connected layer to generate fusion features;

[0014] The fusion features are input into the cycle optimization layer to generate component data.

[0015] Step 4: Use the cross-trained prediction model to predict the components of the polishing liquid.

[0016] In the technical solution provided in the present application, when monitoring the effective components of the polishing liquid, the method of "filtering, separating, extracting and quantifying" the samples of the polishing liquid to extract various components in the polishing liquid and then calculating the content of various components in the polishing liquid is not used. Instead, the labeled data is used as prediction information, the component data is used as a label, a large amount of prediction information and labels are used to train the prediction model, and the internal relationship between the prediction information and the label is found by using the prediction model. Therefore, when monitoring the components of the polishing liquid, the labeled data which is easy to obtain can be used to predict the component data of the polishing liquid, and then the changes of various components in the polishing liquid can be obtained along with the use process of the polishing liquid, so that the polishing liquid can be replaced in time to avoid the occurrence of defective products during polishing. Meanwhile, the prediction model constructed in the present application includes an LSTM neural network and a convolutional neural network. These two neural network models are respectively used to input sequence data related to time and image data. The LSTM neural network can capture the information before and after the time-temperature data and the time-PH data to obtain the change characteristics of the components dissolved in the polishing liquid, and the convolutional neural network can capture the corresponding image characteristics to obtain the change characteristics of the suspended matter in the polishing liquid which affects the scattering of light.

[0017] Further, the polishing liquid includes aluminum oxide micro powder, CTAB, hydrochloric acid, sodium hydroxide and deionized water.

[0018] The types of substances in the polishing liquid are very diverse, and the content of each substance has a certain disorder. If the content of each substance is considered in the prediction model, the precision requirement of the prediction model will be high, and more feature relationships need to be considered, resulting in high cost of model construction. Therefore, the technical scheme is provided as follows:

[0019] Further, step 1 comprises the following steps:

[0020] Step 11: Pre-acquire the same kind of polishing liquid, the content of each component in the use process and the corresponding PH value and temperature data;

[0021] Step 12: According to the use requirements of the polishing liquid, set the sampling range and prediction accuracy of each component, and generate each component mark point according to the sampling range and prediction accuracy of each component.

[0022] In the technical scheme provided by the present application, the sampling range and prediction accuracy of each component are set, so the invalid data of the concentration of each component in the polishing liquid is removed, so that the results considered by the prediction model will be less, reducing the accuracy requirement of the prediction model.

[0023] The spectral information is affected by other factors during measurement. Therefore, the present application provides the following scheme:

[0024] Further, the spectral information is acquired by performing the following steps:

[0025] S1: Obtain a polishing liquid sample, irradiate the polishing liquid sample with a fixed wavelength laser, and collect a spectrum of the laser transmitted through the polishing liquid sample;

[0026] S2: According to the light intensity signal distribution in the spectrum, the spectrum is intercepted to obtain a region of interest;

[0027] S3: Median filtering is performed on the region of interest to obtain a filtered spectrum;

[0028] S4: The filtered spectrum is binarized, and then the center of the filtered spectrum is extracted;

[0029] S5: Extract the maximum light spot and the saturated light spot of the filtered spectrum, label the center, the maximum light spot and the saturated light spot in the region of interest, and take the region of interest as the spectral information.

[0030] In the technical scheme provided by the present application, when collecting spectral information, the spectrum is first intercepted to remove a lot of irrelevant information, and then median filtering is performed, which can reduce noise interference and increase the effectiveness of information. Labeling the maximum light spot, the saturated light spot and the center in the region of interest can reflect the scattering of the polishing liquid sample to the spectrum, so as to reflect the dispersion of the suspended substances in the polishing liquid, and further reflect the concentration of the suspended substances.

[0031] The spectral information is affected by the optical path of the laser in the polishing liquid, so there are some dark spots, which are mainly caused by the laser being blocked by the suspended matter in the polishing liquid during scattering. The distribution of these dark spots has no regularity and cannot reflect the concentration of the suspended matter in the polishing liquid. However, they will affect the reading of the spectral information. Therefore, the application provides the following technical solutions:

[0032] Further, the window of the median filtering is 3x3.

[0033] In the technical solution provided by the application, the 3x3 window median filtering is set, so for each pixel point, the median of the light intensity of the surrounding 8 points plus itself is obtained, thereby the special dark spots in the spectral information can be accurately filtered, and the distribution of the low light intensity in the spectral information is not blurred, and the reading ability of the spectral information is increased.

[0034] The spectral information of the polishing liquid is affected by many factors, especially the spectral information of the edge part of the spectral information. Because the laser is reflected and absorbed many times, there is a certain disorder. Therefore, it is difficult to distinguish the scattering characteristics of the laser. Therefore, the application provides the following technical solutions:

[0035] Further, when extracting the center of the filtered spectrum, the threshold value of the binary processing is the median of the intensity range.

[0036] In the technical solution provided by the application, in order to accurately reflect the influence of the concentration of the suspended matter in the polishing liquid on the laser scattering, the median of the intensity range is used for binary processing and positioning the center position, so the center position can reflect the range distribution of the laser scattering in the polishing liquid, and the relationship between the concentration of the suspended matter in the polishing liquid and the laser scattering can be extracted.

[0037] Further, when extracting the maximum light spot, the threshold value of the binary processing is the maximum value of the intensity range of the background area.

[0038] In the present solution, the "maximum value of the intensity range of the background area" is used as the maximum light spot, so the scattering range of the laser in the polishing liquid can be accurately reflected.

[0039] Further, when extracting the saturated light spot, the threshold value of the binary processing is three quarters of the maximum value of the intensity range.

[0040] In the present solution, the "three quarters of the maximum value of the intensity range" is used to extract the saturated light spot, so the saturated light spot is the center area of the laser after being scattered in the polishing liquid, which is convenient for the prediction model to learn and locate the center position of the spectral information.

[0041] The center of the filtered spectrum, the maximum spot, and the saturated spot can accurately reflect the scattering effect of the polishing liquid sample on the spectrum, distinguish effective information, and the light transmission intensity.

[0042] Further, the convolutional network model comprises:

[0043] The input layer has 3 channels, respectively for inputting the center mask of the filtered spectrum, the maximum spot mask, and the saturated spot mask of the filtered spectrum in the region of interest;

[0044] The convolutional layer is used for convolution operation on the information input by the input layer;

[0045] The pooling layer performs pooling operation on the output of the convolutional layer to output a feature map;

[0046] The fully connected layer classifies the feature map to generate a second high-level feature.

[0047] In the technical solution provided by the present application, when generating the second high-level feature, an input layer with three channels is selected, and the input information is the center mask of the filtered spectrum, the maximum spot mask, and the saturated spot mask of the filtered spectrum in the region of interest. Therefore, when generating the second high-level feature, the convolutional network model can comprehensively consider the related features of the center, the maximum spot, and the saturated spot.

[0048] As a second aspect of the present application, the present application provides a polishing liquid monitoring system, comprising:

[0049] The information acquisition module is used for acquiring the mark information; and the monitoring module is used for monitoring the concentrations of the components of the polishing liquid based on the polishing liquid monitoring method. BRIEF DESCRIPTION OF DRAWINGS

[0050] The accompanying drawings, which form a part of the present application, are used to provide further understanding of the present application, so that other features, purposes and advantages of the present application become more apparent. The illustrative embodiment drawings of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application.

[0051] In addition, throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic, and the elements and elements are not necessarily drawn to scale.

[0052] In the drawings:

[0053] Figure 1 The flowchart of the polishing liquid monitoring method.

[0054] Figure 2 The effect diagram of the laser transmitting through the polishing liquid sample. DETAILED DESCRIPTION

[0055] Embodiments of the present application will be described in more detail with reference to the drawings. While the present application is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that the drawings are not intended to limit the application to the particular embodiments disclosed, but on the contrary, the intention is to cover all modifications, equivalents and alternatives falling within the spirit and scope of the application as defined by the appended claims.

[0056] It should also be understood that, unless clearly indicated otherwise, the drawings are not necessarily drawn to scale. In addition, it is to be understood that the phraseology or terminology herein is for the purpose of description and not of limitation, as the scope of the application can be further defined only by the appended claims. Furthermore, it is to be understood that the use of "including," "comprising," "having," "containing," or "including" and variations thereof herein are not intended to be construed as having a limiting meaning, and that the use of such terms is not intended to exclude other additives, components, elements or features. In addition, it should be understood that where the application is defined by a plurality of separate components or elements, it is intended that these components or elements can be combined together into a single unitary component or element, and that statements that a component or element is separate are intended to be construed as a statement that the component or element is separate and distinct from another component or element.

[0057] The application will be described in more detail with reference to the drawings and embodiments.

[0058] Referring to Figure 1 The polishing liquid monitoring method comprises the following steps:

[0059] Step 1: Obtain the component data of the polishing liquid along with the use process and the mark data corresponding to the component data, wherein the component data comprises the concentration of each component, and the mark data comprises the PH value, spectral information and temperature data.

[0060] In metal processing, the polishing liquid can make the metal product exceed the original luster, and improve the surface quality of the product. In the configuration of the polishing liquid, the concentration of each component is determined. With the use of the polishing liquid, each component in the polishing liquid will gradually change under various chemical and physical reactions until it loses its effect.

[0061] The polishing liquid generally comprises suspended matter and soluble matter. For example, a polishing liquid mainly composed of aluminum oxide micro powder, CTAB, hydrochloric acid, sodium hydroxide and deionized water, wherein the aluminum oxide micro powder is suspended matter, and the CTAB, hydrochloric acid and sodium hydroxide are soluble matter. Although there is no inevitable chemical or physical reaction relationship between the suspended matter and the soluble matter, the concentration of the suspended matter and the concentration of the soluble matter change simultaneously with the polishing process during the use of the polishing liquid.

[0062] In order to avoid the appearance of defective products during processing, it is necessary to monitor the concentration of each component of the polishing liquid and replace the polishing liquid in time. During the use of the polishing liquid, the concentration of each component in the polishing liquid will change, and the corresponding temperature and PH value of the polishing liquid will also change. At the same time, after the concentration of each component in the polishing liquid changes, the spectral information will also change. The PH value and temperature are affected by the soluble matter in the polishing liquid, while the spectral information is mainly affected by the suspended matter in the polishing liquid, that is, the PH value and temperature reflect the concentration information of the soluble matter in the polishing liquid, and the spectral information affects the information of the insoluble matter in the polishing liquid. The reason is that the soluble matter will affect the PH value of the polishing liquid, and the chemical reaction generated during the use of the polishing liquid will also affect the temperature of the polishing liquid, resulting in a change in the temperature of the polishing liquid. The insoluble matter in the polishing liquid is dispersed in the polishing liquid in a suspended manner, and after being irradiated by laser, the laser and the suspended matter generate scattering, absorption and other information, and then generate spectral information.

[0063] In this way, the PH value and temperature can reflect the concentration information of the soluble matter in the polishing liquid, and the spectral information can reflect the concentration information of the suspended matter in the polishing liquid.

[0064] Further, the marking data in the present scheme includes PH value, spectral information and temperature data, which utilizes the inherent relationship between these information and the main components in the polishing liquid.

[0065] Because the temperature is affected by the environmental temperature and heat dissipation, the polishing liquid needs to be monitored during continuous use in the present scheme, so as to avoid the fact that the polishing liquid cannot reflect the heat absorption and heat release information during the use of the polishing liquid when the polishing liquid is used intermittently.

[0066] Further, step 1 includes the following steps:

[0067] Step 11: Pre-acquire the same kind of polishing liquid, the content of each component during use and the corresponding PH value and temperature data.

[0068] There are various kinds of polishing liquids, and each polishing liquid has different components and reactions, so each polishing liquid needs to be trained separately during polishing liquid prediction.

[0069] After the polishing liquid is configured, the effective components in the polishing liquid will change with the reaction, and for each stage, the component content, PH value, temperature data and spectral information can be extracted. Among them, the component content needs to be sampled from the polishing liquid, and then the corresponding component test in the laboratory can be performed. The PH value and temperature data are directly monitored by sensors. In this way, the change information of the component content under normal use of the polishing liquid and the PH value and temperature corresponding to the change information of the component content can be obtained.

[0070] For example, the initial component information of the polishing liquid is a1; the PH value at this time is a2, the temperature is a3, and the spectral information is a4. After the polishing liquid is used for 1 hour, the component information changes from a1 to b1, the PH value changes to b2, the temperature changes to b3, and the spectral information changes to b4. The PH value, temperature, spectral information, and component information are continuously recorded as the use process of the polishing liquid continues. From the completion of the configuration to the inability to use the polishing liquid, the component content and the PH value and temperature data corresponding to the component content can be obtained, that is, a plurality of data groups can be obtained, each data group including the PH value, temperature, spectral information, and component information.

[0071] Step 11 is to obtain the change of the component content, PH value, temperature, and spectral information of a large amount of polishing liquid during use. The component content is obtained through standard chemical quantitative test detection, and the PH value, temperature, and spectral information are directly obtained through sensors or corresponding devices. The specific acquisition methods belong to the prior art and will not be further described here.

[0072] The core idea of the present scheme is that for each polishing liquid, a large number of data groups are obtained in advance through step 11. These data groups are actually training samples of the prediction model. The component content is the information that needs to be predicted by the prediction model, and the PH value, temperature, and spectral information are the input information of the prediction model.

[0073] Step 12: According to the use requirements of the polishing liquid, set the sampling range and prediction accuracy of each component, and generate each component marker point according to the sampling range and prediction accuracy of each component.

[0074] During the use of the polishing liquid, the more frequently the component content is detected, the more dense the corresponding data is, the more training samples there are, and at the same time, the relationship between the component content and the marker data can be better exhibited. In subsequent training, the accuracy of the prediction model is higher. For example, assuming that a certain polishing liquid has a time life of 1 hour under specific working conditions after being configured, if it is detected only three times within 1 hour, only 3 data groups can be obtained; if it is detected 10 times, 10 detection groups can be obtained. The more frequently the detection is, the more tedious the component content detection work is, but the more the number of detections is, the more data groups there are, and the better the training effect of the prediction model is.

[0075] Therefore, the extraction density when collecting data in step 11 needs to be controlled to control the difficulty of obtaining training samples. The reason for such design is as follows:

[0076] Suppose the total number of training samples is set to 1000, if the 1000 training samples are extracted in the use cycle of a polishing liquid. Then this polishing liquid can extract 1000 data groups from the beginning of use to the final use, which can well reflect the relationship between the component information and the label data of the polishing liquid in the use process. However, the 1000 data groups are only the detection results of the use cycle of one polishing liquid, and cannot well represent all the polishing liquids. Therefore, if the 1000 training samples are extracted in the use cycle of 10 polishing liquids, only 100 data groups are involved in each use cycle of the 10 polishing liquids. Compared with the 1000 data groups in the use cycle of one polishing liquid before, the total number of data groups is reduced, but the 1000 training samples involve the use cycle of 10 polishing liquids, which is more representative. Therefore, the extraction density of data needs to be controlled, and more use cycles of polishing liquids are involved under the same data extraction difficulty. Therefore, in practice, it needs to be set according to the use requirements of the polishing liquid.

[0077] For example, the polishing liquid with effective component "alumina micropowder" is selected, and the concentration of the substance must be higher than 10%. When configuring, the initial concentration of the substance is 20%. After polishing a workpiece, the maximum consumption of alumina micropowder is 1%. Therefore, in order to avoid the content of alumina micropowder being lower than 10% when the workpiece is polished by the polishing liquid, the difference between the contents of alumina micropowder between adjacent sampling points needs to be less than 1% when collecting data groups in step 11. In this way, the prediction model can try to achieve an error of less than 1% in prediction accuracy, so that the concentration of alumina micropowder in the polishing liquid can be monitored when the concentration of alumina micropowder is 11% to 12%, and the concentration of alumina micropowder is avoided to be insufficient during polishing. Therefore, the extraction density when collecting data in step 11 is related to the error accuracy, if the error needs to be less than 1%, the extraction density also needs to be less than 1%.

[0078] Thus, in step 11, when the relevant data groups are extracted, it is necessary to ensure that the obtained component content and the corresponding PH value, temperature data and spectral information density are lower than the concentration change of 1% of the "alumina micro powder". For example, during the use of the polishing liquid, the average consumption of alumina micro powder is 0.5% per hour, so when collecting information, the density of the component content and the corresponding PH value, temperature data and spectral information needs to be lower than the concentration change of 1% of the "alumina micro powder". In the obtained component content and the corresponding PH value, temperature data and spectral information, the content (concentration) change value of the alumina micro powder in the component content is about 0.5% (less than 1%). At this time, the component content can be extracted as needed, so that the content (concentration) change value of the alumina micro powder in the adjacent component content is about 0.9%.

[0079] In essence, step 11 is a process of obtaining information with greater density, and step 12 is a process of extracting information with greater density. Therefore, the marker points in step 12 are the pre-set data points. For example, the data points of 11%, 12%, …20% of the alumina micro powder can be set as marker points. In practice, if the component content data extracted from a batch of polishing liquid only contains data points of 11.1%, 12.1%, …20.1%, and does not contain data points of 11%, 12%, …20%, then the data points of 11.1%, 12.1%, …20.1% are used to replace the pre-set data points of 11%, 12%, …20%. If the number of data groups obtained from a batch of polishing liquid is higher than the number of data points, then the redundant data points are deleted so that the number of data points is the same as the pre-set number.

[0080] Step 2: Obtain the time temperature data, time PH data and time spectral information of the polishing liquid to be tested as prediction information. Step 2 mainly uses the prediction model to collect the required information after the prediction model is trained. The prediction information here is actually the input information of the prediction model in step 1.

[0081] In steps 1 and 2, spectral information needs to be collected, and the spectral information can be obtained by the following steps:

[0082] S1: Obtain a polishing liquid sample, irradiate the polishing liquid sample with a fixed wavelength laser, and collect the spectral diagram of the laser passing through the polishing liquid sample.

[0083] Since the influence of natural light needs to be avoided when obtaining spectral information, a sample of the polishing liquid needs to be obtained. In practice, the polishing liquid can be obtained manually or by a machine from the container (water tank) of the polishing liquid, or the sample of the polishing liquid can be automatically obtained by a pipeline system.

[0084] After obtaining a sample of the polishing fluid, the sample is placed in a test tube and illuminated with a laser. Spectral information is collected after the laser passes through the test tube. The principle is as follows: when the laser is irradiated on pure water, the laser will basically pass through the pure water, forming a light spot slightly larger than the initial diameter of the laser. When the laser is irradiated on the polishing fluid, scattering and absorption phenomena will occur. These phenomena are closely related to the concentration of the corresponding substances in the polishing fluid. When the laser passes through the polishing fluid, it will interact with the particles or molecules in the polishing fluid, causing the propagation direction and intensity of the light to change. As a result, the laser will be diffused, and ultimately form a light spot much larger than the initial diameter of the laser. This light spot is the spectral information. Therefore, the spectral information is more closely related to the content of each component in the polishing fluid.

[0085] In practice, because the soluble particles in the polishing liquid are small and evenly dispersed, they have little effect on the scattering of the laser. However, the suspended matter in the polishing liquid has a large particle size and is more likely to scatter and reflect the laser noise, affecting the spectral information. In order to increase the correlation between the spectral information and the concentration of the suspended matter, in this scheme, the frequency of the laser is f, and the absorption peak of the suspended matter is v, f = c*v, where c is the speed of light. For example, the absorption peak of the Al-O bond vibration peak in aluminum oxide is approximately 400-600 cm -1 Within the range, the laser frequency f = 400cm -1 ×3×10 10 cm / s=1.2×10 13 Hz.

[0086] In this scheme, the frequency of the laser is made equal to the absorption peak of the suspended matter. Therefore, the light intensity in the final spectrum has a greater correlation with the concentration of the corresponding suspended matter. When the concentration of the suspended matter changes, the corresponding changes can be obtained more clearly in the spectrum. At the same time, the concentration changes of other substances will reduce the impact on the light intensity in the spectrum.

[0087] In S1, the thickness (optical path) of the polishing liquid sample also needs to be controlled, that is, the polishing liquid needs to be avoided from being too thick, and then after the laser passes through the polishing liquid, most of the laser energy is scattered and consumed, and no effective spectral information can be detected. Therefore, for each polishing liquid, the thickness of the polishing liquid sample needs to be controlled in advance. It is mainly judged based on the detected intensity (light intensity). In this solution, the light intensity is controlled between 0.1 and 0.2 of the initial light intensity to allow the laser to be fully dispersed in the polishing liquid and avoid other problems. Figure 2 As shown, Figure 2 This is a diagram showing the scattering effect of infrared light irradiating the polishing liquid sample.

[0088] S2: According to the light intensity signal distribution in the spectrum, the spectrum is intercepted to obtain the region of interest.

[0089] The pixels at the edge of the spectrum are irrelevant to spectral scattering, so the spectrum needs to be intercepted to obtain the region of interest. When obtaining the region of interest, the rate of change of intensity (light intensity) is mainly used to determine the region of interest. The region with an intensity change rate below a first preset value is set as the region of interest.

[0090] For example, in a spectrum graph, the intensity increases from 100 to 1000 from the center to the edge. The pixels on the edge with an intensity less than 100 have no usable information because the intensity has not changed. Therefore, the area with increased intensity is regarded as the region of interest.

[0091] S3: Perform median filtering on the region of interest to obtain a filtered spectrum.

[0092] The region of interest is the middle position of the spectral information, but since there is still a lot of noise in the spectral information, a median filter is required. The window of the median filter is 3×3. During the median filter, for each pixel, the original pixel value is replaced by the median of the pixel values ​​in its neighborhood (3×3 range). In this solution, the reason for using a 3×3 window is mainly to avoid the influence of some special dark spots on the spectral information. The 3×3 window is actually the pixels directly adjacent to the pixel point. The intensity of the center point is replaced by the intensity of these pixels, which minimizes the impact on the spectral information and can effectively eliminate special dark spot information.

[0093] S4: Binarize the filtered spectrum and then extract the center of the filtered spectrum.

[0094] S5: Extract the maximum spot and saturation spot of the filtered spectrum, mark the center of the circle, the maximum spot and the saturation spot in the region of interest, and use the region of interest as spectral information.

[0095] S4 and S5 are mainly used to obtain effective information from the filtered spectrum. The effective information here mainly refers to the center of the filtered spectrum, the maximum spot, and the saturated spot. When extracting the above three pieces of information, the main method is to use a binarization threshold. After the binarization threshold is set, information extraction is performed.

[0096] In the acquisition of the center of the filtered spectrum, the threshold of the binary processing is the median of the intensity range. For example, in the filtered spectrum, the intensity range of all pixel grids is 1000-2000, and the median of the intensity range is 1500. The pixel grids with intensity values (light intensity) higher than 1500 in the filtered spectrum are set to 1, and the pixel grids with intensity values (light intensity) lower than 1500 are set to 0. For the pixel points set to 1, the center point is found, and the center point is taken as the center of the filtered spectrum. In the acquisition of the maximum light spot and the saturated light spot, the scheme is the same, except that the critical point of the binary processing is different. Specifically, in the acquisition of the maximum light spot, the threshold of the binary processing is the maximum value of the intensity of the background region.

[0097] The background region refers to the edge region in the filtered spectrum. In the filtered spectrum, the intensity of the center region is very high, and the intensity changes obviously. In the edge part of the filtered spectrum, the intensity is low, and the intensity changes slowly. The region of interest is the part in the filtered spectrum with an intensity change lower than a second preset value. For example, if the intensity change of adjacent 10 pixel grids is less than 100, the demarcation line of the background region is taken as the second preset value.

[0098] After obtaining the demarcation line of the background region, the maximum value of the intensity of the background region can be obtained, and then the pixel grids higher than the value are set to 1, and the pixel grids lower than the value are set to 0. For the pixel points set to 1, the pixel points are taken as the maximum light spot.

[0099] In the extraction of the saturated light spot, the threshold of the binary processing is the three-fourth of the maximum intensity in the filtered spectrum. For example, in the filtered spectrum, the intensity range of all pixel grids is 1000-2000, and the pixel grids higher than 1750 are set to 1, and the pixel grids lower than 1750 are set to 0. For the pixel points set to 1, the pixel points are taken as the saturated light spot.

[0100] Therefore, in the scheme, the maximum light spot is the region in the region of interest most affected by the scattering of the polishing liquid, and the saturated light spot is the core region of the laser scattering.

[0101] Step 3: Construct a component prediction model, take the prediction information as the input of the prediction model, and take the component data as the output, and cross-train the prediction model.

[0102] The prediction model includes an LSTM neural network, a convolutional neural network, a fully connected layer, and a cyclic optimization layer.

[0103] The time-temperature data and the time-PH data are input into the LSTM neural network to obtain first high-level features; the spectral information is input into the convolutional neural network to obtain second high-level features; the first high-level features and the second high-level features are input into the fully connected layer to generate fusion features; and the fusion features are input into the cyclic optimization layer to generate component data.

[0104] The LSTM neural network comprises:

[0105] An input network for inputting PH value and temperature data.

[0106] An LSTM layer comprising one or more LSTM units. Each LSTM unit has three main gates: a forget gate, an input gate, and an output gate, and a cell state for storing long-term information.

[0107] A fully connected layer for outputting first high-level features.

[0108] The convolutional network model comprises:

[0109] An input layer having three channels for inputting a center mask of a region of interest and a filtered spectrum, a maximum spot mask, and a saturated spot mask of the region of interest and the filtered spectrum.

[0110] A convolutional layer for performing convolutional operations on the information input by the input layer.

[0111] A pooling layer for performing pooling operations on the output of the convolutional layer to output a feature map.

[0112] A fully connected layer for classifying the feature map to generate second high-level features.

[0113] The LSTM neural network and the convolutional network model are mainly used to extract high-level features related to the content of each component. Therefore, when building a prediction model, the convolutional network model and the LSTM neural network need to be trained first. The training method is to input the PH value and temperature data as input information of the LSTM neural network, and input the spectral information as input information of the convolutional network model, and then use the content of each component as the supervision information of the LSTM neural network and the convolutional network model. The LSTM neural network outputs first high-level features, and the convolutional network model outputs second high-level features. In this way, the first high-level features and the second high-level features have a greater correlation with the content of each component.

[0114] The cycle optimization layer is actually a support vector machine model. In practice, the concentrations (contents) of substances in the polishing liquid are not disordered, but have certain rules. For example, if the concentration of aluminum oxide powder is measured, it can actually represent the current reaction state of the polishing liquid, and the concentrations of CTAB, hydrochloric acid, and sodium hydroxide are also basically the same.

[0115] Therefore, the cycle optimization layer used in the present scheme is to use a support vector machine to regress the fused features into the corresponding component information. That is, in the cycle optimization layer, the label is the component information, and the predicted information (input information) is the labeled data. Among them, the type of component information is limited, and the type of component information is mainly related to the prediction accuracy set in step 12. For example, if the polishing liquid with effective component "alumina micro powder" is selected, the concentration of the substance must be higher than 10%, the initial concentration of the substance is 20% when configured, and the error is less than 1%, then the sample type is 11%, 12%, …20%. As for the remaining components, in fact, after the content of alumina micro powder is determined, the content of the remaining components is basically determined. Therefore, the type of component information in the present scheme is limited. In this way, the cycle optimization layer in the present scheme is actually based on a support vector machine to distinguish the fused features and regress the fused features into the corresponding component information.

[0116] Specifically, the training process of the cycle optimization layer is as follows:

[0117] S01: initialize the weight vector w, the bias term b, the bias vector e, and the regularization parameter λ;

[0118] S02: set the objective function:

[0119] T represents the transpose of the vector;

[0120] Set the constraint condition:

[0121]

[0122] y i represents the true output of the i-th sample, represents a non-linear function that maps the input vector X i to a high-dimensional space, e i represents the prediction error of the i-th sample;

[0123] Set the Lagrange function:

[0124]

[0125] i and j represent the index of the sample, a i represents the Lagrange multiplier, L represents the Lagrange function, and K represents the kernel function;

[0126] S03: take the partial derivative of the Lagrange function and set it to zero to obtain a set of optimization conditions; the conditions include:

[0127]

[0128] γe i -ai = 0;

[0129]

[0130] S04: Obtain a linear equation set by arranging the above optimization conditions;

[0131] I represents a unit matrix, represents an m x m matrix, where m is the number of training samples, the element of is the value of the kernel function K(x i , x j ), which represents the Euclidean distance between the training samples x i and x j .

[0132] S05: Solve the above linear equation set using Gaussian elimination method to obtain the parameters of the model, including the weight vector w, the bias term b and the Lagrange multiplier vector a;

[0133] S06: Evaluate the trained model using the training set data and optimize the model.

[0134] Through the above scheme, the loop optimization layer can be optimized.

[0135] Step 4: Use the cross-trained prediction model to predict the components of the polishing liquid.

[0136] Embodiment 2: The application also provides a polishing liquid monitoring system. The polishing liquid monitoring system comprises an information acquisition module for acquiring mark information and a monitoring module for monitoring the concentrations of components of the polishing liquid based on the polishing liquid monitoring method of embodiment 1.

[0137] The above description is only some of the preferred embodiments of the present application and the explanation of the technical principles applied. Those skilled in the art should understand that the scope of the application involved in the embodiments of the present application is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with the technical features disclosed in the embodiments of the present application (but not limited to) having similar functions to form technical solutions.

Claims

1. A polishing liquid monitoring method, characterized in that: The steps include: Step 1: Obtaining composition data of the polishing liquid as it is used and label data corresponding to the composition data, wherein the composition data includes the concentration of each component, and the label data includes pH value, spectral information, and temperature data; Step 2: Obtain time-temperature data, time-pH data, and time-spectrum information of the polishing liquid to be tested as prediction information; Step 3: Build a component prediction model, use the prediction information as the input of the prediction model, and the component data as the output, and cross-train the prediction model; Among them, the prediction model includes LSTM neural network, convolutional neural network, fully connected layer and loop optimization layer; The time-temperature data and time-pH data are input into the LSTM neural network to obtain the first high-level features; The temporal spectral information is input into the convolutional neural network to obtain the second high-level features; The first high-level feature and the second high-level feature are input to the fully connected layer to generate a fusion feature; The fused features are input to the loop optimization layer to generate component data; Step 4: Use the cross-trained prediction model to predict the composition of the polishing liquid; The following steps are performed when acquiring spectral information: S1: Obtain a polishing liquid sample, irradiate the polishing liquid sample with a laser of a fixed wavelength, and collect a spectrum of the laser passing through the polishing liquid sample; S2: According to the light intensity signal distribution in the spectrum, the spectrum is intercepted to obtain the region of interest; S3: median filtering is performed on the region of interest to obtain a filtered spectrum; S4: Binarize the filtered spectrum and extract the center of the filtered spectrum; S5: Extract the maximum spot and saturation spot of the filtered spectrum, mark the center of the circle, the maximum spot and the saturation spot in the region of interest, and use the region of interest as spectral information.

2. The polishing liquid monitoring method according to claim 1, wherein: Step 1 includes the following steps: Step 11: Obtain in advance the same type of polishing liquid, the content of each component during use, and the corresponding pH value and temperature data; Step 12: According to the requirements of the polishing liquid, set the sampling range and prediction accuracy of each component, and generate the marking points of each component according to the sampling range and prediction accuracy of each component.

3. The polishing liquid monitoring method according to claim 1, wherein: The window of the median filter is 3×3.

4. The polishing liquid monitoring method according to claim 1, wherein: When extracting the center of the filtered spectrum, the threshold of the binarization process is the median of the intensity range.

5. The polishing liquid monitoring method according to claim 1, wherein: When extracting the maximum light spot, the threshold of the binarization process is the maximum value of the intensity of the background area within the intensity range.

6. The polishing liquid monitoring method according to claim 1, wherein: When extracting saturated light spots, the threshold of the binarization process is three-quarters of the maximum value of the intensity range.

7. The polishing liquid monitoring method according to claim 1, wherein: Convolutional network models include: Input layer, the input layer has three channels, which are used to input the center mask of the region of interest and the filter spectrum, the mask of the region of interest and the maximum spot, and the saturation spot mask of the region of interest and the filter spectrum; Convolutional layer, used to perform convolution operation on the information input by the input layer; Pooling layer, which performs pooling operation on the output of the convolutional layer and outputs feature maps; The fully connected layer classifies the feature map to generate the second high-level features.

8. A polishing liquid monitoring system, characterized in that: include: An information acquisition module, used to obtain marking information; The monitoring module monitors the concentration of each component of the polishing liquid based on the polishing liquid monitoring method according to any one of claims 1 to 7.

Images