Method and system for purifying aluminum oxide mother liquor

Abstract

The invention relates to the technical field of hyperspectrum, in particular to an aluminum oxide mother liquor purification method and system.The method specifically comprises the steps that hyperspectral images of current evaporation mother liquor at all moments are collected; constructing a noise source coefficient for the similarity of the pixel hyperspectral data and the regularity of the hyperspectral data of each pixel; clustering pixels in each hyperspectral image, and determining a noise exact coefficient of the current evaporation mother liquor by combining the noise source coefficient according to pixel distribution discrete characteristics in a cluster and the number of pixel categories; on the basis of data differences inside and outside the clustering clusters and data differences inside the clustering clusters at different moments, impurity detection is carried out on the current evaporation mother liquor in combination with a noise exact coefficient, and the mother liquor with impurities detected out is purified; the impurity influence is reduced, the sodium carbonate content in the current evaporation mother liquor can be more accurately judged, and the accuracy of a mother liquor detection result is improved.

Description

Technical Field

[0001] This application relates to the field of hyperspectral technology, specifically to a method and system for purifying alumina mother liquor. Background Technology

[0002] my country boasts abundant bauxite reserves and is the world's largest producer of alumina. The Bayer process and its modifications are widely used alumina production methods both domestically and internationally. In the Bayer process, the purification of the alumina mother liquor mainly includes washing, evaporation, and separation, ultimately producing evaporation mother liquor and a series of other substances. The evaporation mother liquor can be recycled after purification. Each purification cycle may still contain a certain amount of sodium carbonate, which can be recovered by adding lime milk for causticization. The amount of lime milk added needs to be adjusted based on the sodium carbonate content in the mother liquor. Therefore, component analysis of the mother liquor is necessary.

[0003] Traditional techniques for component detection in evaporation mother liquor typically employ hyperspectral imaging equipment to acquire hyperspectral data and adjust the amount of lime slurry added based on actual conditions. However, during the hyperspectral data acquisition process, noise interference may occur, leading to some deviation in the determination of sodium carbonate content. Therefore, noise reduction algorithms are used for processing. But in addition to noise interference, the evaporation mother liquor may still contain fine impurities suspended in it due to incomplete evaporation, solubility changes caused by temperature variations, etc. The hyperspectral data of these impurities may interfere with the detection results, ultimately reducing the accuracy of sodium carbonate content detection and affecting subsequent production. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide a method and system for purifying alumina mother liquor, the specific technical solution of which is as follows: In a first aspect, embodiments of this application provide a method for purifying alumina mother liquor, the method comprising the following steps: Acquire hyperspectral images of the currently evaporated mother liquor at various time points; For all hyperspectral images, the noise source coefficient of the current evaporated mother liquor is constructed based on the similarity between hyperspectral data of pixels at the same location at different times and the overall regularity of the distribution of hyperspectral data of each pixel. Cluster all pixels in each hyperspectral image, and determine the exact noise coefficient of the current evaporated mother liquor based on the degree of disorder in the distribution of hyperspectral data differences between pixels in each cluster and the number of pixel categories, combined with the noise source coefficient. Based on the difference in intra-cluster distance between the hyperspectral data at the beginning and end of the time, and the degree of difference in hyperspectral data between intra-cluster and extra-cluster pixels, the impurity interference coefficient of the current evaporated mother liquor is constructed. Based on the noise exact coefficient and the impurity interference coefficient, impurities are detected in the current evaporated mother liquor, and the mother liquor containing impurities is purified.

[0005] In one embodiment, the process of obtaining the noise source coefficient is as follows: The similarity between hyperspectral data of pixels with the same coordinates in the hyperspectral image at each time step and the previous time step is calculated by a similarity algorithm and denoted as the first similarity. The fitting curves of hyperspectral data for each pixel in each hyperspectral image are obtained by curve fitting algorithm, and the mean goodness of fit of the fitting curves of pixels with the same coordinates in the hyperspectral image at each time step and the previous time step is calculated. The noise source coefficient of the current evaporated mother liquor is determined based on the mean goodness of fit of all pixels at all times and the similarity. The noise source coefficient is positively correlated with the similarity and negatively correlated with the mean goodness of fit.

[0006] In one embodiment, the process of obtaining the noise source coefficient of the current evaporated mother liquor is as follows: Calculate the ratio of the first similarity of each pixel at each time step to the mean of the goodness of fit, and calculate the fusion value of the ratio of all pixels at all times step by step, which is used as the noise source coefficient of the current evaporated mother liquor.

[0007] In one embodiment, the clustering of all pixels in each hyperspectral image specifically involves: The mean reflectance of the hyperspectral data of each pixel is calculated, and the mean reflectance of all pixels in each hyperspectral image is used as the input of the clustering algorithm to perform clustering.

[0008] In one embodiment, the process of obtaining the exact noise coefficient is as follows: The difference distance between the hyperspectral data of any two pixels in each cluster is calculated using a distance metric algorithm; the variance of all the difference distances in each cluster is then calculated. The noise exact coefficient of the current evaporating mother liquor is positively correlated with the noise source coefficient, and negatively correlated with the number of clusters in each hyperspectral image and the variance of each cluster.

[0009] In one embodiment, the process of obtaining the impurity interference coefficient is as follows: The hyperspectral data of all pixels in each cluster are fitted using a curve fitting algorithm to obtain the fitting curve of each cluster, which is denoted as the representative spectral data curve. In the hyperspectral image at the last moment, the similarity between the hyperspectral data fitting curve of each pixel outside any cluster and the representative spectral data curve of any cluster is calculated and denoted as the second similarity. Calculate the mean of all the difference distances within each cluster, and denot it as the first mean; compare the magnitude of the first mean of the corresponding clusters at the first time point and the last time point, and determine the impurity interference coefficient of the current evaporated mother liquor based on the comparison result and the second similarity.

[0010] In one embodiment, the process of obtaining the impurity interference coefficient is as follows: Calculate the sign function of the difference between the first mean of each cluster at the first time step and the first mean of the corresponding cluster at the last time step; calculate the fusion value of the second similarity of all clusters at the last time step; and perform an inverse proportional mapping between the sign function corresponding to each cluster at the last time step and the fusion value. The impurity interference coefficient is proportional to the result of the inverse proportional mapping of all clusters at the last time step.

[0011] In one embodiment, the process of detecting impurities in the current evaporation mother liquor is as follows: The normalized value of the ratio of the impurity interference coefficient to the noise exact coefficient is calculated and denoted as the mother liquor reprocessing coefficient. The presence of impurities in the currently evaporated mother liquor is determined by the mother liquor reprocessing coefficient.

[0012] In one embodiment, determining whether there are impurities in the current evaporated mother liquor using the mother liquor reprocessing coefficient specifically involves: If the mother liquor reprocessing coefficient is less than the preset segmentation threshold, it is determined that there are no impurities in the current evaporated mother liquor; otherwise, it is determined that there are impurities in the current evaporated mother liquor.

[0013] Secondly, embodiments of this application also provide an alumina mother liquor purification system, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0014] The embodiments of this application have at least the following beneficial effects: This application acquires hyperspectral images of the current evaporating mother liquor at various times. Based on the similarity of pixel hyperspectral data and the regularity of each pixel's hyperspectral data, a noise source coefficient is constructed to initially distinguish the interference of noise and impurities on the mother liquor's hyperspectral data. Pixels in each hyperspectral image are clustered. Based on the discrete characteristics of pixel distribution within clusters, the differences in data inside and outside clusters, and the differences in data within clusters at different times, combined with the aforementioned noise source coefficient, impurity detection is performed on the current evaporating mother liquor. This considers the pixel hyperspectral distribution characteristics when impurities are present in the mother liquor, further distinguishing the interference of noise and impurities on the mother liquor's hyperspectral data. The type of interference present in the currently acquired mother liquor hyperspectral image data can be determined, and corresponding methods can be used to eliminate the interference, thereby more accurately determining the sodium carbonate content in the current evaporating mother liquor. This solves the problem that current traditional methods for determining the sodium carbonate content in evaporating mother liquor cannot distinguish whether the hyperspectral image data of the evaporating mother liquor is affected by noise or impurity hyperspectral data, which may lead to impurity interference affecting subsequent processing. The mother liquor containing impurities is purified to reduce the impact of impurities and improve the accuracy of mother liquor test results. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating the steps of an alumina mother liquor purification method provided in one embodiment of this application; Figure 2 This is a schematic diagram illustrating the process of obtaining the noise source coefficient. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an alumina mother liquor purification method and system proposed according to this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of the alumina mother liquor purification method and system provided in this application.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a method for purifying alumina mother liquor according to an embodiment of this application. The method includes the following steps: Step S1: Acquire hyperspectral images of the current evaporated mother liquor at each time point.

[0021] A high-definition hyperspectral camera is used to acquire hyperspectral image data of the alumina mother liquor after the evaporation and separation process, with a time interval T. In this application, data acquisition begins after the separation is completed, and hyperspectral image data is acquired at a total of Z time points. It should be noted that the values ​​of T and Z can be set by the implementer according to the actual situation. Preferably, in this embodiment, the value of T is set to 0.5s and the value of Z is set to 50.

[0022] Step S2: For all hyperspectral images, construct the noise source coefficient of the current evaporated mother liquor based on the similarity between hyperspectral data of pixels at the same location at different times and the overall regularity of the distribution of hyperspectral data of each pixel.

[0023] For hyperspectral data of evaporation mother liquor, when noise interference is present, the reflectance data of each pixel in the acquired hyperspectral data will contain numerous reflection peaks of varying sizes, exhibiting high randomness. However, once the noise is generated, the overall change is relatively small over time. When there are many impurities in the evaporation mother liquor, the liquid flow within the mother liquor causes random displacement of the impurity components. Consequently, various components will appear at different locations over time, creating a hyperspectral data variation similar to that caused by random noise interference. While noise interference can be easily addressed by using denoising algorithms to determine the sodium carbonate content in the mother liquor, impurity interference present requires more sophisticated denoising algorithms. These algorithms not only fail to accurately remove interference from the hyperspectral data but may also discard valid data, affecting data integrity and leading to misjudgments of the current state of the mother liquor. Therefore, it is necessary to first distinguish between noise and impurity interference in the hyperspectral data of the mother liquor.

[0024] Specifically, when the interference source in the hyperspectral data of the evaporating mother liquor is noise rather than impurities, the hyperspectral data mainly consists of sodium carbonate and noise. This manifests primarily as spikes in the hyperspectral data curve of a single pixel. Since the acquisition conditions for the hyperspectral data are consistent, the degree of noise impact is similar. Therefore, for a single pixel in the hyperspectral image data, the morphological changes of the hyperspectral data at multiple time points are relatively small, with the main change being small variations in noise. Furthermore, when the interference source is impurities rather than noise, the evaporating mother liquor actually contains many different substances. Each substance has different reflectance in different bands of the hyperspectral data, resulting in significant differences in the overall morphological characteristics of the hyperspectral data between different substances. As substances move within the evaporating mother liquor, the location of a single pixel in the hyperspectral image may contain different substances at different time points, leading to significant differences in the morphological characteristics of the hyperspectral data contained in a single pixel at different times.

[0025] To characterize the above features, for two pixels with the same coordinates in the hyperspectral image at each time step and the previous time step, the cosine similarity between the hyperspectral data of these two pixels is calculated and denoted as the first similarity. The cosine similarity is a well-known technique, and its specific process will not be elaborated further.

[0026] It should be noted that this application provides only one similarity algorithm for the similarity between hyperspectral data of two pixels. There are many existing similarity algorithms, and implementers may also use other similarity algorithms to calculate the similarity between hyperspectral data of two pixels. This application does not impose any specific restrictions.

[0027] For a single pixel in the hyperspectral image data at a single time point, the hyperspectral data it contains is used as input, and a curve fitting algorithm is used to perform curve fitting to obtain the goodness of fit of the fitted curve. Furthermore, the average goodness of fit of the fitted curves for pixels at the same coordinates in the hyperspectral image at each time point and the previous time point is calculated. The goodness of fit method is a well-known technique, and the specific process will not be elaborated further.

[0028] It should be noted that there are many existing curve fitting algorithms. Preferably, in this application embodiment, the least squares method is used to perform curve fitting on the hyperspectral data of a single pixel. Implementers may also use other curve fitting algorithms to perform curve fitting on the hyperspectral data of a single pixel. This application does not impose any specific restrictions.

[0029] Furthermore, the ratio of the first similarity of each pixel at each time point to the mean of the goodness of fit is calculated, and the fusion value of the ratio of all pixels at all time points is calculated as the noise source coefficient of the current evaporated mother liquor, to characterize the source of interference to the current evaporated mother liquor. The fusion can be addition, multiplication, averaging, etc. Preferably, in this embodiment, the expression for the noise source coefficient is:

[0030] In the formula, The noise source coefficient of the current evaporated mother liquor; I is the total number of pixels in a single hyperspectral image data; J is the total number of sampling times; The first similarity is the difference between the i-th pixel in the hyperspectral image at time j and the i-th pixel in the hyperspectral image at time j-1. The mean of the goodness of fit of the hyperspectral data fitting curves for the i-th pixel in the hyperspectral image data at time j and time j-1 is given.

[0031] When the hyperspectral image data of the evaporated mother liquor is subject to noise interference, for a single pixel, the overall spectral data at multiple time points are more approximate, and the better the curve fitting of the hyperspectral data contained in a single pixel, the lower the goodness of fit. The larger the value, the greater the noise source coefficient, indicating that there is a greater possibility of noise in the current hyperspectral image data and a smaller possibility of interference from impurities.

[0032] Step S3: Cluster all pixels in each hyperspectral image. Based on the degree of disorder in the distribution of hyperspectral data differences between pixels in each cluster and the number of pixel categories, combined with the noise source coefficient, determine the exact noise coefficient of the current evaporated mother liquor.

[0033] If there are not many impurities in the mother liquor, the main component of the mother liquor is sodium carbonate, and the overall distribution is relatively uniform. Even with noise interference, the approximation of each pixel in the hyperspectral image data is relatively high, with only a few pixels that are significantly affected by noise. However, if there are many impurities in the mother liquor, the composition of the mother liquor is complex, and various substances have a certain proportion in the hyperspectral image data. Ultimately, this manifests as a discrete distribution of pixels containing multiple substances in the hyperspectral image data of the mother liquor.

[0034] To characterize the aforementioned features, for a single pixel in the hyperspectral image data of the evaporated mother liquor at a single time point, the mean value of the reflectance of all wavelengths in the hyperspectral data of that pixel is calculated as the average reflectance of that pixel. Furthermore, using the average reflectance of all pixels as input, the DPC clustering algorithm is used. The absolute value of the difference in average reflectance between pixels is used as the metric distance between pixels, with a cutoff distance of 3. The output consists of multiple pixel clusters, and each cluster represents a pixel of one substance in the evaporated mother liquor. The DPC clustering algorithm is a well-known technique, and its specific process will not be elaborated further.

[0035] It should be noted that this application only provides one clustering method for clustering all pixels in a hyperspectral image. There are many existing clustering methods, and implementers may also use other clustering algorithms to cluster all pixels in a hyperspectral image. This application does not impose any specific restrictions.

[0036] The difference distance between the hyperspectral data of any two pixels within each cluster is calculated. Preferably, in this embodiment, the difference distance is the Euclidean distance. In other embodiments of this application, the implementer may also use other distance metric algorithms to calculate the difference distance between the hyperspectral data of two pixels. Further, the degree of disorder in the distribution of all the difference distances within each cluster is analyzed. The degree of disorder can be variance, standard deviation, and information entropy, etc. Preferably, in this embodiment, the degree of disorder is the variance of all the difference distances within each cluster, denoted as the intra-cluster distance variance.

[0037] Based on the above analysis, the exact noise coefficient of the current evaporated mother liquor is calculated to characterize the sources of interference in the hyperspectral image data of the evaporated mother liquor. Preferably, in this embodiment, the expression for the exact noise coefficient of the current evaporated mother liquor is:

[0038] In the formula, The exact noise coefficient of the current evaporated mother liquor; The noise source coefficient of the current evaporated mother liquor; J is the total number of sampling times; Let be the total number of pixel clusters in the hyperspectral image data at time j; Let be the mean of the intra-cluster distance variances of all clusters in the hyperspectral image data at time j.

[0039] In other embodiments of this application, the expression for the exact noise coefficient may also be: ,in, for The normalized value, for The normalized value is obtained by using the Z-score function as the normalization function in this embodiment. The normalized value of the value at each time moment is obtained by using the values ​​at all times as input to the Z-score function.

[0040] If the interference in the hyperspectral image data of the evaporating mother liquor is noise, and the mother liquor contains few impurities and has a uniform distribution of substances (i.e., the fewer the total number of clusters obtained after dividing the pixels into clusters according to the differences in hyperspectral data reflectance), and the more even the spacing between pixels within each cluster, the smaller the mean variance of the Euclidean distance between pixels within each cluster, the smaller the total number of clusters, and the larger the noise exact coefficient, the greater the likelihood that the interference in the current hyperspectral image data of the evaporating mother liquor is noise, and a denoising algorithm should be used directly for subsequent processing.

[0041] Step S4: Based on the difference in intra-cluster distance between the hyperspectral data at the beginning and end of the time interval, and the degree of difference in hyperspectral data between intra-cluster and extra-cluster pixels, construct the impurity interference coefficient of the current evaporating mother liquor.

[0042] During the evaporation and separation process of the mother liquor, the temperature of the mother liquor changes significantly, and the precipitation of various components varies. Consequently, components that were initially only weakly precipitated or not precipitated at all may gradually precipitate after separation (e.g., separation is performed after evaporation, but the temperature of the mother liquor is still high, so some substances will continue to precipitate after further water loss). This results in the presence of certain impurities in the mother liquor, and further precipitation may occur on the surface of these impurities, eventually leading to the formation of relatively large crystals of certain impurities. Due to their relatively large size, these impurities move slowly in the mother liquor, and as the mother liquor gradually becomes calmer over time, they gradually settle due to gravity. Consequently, during the sampling time of hyperspectral image data, the hyperspectral data of multiple pixels in a local area at multiple times are highly approximate. However, when calculating using only the above steps, there is still a certain deviation, so further analysis is required.

[0043] Specifically, when impurities still precipitate after the evaporation and separation process of the mother liquor, and the degree of aggregation gradually increases, in the hyperspectral image data of the mother liquor, the pixels contained in a certain cluster gradually aggregate into a certain area over time, mainly manifested as the distance between pixels within the cluster gradually decreasing. Furthermore, if there is interference from impurities in the mother liquor, the differences in hyperspectral data between impurities are significant due to their obvious compositional differences. To characterize these features, the hyperspectral data of all pixels in a single cluster are merged. In this embodiment, the merging method involves calculating the average reflectance of the same wavelength in the hyperspectral data of all pixels in the cluster to obtain the merged result of the hyperspectral data of all pixels in the cluster. The merged hyperspectral data is then curve-fitted, and the fitting result is used as the representative spectral data curve of the cluster.

[0044] For the hyperspectral image at the last moment, calculate the cosine similarity between the representative spectral data curve of each cluster and the fitted hyperspectral data curve of each pixel outside the cluster, and denote it as the second similarity.

[0045] It should be noted that this application provides only one similarity algorithm for the similarity between the representative spectral data curve and the fitted curve of each pixel. There are many existing similarity algorithms, and implementers may also use other similarity algorithms to calculate the similarity between the representative spectral data curve and the fitted curve of each pixel. This application does not impose any specific restrictions.

[0046] The difference distances within each cluster at the first time step are compared with the difference distances within the corresponding clusters at the last time step. Specifically, the difference distance between any two pixels within each cluster is obtained, and the mean of all such difference distances within each cluster is calculated and denoted as the first mean. Further, a sign function is calculated for the difference between the first mean of each cluster at the first time step and the first mean of the corresponding cluster at the last time step. The comparison result is determined by the sign function. The sign function is a known technique, and the specific process will not be elaborated further.

[0047] Calculate the fusion value of the second similarity for all clusters at the last time step; then perform an inverse proportional mapping between the sign function corresponding to each cluster at the last time step and the fusion value. The inverse proportional mapping can be subtraction, division, negative exponentiation, etc.

[0048] The impurity interference coefficient of the current evaporating mother liquor is calculated based on the sign function and the result of the inverse proportional mapping to characterize whether there is impurity interference in the current evaporating mother liquor. Preferably, in this embodiment of the application, the expression for the impurity interference coefficient of the current evaporating mother liquor can be:

[0049] In the formula, R is the impurity interference coefficient of the current evaporated mother liquor; R is the total number of clusters in the hyperspectral image at time Z. Let be the first mean of the r-th cluster in the hyperspectral image at time Z, where r is the number of the r-th cluster in the hyperspectral image; The first mean of the r-th cluster in the hyperspectral image at the first time step; Let be the cosine similarity between the representative spectral data curve of the r-th cluster and the hyperspectral data curve of the p-th pixel outside the cluster in the hyperspectral image at time Z; P is the total number of pixels in the hyperspectral image at time Z excluding the pixels of the r-th cluster. It is a symbolic function.

[0050] In this embodiment, the numbering method for each cluster in the hyperspectral image is as follows: the coordinates of the pixel corresponding to the cluster center of each cluster are obtained as the coordinates of that cluster. All clusters in the hyperspectral image are sorted and numbered sequentially according to their coordinates from top to bottom and left to right, thereby obtaining the number of each cluster in the hyperspectral image. Implementers may also use other methods for numbering, and this application does not impose specific restrictions.

[0051] When impurities still precipitate after evaporation and separation of the mother liquor, the pixel clusters in the hyperspectral image data of the mother liquor will become denser over time. Moreover, due to the complexity of the impurity composition, the hyperspectral data contained in the pixel is more significantly different from the hyperspectral data of other pixels, and the similarity is lower. This indicates that the higher the probability of impurities in the current mother liquor, the greater the possibility that the hyperspectral data will be affected by the spectral data of impurities.

[0052] In other embodiments of this application, the expression for the impurity interference coefficient may also be: ,in, is the mean of all the second similarities in the r-th cluster of the hyperspectral image at time Z.

[0053] Step S5: Based on the noise exact coefficient and the impurity interference coefficient, perform impurity detection on the current evaporated mother liquor, and purify the mother liquor containing impurities.

[0054] Based on the above analysis, the reprocessing coefficient of the current evaporated mother liquor is calculated as follows:

[0055] In the formula, The mother liquor reprocessing coefficient is the current evaporation mother liquor. The impurity interference coefficient of the mother liquor; The exact coefficient of noise in the evaporated mother liquor; This is the normalization function. Wherein, This is denoted as the first ratio. In this embodiment, the first ratio of all historically evaporated mother liquors is obtained, and then normalized using a maximum-minimum normalization method to the first ratio of the current evaporated mother liquor. Many existing normalization methods exist, and implementers may also use other normalization algorithms to normalize the first ratio; this application does not impose specific limitations.

[0056] When the likelihood of impurities in the mother liquor is high and the likelihood of noise is low, the determination of the sodium carbonate content in the mother liquor is more likely to be interfered with by the hyperspectral data of impurities. In this case, the mother liquor should be further purified, such as by re-evaporation or separation, in order to eliminate the influence of impurities on the determination of the sodium carbonate content in the mother liquor.

[0057] Acquire labeled hyperspectral image data of all historical evaporation mother liquors, calculate the mother liquor reprocessing coefficient for each historical evaporation mother liquor, use all mother liquor reprocessing coefficients as input, and use cross-validation to output the segmentation threshold of the mother liquor reprocessing coefficient.

[0058] If the reprocessing coefficient of the current evaporating mother liquor is greater than or equal to the segmentation threshold, it indicates that there are impurities in the current evaporating mother liquor. When judging the sodium carbonate content in the mother liquor, it is easy to be interfered with by the hyperspectral data of impurities. Further purification treatment (evaporation, separation, etc.) is carried out on the mother liquor until the reprocessing coefficient of the mother liquor is less than the segmentation threshold to eliminate the interference of impurities.

[0059] When the mother liquor reprocessing coefficient is less than the segmentation threshold, it is determined that there are no impurities in the current evaporated mother liquor, only noise interference. In this case, a denoising algorithm can be used to denoise the hyperspectral image data, and then the sodium carbonate content in the mother liquor can be further determined. The denoising algorithm used in this application embodiment is non-local mean denoising. There are many existing denoising algorithms, and implementers can also use other denoising algorithms to denoise the hyperspectral image data. This application does not impose specific limitations.

[0060] The purification process is as follows: 1. The mother liquor sent from the decomposition workshop enters the evaporation raw liquid tank. Part of the mother liquor is used to adjust the solid content of the slurry, and the other part enters the evaporation system.

[0061] 2. The mother liquor enters the evaporation station, which consists of a multi-effect tubular falling film evaporator and a flash evaporator, using a counter-current process. The raw liquor is heated and evaporated in stages, while the discharge temperature is monitored by a smart temperature sensor. The discharge temperature is about 140°C, and the liquor enters the flash evaporation system. The flash evaporation discharge temperature is about 82°C, and the liquor is sent to the salt settling tank.

[0062] 3. After evaporation, the bottom flow of the salt settling tank is separated by a centrifuge, and the solid particles enter the causticizing tank for causticizing (i.e., lime milk is added and sodium carbonate monohydrate is reacted).

[0063] Furthermore, after eliminating potential impurities or noise interference in the hyperspectral data, standard hyperspectral data of the evaporation mother liquor is obtained. Combined with the hyperspectral data of the currently processed evaporation mother liquor, the sodium carbonate content in the current mother liquor is calculated using the characteristic band iPLS model and Beer-Lambert law. Based on the calculation results and the addition ratio of sodium carbonate to lime slurry, the alumina production system sends a control signal to the lime slurry metering pump. The frequency converter of the lime slurry metering pump, based on the control signal, reduces or increases the speed of the lime slurry metering pump, thereby reducing or increasing the amount of lime slurry added. At the same time, an intelligent sensor module monitors the actual concentration of lime slurry. If the concentration is less than the preset value, the flow rate of the metering pump needs to be increased synchronously; otherwise, the flow rate of the metering pump is reduced to ensure the accuracy of the actual total amount of lime slurry delivered.

[0064] The purification of the mother liquor, the calculation of sodium carbonate content, and the adjustment of lime slurry dosage are all well-known technologies, and the specific processes will not be described in detail.

[0065] A schematic diagram of the process for obtaining the noise source factor is shown below. Figure 2 As shown.

[0066] Based on the same inventive concept as the above method, this application embodiment also provides an alumina mother liquor purification system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described alumina mother liquor purification methods.

[0067] In summary, this application provides a method for purifying alumina mother liquor. The method involves acquiring hyperspectral images of the currently evaporated mother liquor at various times. Based on the similarity and regularity of the hyperspectral data of each pixel, a noise source coefficient is constructed to initially distinguish between noise and impurities interfering with the mother liquor's hyperspectral data. Pixels in each hyperspectral image are clustered. Based on the discrete characteristics of pixel distribution within clusters, data differences inside and outside clusters, and data differences within clusters at different times, combined with the noise source coefficient, impurity detection is performed on the current evaporated mother liquor. This method considers the pixel hyperspectral distribution characteristics when impurities are present in the mother liquor, further distinguishing between noise and impurities interfering with the mother liquor's hyperspectral data. It can determine the type of interference present in the currently acquired mother liquor hyperspectral image data, thereby using corresponding methods to eliminate interference and more accurately determining the sodium carbonate content in the current evaporated mother liquor. This solves the problem that traditional methods for determining the sodium carbonate content in evaporated mother liquor cannot distinguish whether the hyperspectral image data is affected by noise or impurities, which could lead to impurities interfering with subsequent processing. The mother liquor containing impurities is purified to reduce the impact of impurities and improve the accuracy of mother liquor test results.

[0068] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this application. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0069] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0070] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A method for purifying alumina mother liquor, characterized in that, The method includes the following steps: Acquire hyperspectral images of the currently evaporated mother liquor at various time points; For all hyperspectral images, the noise source coefficient of the current evaporated mother liquor is constructed based on the similarity between hyperspectral data of pixels at the same location at different times and the overall regularity of the distribution of hyperspectral data of each pixel. Cluster all pixels in each hyperspectral image, and determine the exact noise coefficient of the current evaporated mother liquor based on the degree of disorder in the distribution of hyperspectral data differences between pixels in each cluster and the number of pixel categories, combined with the noise source coefficient. Based on the difference in intra-cluster distance between the hyperspectral data at the beginning and end of the time, and the degree of difference in hyperspectral data between intra-cluster and extra-cluster pixels, the impurity interference coefficient of the current evaporated mother liquor is constructed. Based on the noise exact coefficient and the impurity interference coefficient, impurities are detected in the current evaporated mother liquor, and the mother liquor containing impurities is purified.

2. The method for purifying alumina mother liquor as described in claim 1, characterized in that, The process of obtaining the noise source coefficient is as follows: The similarity between hyperspectral data of pixels with the same coordinates in the hyperspectral image at each time step and the previous time step is calculated by a similarity algorithm and denoted as the first similarity. The fitting curves of hyperspectral data for each pixel in each hyperspectral image are obtained by curve fitting algorithm, and the mean goodness of fit of the fitting curves of pixels with the same coordinates in the hyperspectral image at each time step and the previous time step is calculated. The noise source coefficient of the current evaporated mother liquor is determined based on the mean goodness of fit of all pixels at all times and the similarity. The noise source coefficient is positively correlated with the similarity and negatively correlated with the mean goodness of fit.

3. The method for purifying alumina mother liquor as described in claim 2, characterized in that, The process for obtaining the noise source coefficient of the current evaporated mother liquor is as follows: Calculate the ratio of the first similarity of each pixel at each time step to the mean of the goodness of fit, and calculate the fusion value of the ratio of all pixels at all times step by step, which is used as the noise source coefficient of the current evaporated mother liquor.

4. The method for purifying alumina mother liquor as described in claim 1, characterized in that, The clustering of all pixels in each hyperspectral image specifically involves: The mean reflectance of the hyperspectral data of each pixel is calculated, and the mean reflectance of all pixels in each hyperspectral image is used as the input of the clustering algorithm to perform clustering.

5. The method for purifying alumina mother liquor as described in claim 1, characterized in that, The process for obtaining the exact noise coefficient is as follows: The difference distance between the hyperspectral data of any two pixels in each cluster is calculated using a distance metric algorithm; the variance of all the difference distances in each cluster is then calculated. The noise exact coefficient of the current evaporating mother liquor is positively correlated with the noise source coefficient, and negatively correlated with the number of clusters in each hyperspectral image and the variance of each cluster.

6. The method for purifying alumina mother liquor as described in claim 5, characterized in that, The process for obtaining the impurity interference coefficient is as follows: The hyperspectral data of all pixels in each cluster are fitted using a curve fitting algorithm to obtain the fitting curve of each cluster, which is denoted as the representative spectral data curve. In the hyperspectral image at the last moment, the similarity between the hyperspectral data fitting curve of each pixel outside any cluster and the representative spectral data curve of any cluster is calculated and denoted as the second similarity. Calculate the mean of all the difference distances within each cluster, and denote it as the first mean; The magnitude of the first mean of the corresponding clusters at the first and last time points is compared, and the impurity interference coefficient of the current evaporated mother liquor is determined based on the comparison result and the second similarity.

7. The method for purifying alumina mother liquor as described in claim 6, characterized in that, The process for obtaining the impurity interference coefficient is as follows: Calculate the sign function of the difference between the first mean of each cluster at the first time step and the first mean of the corresponding cluster at the last time step; calculate the fusion value of the second similarity of all clusters at the last time step; and perform an inverse proportional mapping between the sign function corresponding to each cluster at the last time step and the fusion value. The impurity interference coefficient is proportional to the result of the inverse proportional mapping of all clusters at the last time step.

8. The method for purifying alumina mother liquor as described in claim 1, characterized in that, The process of detecting impurities in the current evaporated mother liquor is as follows: The normalized value of the ratio of the impurity interference coefficient to the noise exact coefficient is calculated and denoted as the mother liquor reprocessing coefficient. The presence of impurities in the currently evaporated mother liquor is determined by the mother liquor reprocessing coefficient.

9. The method for purifying alumina mother liquor as described in claim 8, characterized in that, The determination of whether there are impurities in the current evaporated mother liquor by using the mother liquor reprocessing coefficient is specifically as follows: If the mother liquor reprocessing coefficient is less than the preset segmentation threshold, it is determined that there are no impurities in the current evaporated mother liquor; otherwise, it is determined that there are impurities in the current evaporated mother liquor.

10. An alumina mother liquor purification system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-9.

Images