Metallic indium purification control method and system based on historical data analysis

Abstract

The invention relates to the field of metal indium purification control, and discloses a metal indium purification control method and system based on historical data analysis, and according to the technology, dimension reduction is carried out on historical equipment parameters in different time periods through principal component analysis (PCA), and dimension reduction equipment parameter vectors are constructed. Determining a reference total operation time length and a corresponding reference dimension reduction vector, calculating a difference value between the current operation time length and the reference operation time length and a parameter distance, and carrying out weighted fusion on the two operation time lengths to obtain a fusion distance; and determining local density based on the fusion distance, sorting according to the density to obtain a density equipment parameter sequence, determining a measurement distance by combining the parameter distance, and clustering the parameter distance to obtain an equipment state association cluster (including a fault-going cluster and a normal cluster). And acquiring real-time equipment parameters, judging the cluster to which the real-time equipment parameters belong, and adjusting the real-time equipment parameters to a normal cluster. According to the method, purification control optimization of the metal indium can be performed in combination with historical data.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of metal indium purification control, and more particularly to a metal indium purification control method and system based on historical data analysis. BACKGROUND

[0002] In the catalyst recovery process, for the recovery and purification of metal indium, the core processes of pretreatment release, selective leaching and separation and refining are used to achieve high recovery rate and high purity. The catalyst structure is destroyed by crushing, roasting or acid-base dissolution to release indium and remove the carrier interference; indium is dissolved into the liquid phase by acid leaching with hydrochloric acid, sulfuric acid and the like; then impurities are removed by extraction, precipitation or ion exchange, and finally high-purity metal indium is obtained by electrolytic refining.

[0003] The existing metal indium purification control method mainly relies on preset fixed parameters or real-time detection data for simple feedback adjustment, and fails to systematically mine and analyze the data such as process parameters, raw material characteristics, product purity, equipment running state and the like accumulated in the historical production process, so as to extract the key rules and related information affecting the purification effect. This leads to the lack of dynamic adaptability and precision of the purification control strategy, and it is difficult to dynamically adjust the control parameters according to the actual production changes such as raw material composition fluctuation and equipment aging, which may cause problems such as unstable product purity, low purification efficiency, high energy consumption and increased production cost, thereby restricting the further upgrading and quality improvement of the metal indium purification process. SUMMARY

[0004] The application aims to provide a metal indium purification control method and system based on historical data analysis, which solves the technical problem of being unable to optimize the purification control of metal indium in combination with historical data, and achieves the technical effect of optimizing the purification control of metal indium in combination with historical data.

[0005] In a first aspect, the embodiments of the present application provide a metal indium purification control method based on historical data analysis, the method comprising: obtaining a plurality of historical device parameters of metal indium purification in different time periods; wherein different time periods correspond to different total device running time lengths, and the historical device parameters include electrolyte temperature, electrolyte concentration, cell voltage, current density, electrolytic device temperature, electrolytic cell vibration parameters, stirring parameters, indium deposition rate, and indium deposition purity; performing principal component analysis dimension reduction on the plurality of historical device parameters in different time periods to determine target device parameter principal components with a cumulative variance contribution rate greater than or equal to a preset cumulative variance contribution rate; constructing a dimension-reduced device parameter vector corresponding to the target device parameter principal components in different time periods; obtaining a reference dimension-reduced device parameter vector corresponding to a reference total device running time length and a normal device state; determining a difference between the total device running time length and the reference total device running time length as a total device running time length difference; determining a distance between the reference dimension-reduced device parameter vector and the dimension-reduced device parameter vector as a dimension-reduced device parameter distance; weighting and fusing the total device running time length difference and the dimension-reduced device parameter distance in different time periods to obtain a fusion distance in different time periods; determining a local density based on the fusion distance, and sorting the local density from high to low as a density device parameter sequence; determining a metric distance corresponding to the plurality of dimension-reduced device parameter distances according to the plurality of dimension-reduced device parameter distances and the density device parameter sequence; clustering the plurality of dimension-reduced device parameter distances according to the metric distance to obtain a plurality of device state associated clusters; determining a device about to fail cluster and a device normal cluster in the plurality of device state associated clusters; obtaining real-time device parameters, and adjusting the real-time device parameters to the device normal cluster when the real-time device parameters belong to the device about to fail cluster, and issuing a prompt information that the device parameters are being adjusted.

[0006] In a possible implementation, according to the plurality of dimension-reduced device parameter distances and the density device parameter sequence, the metric distance corresponding to the plurality of dimension-reduced device parameter distances is determined, comprising: determining a density penalty factor corresponding to the plurality of historical device parameters according to the density device parameter sequence; wherein the greater the density of the device parameters in the density device parameter sequence, the smaller the density penalty factor; determining a product of the dimension-reduced device parameter distance and the density penalty factor corresponding to the plurality of historical device parameters as the metric distance corresponding to the plurality of dimension-reduced device parameter distances.

[0007] In another possible implementation, the first target principal component includes electrolytic device temperature and electrolytic cell vibration parameters; and the second target principal component includes electrolyte concentration and current density.

[0008] In another possible implementation, the method further includes: obtaining a plurality of historical device parameters of the purification of indium metal in different time periods; wherein the different time periods correspond to different total device running lengths, and the historical device parameters include electrolyte temperature, electrolyte concentration, cell voltage, current density, stirring parameters, deposition rate, and deposition purity; performing principal component analysis dimension reduction on the plurality of historical device parameters in the different time periods to determine target principal components with a cumulative variance contribution rate greater than or equal to a preset contribution rate; constructing a dimension-reduced device parameter vector corresponding to the target principal components in the different time periods; obtaining a reference dimension-reduced device parameter vector corresponding to a reference total device running length and a normal device state; determining a difference between the total device running length and the reference total device running length as a total device running length difference value; determining a distance between the reference dimension-reduced device parameter vector and the dimension-reduced device parameter vector as a dimension-reduced device parameter distance; weighting and fusing the total device running length difference values and the dimension-reduced device parameter distances in the different time periods to obtain a fusion distance corresponding to the plurality of historical device parameters; determining a local density of each fusion distance based on the fusion distance, and sorting the local densities from high to low as a density device parameter sequence; determining a mean value of the variance contribution rates of the principal components of the same dimension of the dimension-reduced device parameter vector; constructing a contribution rate weight matrix according to the mean value of the variance contribution rates of the principal components of the same dimension; determining a weighted device parameter vector corresponding to the different time periods according to the dimension-reduced device parameter vector and the contribution rate weight matrix; determining a distance between the weighted device parameter vectors corresponding to the different time periods as a dimension-reduced device parameter difference degree corresponding to the plurality of historical device parameters; determining a metric distance corresponding to the plurality of dimension-reduced device parameter difference degrees according to the plurality of dimension-reduced device parameter difference degrees and the density device parameter sequence; clustering the plurality of dimension-reduced device parameter difference degrees according to the metric distance to obtain a plurality of device state associated clusters; determining a device about to fail cluster and a device normal cluster in the plurality of device state associated clusters; obtaining real-time device parameters, and adjusting the real-time device parameters to the device normal cluster when the real-time device parameters belong to the device about to fail cluster, and issuing prompt information that the device parameters are being adjusted.

[0009] In another possible implementation, the determination of the metric distance corresponding to the plurality of dimension-reduced device parameter difference degrees according to the plurality of dimension-reduced device parameter difference degrees and the density device parameter sequence includes: determining a density penalty factor corresponding to the plurality of historical device parameters according to the density device parameter sequence; wherein the greater the density of the device parameters in the density device parameter sequence, the smaller the density penalty factor; and determining a product of the dimension-reduced device parameter difference degree and the density penalty factor corresponding to the plurality of historical device parameters as the metric distance corresponding to the plurality of dimension-reduced device parameter difference degrees.

[0010] In another possible implementation, the first target principal component includes electrolyte concentration and electrolyte temperature, and the second target principal component includes cell voltage and current density.

[0011] In another possible implementation, the method further includes: obtaining a plurality of historical recovery parameters of a metal indium recovery raw material preprocessing link; wherein the historical recovery parameters include leaching agent concentration, leaching agent temperature, indium concentration after purification, and impurity content; performing principal component analysis dimension reduction on the plurality of historical recovery parameters to determine target recovery parameter principal components with a cumulative variance contribution rate greater than or equal to a preset cumulative variance contribution rate; constructing a dimension-reduced recovery parameter vector corresponding to the target recovery parameter principal components in different time periods; clustering the plurality of dimension-reduced recovery parameter vectors to obtain a plurality of recovery parameter association clusters; obtaining a plurality of equipment state association clusters; determining, in the plurality of recovery parameter association clusters, a best recovery parameter association cluster corresponding to the equipment state association cluster; obtaining a current recovery parameter and a current equipment parameter; determining a current equipment state association cluster to which the current equipment parameter belongs, and determining a best recovery parameter association cluster corresponding to the current equipment state association cluster; and adjusting the current recovery parameter to a recovery parameter corresponding to the best recovery parameter association cluster when the current recovery parameter deviates from the best recovery parameter association cluster.

[0012] In another possible implementation, in the plurality of recovery parameter association clusters, determining the best recovery parameter association cluster corresponding to the current equipment state association cluster includes: determining center equipment state parameters corresponding to the plurality of equipment state association clusters, and determining center recovery parameters corresponding to the plurality of recovery parameter association clusters; associating the center equipment state parameters corresponding to the plurality of equipment state association clusters and the center recovery parameters corresponding to the plurality of recovery parameter association clusters to a unified feature space, and determining a center parameter distance between the center equipment state parameters and the center recovery parameters in the unified feature space; in the unified feature space, determining a current center equipment state parameter corresponding to the current equipment state association cluster, and obtaining all center parameter distances corresponding to the current center equipment state parameter; in all the center parameter distances, determining a minimum center parameter distance, and determining a recovery parameter association cluster corresponding to the minimum center parameter distance as the best recovery parameter association cluster.

[0013] In another possible implementation, when the current recovery parameter deviates from the best recovery parameter association cluster, adjusting the current recovery parameter to a recovery parameter corresponding to the best recovery parameter association cluster includes: determining a center recovery parameter of the best recovery parameter association cluster, and determining a distance between the center recovery parameter and the current recovery parameter as a recovery parameter distance; when the recovery parameter distance is greater than or equal to a preset recovery parameter distance, it is determined that the current recovery parameter deviates from the best recovery parameter association cluster, and the current recovery parameter is adjusted to the recovery parameter corresponding to the best recovery parameter association cluster.

[0014] In a second aspect, the embodiments of the present application provide a metal indium purification control system based on historical data analysis, including units for implementing the above method.

[0015] The beneficial effects of the embodiments of the present application compared with the prior art are: The embodiments of the present application provide a metal indium purification control method based on historical data analysis, which retains the core features of historical equipment parameters, eliminates the influence of irrelevant or repeated parameters, and obtains more accurate equipment state association clusters through clustering by fusing time and parameter features, so that the normal state and the state about to fail of the equipment can be effectively distinguished, the parameters are adjusted to the normal range before the equipment is about to fail, the influence of equipment failure on the metal indium purification process is avoided, the stable operation of the purification process is ensured, and the purification efficiency of indium is improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0017] Figure 1 The flowchart of the first metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in the figure. Figure 2 The working flowchart of the first metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in the figure. Figure 3 The flowchart of the second metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in the figure. Figure 4 The working flowchart of the second metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in the figure. Figure 5 The flowchart of the third metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in the figure. Figure 6 The working flowchart of the third metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in the figure. Figure 7 The flowchart of the fourth metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in the figure. Figure 8 The working flowchart of the fourth metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in the figure. Figure 9 The flowchart of the fifth metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in the figure. Figure 10A fifth working flow diagram of a metal indium purification control method based on historical data analysis is provided for the embodiments of the present application. Figure 11 A sixth flow diagram of a metal indium purification control method based on historical data analysis is provided for the embodiments of the present application. Figure 12 A logic structure diagram of a metal indium purification control system based on historical data analysis is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0018] It should be understood that the term "comprises" as used in the specification and the appended claims indicates the presence of the described features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0019] It should also be understood that the term "and / or" as used in the specification and the appended claims, means any one or more of the associated listed items, as well as all possible combinations of the items.

[0020] As used in the specification and the appended claims, the term "if' can be interpreted as meaning "when" or "once" or "in response to a determination" or "in response to a detection" depending on the context. Similarly, the phrase "if determined" or "if detected [the described condition or event]" can be interpreted as meaning "once determined" or "in response to a determination" or "once detected [the described condition or event]" or "in response to a detection [the described condition or event]" depending on the context.

[0021] In addition, in the description of the specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish descriptions, and cannot be understood as indicating or implying relative importance.

[0022] The reference in the specification to "one embodiment" or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. Thus, the appearance of the phrases "in one embodiment", "in some embodiments", "in other embodiments", "in additional embodiments", etc. in various places in the specification is not necessarily all referring to the same embodiment, but means "one or more but not all embodiments", unless otherwise specifically stated. The terms "comprise", "include", "have" and their conjugates mean "including but not limited to", unless otherwise specifically stated.

[0023] The existing metal indium purification control method mainly depends on preset fixed parameters or real-time detection data for simple feedback adjustment, which may cause unstable product purity, low purification efficiency, high energy consumption, and increased production cost.

[0024] Based on the above reasons, the embodiment of the present application provides a metal indium purification control method based on historical data analysis, which comprises: performing principal component analysis dimension reduction on a plurality of historical equipment parameters in different time periods to determine target equipment parameter principal components with a cumulative variance contribution rate greater than or equal to a preset cumulative variance contribution rate; constructing a dimension-reduced equipment parameter vector corresponding to the target equipment parameter principal components in different time periods; obtaining a reference dimension-reduced equipment parameter vector corresponding to a reference total equipment running time and a normal equipment state; determining the difference between the total equipment running time and the reference total equipment running time as a total equipment running time difference; determining the distance between the reference dimension-reduced equipment parameter vector and the dimension-reduced equipment parameter vector as a dimension-reduced equipment parameter distance; weighting and fusing the total equipment running time difference and the dimension-reduced equipment parameter distance in different time periods to obtain a fusion distance in different time periods; determining a local density based on the fusion distance, and sorting the local density from high to low as a density equipment parameter sequence; determining a metric distance corresponding to the plurality of dimension-reduced equipment parameter distances according to the plurality of dimension-reduced equipment parameter distances and the density equipment parameter sequence; clustering the plurality of dimension-reduced equipment parameter distances according to the metric distance to obtain a plurality of equipment state associated clusters; determining an equipment about to fail cluster and an equipment normal cluster in the plurality of equipment state associated clusters; obtaining real-time equipment parameters, and adjusting the real-time equipment parameters to the equipment normal cluster when the real-time equipment parameters belong to the equipment about to fail cluster, and issuing a prompt information that the equipment parameters are being adjusted. In the embodiment of the present application, the core features of the historical equipment parameters are retained, the influence of irrelevant or repetitive parameters is eliminated, and the equipment state associated clusters obtained by fusing time and parameter features are more accurate, which can effectively distinguish between the normal state and the state about to fail, adjust the parameters to the normal range before the equipment fails, avoid the influence of equipment failure on the metal indium purification process, ensure the stable operation of the purification process, and improve the purification efficiency of indium.

[0025] In other scenarios, the metal indium purification control method based on historical data analysis provided by the embodiment of the present application can also be applied to indium purification optimization based on historical data, which can improve the optimization effect of indium purification.

[0026] The metal indium purification control method based on historical data analysis provided by the embodiment of the present application will be described in detail below with specific examples.

[0027] Figure 1 The flowchart of the first metal indium purification control method based on historical data analysis provided by the embodiment of the present application is shown in Figure 1As shown, the embodiment of the present application provides a metal indium purification control method based on historical data analysis, the method S110 to S140, the following S110 to S140 are specifically explained.

[0028] S110, obtain a plurality of historical equipment parameters of metal indium purification in different time periods. Wherein, different time periods correspond to different total equipment running time, and the historical equipment parameters include electrolyte temperature, electrolyte concentration, cell voltage, current density, electrolytic equipment temperature, electrolytic cell vibration parameter, stirring parameter, indium deposition rate and indium deposition purity.

[0029] Figure 2 The working flow chart of the first metal indium purification control method based on historical data analysis provided by the embodiment of the present application is shown in Figure 2 As shown in the present implementation, the temperature sensor, concentration sensor, voltage and current meter, vibration sensor, speed sensor and other monitoring devices on the metal indium purification equipment can be used to continuously collect and store a plurality of historical equipment parameters in different time periods, which cover the whole link information of process conditions, equipment state and product indicators in the purification process, providing basic data support for subsequent equipment state analysis.

[0030] It should be noted that different time periods correspond to different total equipment running time, for example, time period 1 is the first 100 hours after the initial start of the equipment, corresponding to the total equipment running time of 100 hours; Time period 2 is 101-200 hours, corresponding to 200 hours; Time period 3 is 201-300 hours, corresponding to 300 hours, and the total running time gradually accumulates with the advancement of time period, reflecting the process from running-in to stable of the equipment.

[0031] In the present implementation, the historical equipment parameters specifically include electrolyte temperature, electrolyte concentration, cell voltage, current density, electrolytic equipment temperature, electrolytic cell vibration parameter, stirring parameter, indium deposition rate and indium deposition purity.

[0032] Exemplarily, the electrolyte temperature is generally controlled at 25-30°C (to ensure the migration rate of indium ions); the electrolyte concentration is 150-200 g / L of indium sulfate solution (to maintain the ionic strength of the electrolysis reaction); the cell voltage is stabilized at 2.5-3.0 V (to avoid the occurrence of side reactions); the current density is 150-200 A / m² (to affect the deposition rate of indium); the temperature of the electrolysis equipment (such as the temperature of the electrolysis tank shell) is maintained at 30-35°C (to prevent overheating of the equipment); the vibration parameters of the electrolysis tank are vibration frequency 0.1-0.5 Hz and vibration amplitude 0.1-0.3 mm (to reflect the stability of the tank structure); the stirring parameters are stirring rate 100-200 rpm (to ensure the uniformity of the electrolyte concentration); the deposition rate of indium is 0.5-1.0 g / min (to reflect the efficiency of electrolysis); and the purity of the deposited indium is 99.95%-99.99% (to directly reflect the product quality).

[0033] S120, principal component analysis is performed on the plurality of historical equipment parameters in different time periods to determine target equipment parameter principal components with a cumulative variance contribution rate greater than or equal to a preset cumulative variance contribution rate. A reduced dimension equipment parameter vector corresponding to the target equipment parameter principal components in different time periods is constructed. A reference reduced dimension equipment parameter vector corresponding to the reference total equipment running time and the normal equipment state is obtained. A difference between the total equipment running time and the reference total equipment running time is determined as a total equipment running time difference. A distance between the reference reduced dimension equipment parameter vector and the reduced dimension equipment parameter vector is determined as a reduced dimension equipment parameter distance.

[0034] In the present implementation, principal component analysis (PCA) can be performed on the plurality of historical equipment parameters in different time periods to calculate the covariance matrix of the 9 original parameters, solve the eigenvalues and eigenvectors, and obtain the variance contribution rate of each principal component. Further, principal components with a cumulative variance contribution rate greater than or equal to a preset cumulative variance contribution rate (such as 90%) can be selected as target equipment parameter principal components.

[0035] For example, after collecting 1000 groups of historical data, the first principal component (covering the associated information of electrolyte temperature, current density, and indium deposition rate) has a variance contribution rate of 45%, the second principal component (covering the associated information of cell voltage and stirring rate) has a variance contribution rate of 30%, and the third principal component (covering the associated information of electrolysis equipment temperature and vibration parameters) has a variance contribution rate of 15%. The cumulative contribution rate of the three is 90%, so the first three principal components can be determined as target principal components, the core features of the original parameters are retained, and redundant information is eliminated.

[0036] In this implementation, based on the scores of the principal components of the target device parameters, dimensionality-reduced device parameter vectors corresponding to different time periods can be constructed. For example, the scores of the three principal components in time period 1 are 0.8, 0.6, and 0.4, corresponding to the dimensionality-reduced vector (0.8, 0.6, 0.4); the scores in time period 2 are 0.7, 0.5, and 0.3, corresponding to the vector (0.7, 0.5, 0.3); and the scores in time period 3 are 0.6, 0.4, and 0.2, corresponding to the vector (0.6, 0.4, 0.2). The dimensionality-reduced vectors simplify high-dimensional parameters into points in low-dimensional space, facilitating subsequent distance calculation and clustering.

[0037] In this implementation, the baseline total equipment runtime can be selected as the typical duration of the equipment in a stable operating phase (e.g., 200 hours, at which point the equipment has completed its break-in and parameter fluctuations are small). The baseline dimensionality-reduced equipment parameter vector is the dimensionality-reduced vector under normal equipment conditions. For example, when the equipment has run for 200 hours, the average scores of the three principal components are 0.7, 0.5, and 0.3, corresponding to the baseline vector (0.7, 0.5, 0.3), which serves as a standard reference point for judging the equipment status.

[0038] In this implementation, the total equipment runtime difference value can be calculated: that is, the difference between the total equipment runtime in different time periods and the baseline total runtime. For example, the total runtime in time period 1 is 100 hours, and the difference value is 100-200=-100 hours; the total runtime in time period 3 is 300 hours, and the difference value is 300-200=100 hours. This value reflects the degree of deviation between the equipment runtime and the "stable phase".

[0039] In this implementation, the distance between the reduced-dimensional device parameters can be calculated, which is the spatial distance (such as Euclidean distance) between the reduced-dimensional device parameter vector and the baseline reduced-dimensional vector at different time periods. For example, the Euclidean distance between the reduced-dimensional vector (0.8, 0.6, 0.4) and the baseline vector (0.7, 0.5, 0.3) in time period 1 is 0.173, which reflects the difference between the parameter state and the normal state.

[0040] S130. Weighted fusion of the total equipment runtime difference and the dimensionality-reduced equipment parameter distance across different time periods is used as the fusion distance for each time period. Local density is determined based on the fusion distance, and the local densities are sorted from high to low as the density equipment parameter sequence. Based on the multiple dimensionality-reduced equipment parameter distances and the density equipment parameter sequence, the corresponding metric distances for the multiple dimensionality-reduced equipment parameter distances are determined. The multiple dimensionality-reduced equipment parameter distances are clustered according to the metric distance to obtain multiple equipment state association clusters. The clusters of devices about to fail and those in normal operation are then identified within these multiple equipment state association clusters.

[0041] In this implementation, the total equipment runtime difference and the distance of the reduced equipment parameters can be weighted and fused to obtain the fused distance. For example, if the weight of the time dimension is set to 0.3 (reflecting the cumulative effect of equipment aging over time) and the weight of the parameter dimension is set to 0.7 (reflecting the real-time deviation of the current state), then the fused distance for time period 1 is |100-200|×0.3 +0.173×0.7≈30.121; the fused distance for time period 2 is |200-200|×0.3 + 0×0.7=0; and the fused distance for time period 3 is |300-200|×0.3 + 0.173×0.7≈30.121. The fused distance integrates the dual-dimensional information of time aging and state deviation, and more comprehensively reflects the equipment status.

[0042] In this implementation, local density can be calculated based on the fusion distance. By counting the number of neighbors within a set distance threshold (e.g., 0.2) for each time period (i.e., other time periods with similar fusion distances), the more neighbors there are, the higher the local density. For example, the fusion distance of time period 2 is 0, and there are multiple neighbors within the threshold, resulting in the highest local density. The fusion distance of time periods 1 and 3 is approximately 30.121, with fewer neighbors and lower local density. Furthermore, the local density can be sorted from high to low to obtain a density device parameter sequence (e.g., time period 2, time period 1, time period 3), highlighting the density of the state distribution and improving clustering accuracy.

[0043] In this implementation, the metric distance is calculated by combining the distance of the dimensionality reduction device parameters and the sequence of density device parameters. For example, if the density of the previous point in the density sequence is higher than that of the next point, the metric distance is the difference between the current dimensionality reduction distance and the previous dimensionality reduction distance (or the distance after density weighting). For example, the metric distance of time period 2 is 0.0 (the highest density, which is used as the benchmark), the metric distance of time period 1 is 0.1 (the difference between the dimensionality reduction distance and the previous point), and the metric distance of time period 3 is 0.2 (the difference between the dimensionality reduction distance and the previous point). The metric distance of points with high density (normal state) is smaller.

[0044] In this implementation, the distance between the reduced device parameters can be clustered according to the metric distance. Time periods with similar metric distances are grouped into the same cluster to obtain the device state association cluster. For example, the metric distance < 0.1 is cluster 1 (including time period 2, corresponding to the stable operation state of the device), the metric distance 0.1 ≤ metric distance < 0.2 is cluster 2 (including time period 1, corresponding to the slight deviation in the break-in stage of the device), and the metric distance ≥ 0.2 is cluster 3 (including time period 3, corresponding to the significant deviation in the aging stage of the device).

[0045] In the present implementation, according to the deviation degree of the parameters in the cluster and the reference state, the device impending failure cluster and the device normal cluster can be determined. For example, the dimension reduction distance of cluster 1 is the smallest, the fusion distance is 0, and the cluster 1 completely matches the reference normal state, so the cluster 1 is the device normal cluster; the dimension reduction distances of cluster 2 and cluster 3 are greater than or equal to 0.173, the fusion distances are greater than or equal to 30.121, and the total running time deviates from the reference, so the cluster 2 and the cluster 3 are the device impending failure clusters.

[0046] In S140, real-time device parameters are acquired, and when it is determined that the real-time device parameters belong to the device impending failure cluster, the real-time device parameters are adjusted to the device normal cluster, and prompt information that the device parameters are being adjusted is sent.

[0047] In the present implementation, the current device parameters (for example, electrolyte temperature 28℃, concentration 185g / L, cell voltage 2.7V, current density 190A / m², and nine parameters) can be collected in real time by sensors on the device, and the current device parameters are converted into a dimension reduction device parameter vector (for example, the real-time vector is (0.6, 0.4, 0.2)).

[0048] In the present implementation, the real-time dimension reduction vector and the device state associated cluster are matched. If the real-time vector belongs to the device impending failure cluster (for example, the matching cluster 3), parameter adjustment is started, for example, the current density is reduced from 190A / m² to 180A / m² (the load of the electrolytic cell is reduced), and the stirring rate is reduced from 200rpm to 150rpm (the vibration of the cell body is reduced), so that the real-time vector returns to the range of the device normal cluster (for example, (0.7, 0.5, 0.3)).

[0049] In the present implementation, prompt information is sent at the same time as the parameter adjustment. The prompt information is popped up on the display screen in the control room in the form of text or is sent in the form of a beeping prompt by a sound and light alarm. The prompt information allows the operator to know the adjustment state in a timely manner, avoids device failure or product quality decline caused by abnormal parameters, and improves the controllability of the purification process.

[0050] Through principal component analysis for dimension reduction, the core features of the historical parameters are retained, and redundant data interference is avoided. Through fusion of the two-dimensional information of time aging and state deviation, the accuracy of device state clustering is improved. Through matching of real-time parameters and clustering clusters, rapid identification of abnormal states and parameter adjustment are realized, the stable operation of the metal indium purification process is ensured, the influence of device failure on production is effectively avoided, and the consistency of product quality is improved.

[0051] Figure 3 A flowchart of a second metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in FIG. 2. Figure 3As shown, in some implementations, in S130, the determination of the plurality of dimension-reduced equipment parameter distances corresponding to the plurality of dimension-reduced equipment parameter distances and the density equipment parameter sequence includes S131-S132, which are described below.

[0052] S131, according to the density equipment parameter sequence, determine the density penalty factor corresponding to the plurality of historical equipment parameters. Wherein, the greater the density of the equipment parameter in the density equipment parameter sequence, the smaller the density penalty factor.

[0053] Figure 4 The working flow diagram of the second metal indium purification control method based on historical data analysis provided by the embodiment of the application is as shown in Figure 4 As shown, in the present implementation, the density penalty factor corresponding to each historical equipment parameter can be assigned according to the density equipment parameter sequence, and the density equipment parameter sequence is a sequence sorted from high to low according to local density, wherein the greater the density of the equipment parameter, the more intensive the state distribution of the parameter, and the smaller the corresponding density penalty factor is set, so as to reduce the weight influence of the intensive state on the subsequent distance calculation.

[0054] It should be noted that the size of the density penalty factor is directly related to the density sorting of the density equipment parameter sequence, and the parameter with higher density corresponds to a smaller penalty factor, which can reduce the interference of distance data of common normal state on the clustering result.

[0055] For example, it is assumed that the top 30% of the parameters in the density equipment parameter sequence have the highest density, corresponding to a density penalty factor of 0.2; the middle 40% of the parameters have medium density, corresponding to a density penalty factor of 0.5; and the last 30% of the parameters have the lowest density, corresponding to a density penalty factor of 1.0, so that the greater the density of the parameter, the smaller the density penalty factor.

[0056] S132, determine the product of the dimension-reduced equipment parameter distance and the density penalty factor corresponding to the plurality of historical equipment parameters as the metric distance corresponding to the plurality of dimension-reduced equipment parameter distances.

[0057] In the present implementation, the dimension-reduced equipment parameter distance of each historical equipment parameter can be multiplied by the density penalty factor corresponding to the parameter to obtain the metric distance corresponding to the dimension-reduced equipment parameter distance, and this calculation method can adjust the distance weight of different density regions, so that the distance difference in the low-density region is more prominent, facilitating subsequent differentiation of equipment state.

[0058] It should be noted that the dimension-reduced equipment parameter distance is the distance between the reference dimension-reduced equipment parameter vector and the dimension-reduced vector of the historical parameter, and after being multiplied by the density penalty factor, the dimension-reduced distance in the high-density region will be reduced in weight, and the dimension-reduced distance in the low-density region will be relatively enlarged in weight.

[0059] Exemplarily, the dimension-reduced equipment parameter distance of a certain historical equipment parameter is 6, and the corresponding density penalty factor is 0.3, and the product of the two is 1.8, that is, the metric distance is 1.8; the dimension-reduced equipment parameter distance of another historical equipment parameter is also 6, but the density penalty factor is 0.7, and the product is 4.2, and the metric distance of the latter is larger, which reflects that the distance contribution of the low-density area is more significant.

[0060] By the present implementation, the weight of the dimension-reduced equipment parameter distance is adjusted by the density penalty factor. The high-density equipment parameter area usually corresponds to the normal or common state, and the distance weight is reduced to avoid the distance data of the high-density area from excessively affecting the subsequent clustering. The obtained metric distance is more consistent with the actual distribution characteristics of the equipment parameter, reduces the interference of the high-density area on the clustering result, improves the accuracy of the subsequent device state associated cluster division, improves the accuracy of the real-time device parameter state judgment, reduces the situation of misadjustment or missed adjustment, makes the parameter adjustment more accurate and effective, ensures the normal operation of the indium purification equipment, and reduces the risk of interruption of the purification process caused by equipment failure.

[0061] By the present implementation, the metric distance corresponding to different equipment states is more significantly different. In clustering, the low-density equipment that is about to fail can be more clearly distinguished from the high-density equipment that is normal, so that the differentiation degree of the equipment state associated cluster is improved, and the high-density data of the normal cluster is prevented from masking the low-density data that is about to fail, so that the cluster division is more accurate.

[0062] In some implementations, the first target principal component includes an electrolytic equipment temperature and an electrolytic cell vibration parameter. The second target principal component includes an electrolyte concentration and a current density.

[0063] In the present implementation, the first target principal component is composed of an electrolytic equipment temperature and an electrolytic cell vibration parameter. Both parameters are directly related to the physical operation state of the equipment and are key physical signs for capturing potential equipment failure. The electrolytic equipment temperature reflects the temperature level of the structure of the electrolytic equipment (such as the electrolytic cell body and heating elements) during operation. Abnormal fluctuations in the temperature may indicate a failure of the heat dissipation or temperature control system of the equipment. The electrolytic cell vibration parameter reflects the vibration characteristics (such as amplitude and frequency) of the electrolytic cell during operation and can reflect the stability of the mechanical structure.

[0064] Exemplarily, the monitoring point of the electrolytic equipment temperature of a certain indium purification equipment is arranged on the side wall of the electrolytic cell. The temperature is maintained at 60-80°C during normal operation. If the temperature continuously exceeds 85°C, it usually indicates that the temperature control module of the heating device is malfunctioning, and the equipment needs to be shut down for maintenance. The electrolytic cell vibration parameter is collected by a bottom vibration sensor. Under normal working conditions, the vibration amplitude does not exceed 0.5 mm / s. If the amplitude suddenly rises above 1 mm / s, it is highly likely that the fixing bolts of the electrolytic cell are loose or the internal electrode frame is deformed, which is a warning signal that the equipment is about to fail.

[0065] In the present embodiment, the second target principal component includes electrolyte concentration and current density, which are core control indicators of the electrolytic purification process and cooperatively affect the chemical equilibrium of the electrolysis reaction and the indium deposition effect. The electrolyte concentration refers to the mass concentration of the effective component (e.g., indium sulfate) of the reaction in the electrolyte, which determines the content of indium ions in the electrolyte. The current density refers to the current intensity passing through the electrode surface per unit area, which directly controls the rate and intensity of the electrolysis reaction.

[0066] For example, in the metal indium purification process in the sulfuric acid system, the electrolyte concentration is usually controlled at 150-200 g / L (calculated based on indium sulfate). If the concentration is reduced to below 120 g / L, the migration rate of indium ions on the cathode will slow down, and the indium deposition rate will decrease from 0.8 kg / h to below 0.5 kg / h. The current density is calculated based on the effective area of the cathode plate, and the normal range is 200-300 A / m². If the current density exceeds 350 A / m², a large amount of hydrogen gas will be generated on the cathode surface, resulting in a loose indium deposition layer, and the purity will decrease from 99.95% to below 99.8%, which cannot meet the product quality requirements.

[0067] Through the present embodiment, the targeted principal component selection can more accurately capture the physical signs of the impending failure of the equipment, improve the accuracy of the identification of the impending failure cluster of the equipment, and reduce the false judgment of the failure caused by the omission of the equipment state parameters. The present embodiment can more accurately reflect the cooperative abnormality of the process parameters, improve the accuracy of the adjustment of the process parameters, and reduce the purity fluctuation of the indium purification caused by the imbalance of the process parameters.

[0068] Through the present embodiment, the interpretability and pertinence of the reduced equipment parameter vector are improved, the weighted fusion of the subsequent total equipment running time difference value and the reduced equipment parameter distance is more accurate, and thus the discrimination accuracy of the impending failure cluster of the equipment and the normal cluster of the equipment is improved, and the error of the equipment state judgment is reduced.

[0069] Figure 5 A flowchart of a third metal indium purification control method based on historical data analysis provided by the present embodiment is shown in Figure 5 As shown in some implementations, the above method further includes S210 to S240, which are described below.

[0070] S210, a plurality of historical equipment parameters of metal indium purification in different time periods are obtained. Different time periods correspond to different total equipment running time, and the historical equipment parameters include electrolyte temperature, electrolyte concentration, cell voltage, current density, stirring parameters, deposition rate, and deposition purity.

[0071] Figure 6 A working flowchart of the third metal indium purification control method based on historical data analysis provided by the present embodiment is shown inFigure 6 As shown, in the present embodiment, the temperature sensor, the concentration sensor, the voltage and current acquisition module, and the stirring controller deployed on the metal indium purification equipment can continuously collect historical data of the equipment at different stages such as starting, stable operation, and high-load operation, so as to ensure that the historical equipment parameters can cover the running state of the equipment in the whole life cycle and provide comprehensive basic data for subsequent analysis.

[0072] In the present embodiment, different time periods are divided according to the cumulative running time of the equipment, for example, 100 hours, 200 hours, and 300 hours of equipment operation correspond to different time periods.

[0073] In the present embodiment, in the historical equipment parameters, the electrolyte temperature is the real-time monitoring temperature of the electrolyte in the electrolytic cell, the electrolyte concentration is the mass concentration of indium ions in the electrolyte, the cell voltage is the voltage value between the anode and the cathode of the electrolytic cell, the current density is the current intensity per unit area of the electrolytic plate, the stirring parameter is the rotating speed of the stirrer, the deposition rate is the deposition thickness per unit time of the indium metal on the cathode surface, and the deposition purity is the mass proportion of impurity elements in the deposited indium.

[0074] S220, performing principal component analysis dimension reduction on the plurality of historical equipment parameters in different time periods to determine target principal components with a cumulative variance contribution rate greater than or equal to a preset contribution rate. A reduced dimension equipment parameter vector corresponding to the target principal components in different time periods is constructed. A reference reduced dimension equipment parameter vector corresponding to the reference total equipment running time and the normal equipment state is obtained. The difference between the total equipment running time and the reference total equipment running time is determined as a total equipment running time difference value. The distance between the reference reduced dimension equipment parameter vector and the reduced dimension equipment parameter vector is determined as a reduced dimension equipment parameter distance.

[0075] In the present embodiment, the historical equipment parameters can be standardized to eliminate dimensional differences, and then the covariance matrix of the parameters is calculated, the eigenvalues and eigenvectors of the covariance matrix are solved, the eigenvectors with a cumulative variance contribution rate corresponding to the eigenvalues reaching a preset value (such as 85%) are selected as the target principal components, and it is ensured that the target principal components can reflect the core information of the historical data.

[0076] In the present embodiment, the historical equipment parameters of each time period can be projected to the target principal components to obtain the score values of each principal component, and the reduced dimension equipment parameter vector is formed by arranging the principal components in order, for example, the first target principal component score of a certain time period is 0.8 and the second principal component score is 0.6, and the reduced dimension vector is [0.8, 0.6].

[0077] In the present embodiment, the reference total device runtime length is selected as the cumulative length of time when the device is in the best stable state (e.g., 500 hours), and the reference dimensionality reduction device parameter vector is the mean of the dimensionality reduction vectors in multiple time periods when the device is normally running (e.g., the mean of the first principal component is 0.75, the mean of the second principal component is 0.55, and the reference vector is [0.75, 0.55]).

[0078] In the present embodiment, the total device runtime length difference value is obtained by subtracting the reference total device runtime length from the total device runtime length in a certain time period, for example, the total runtime length in a certain time period is 600 hours, and the difference value is 100 hours.

[0079] In the present embodiment, the Euclidean distance is used to calculate the distance between the reference dimensionality reduction vector and the dimensionality reduction vector in each time period, that is, the square root of the sum of the squares of the corresponding dimension difference, for example, the distance between the reference vector [0.75, 0.55] and the dimensionality reduction vector [0.8, 0.6] is about 0.0707.

[0080] S230, the total device runtime length difference value and the dimensionality reduction device parameter distance in different time periods are fused to obtain a fusion distance corresponding to the multiple historical device parameters. The local density of each fusion distance is determined based on the fusion distance, and the local density is sorted from high to low as a density device parameter sequence. The mean of the variance contribution rate of the principal component of the same dimension of the dimensionality reduction device parameter vector is determined. According to the mean of the variance contribution rate of the principal component of the same dimension, a contribution rate weight matrix is constructed. According to the dimensionality reduction device parameter vector and the contribution rate weight matrix, a weighted device parameter vector corresponding to different time periods is determined. The distance between the weighted device parameter vectors corresponding to different time periods is determined as the dimensionality reduction device parameter difference degree corresponding to the multiple historical device parameters. According to the multiple dimensionality reduction device parameter difference degrees and the density device parameter sequence, a metric distance corresponding to the multiple dimensionality reduction device parameter difference degrees is determined. The multiple dimensionality reduction device parameter difference degrees are clustered according to the metric distance to obtain multiple device state associated clusters. And determine the device in the multiple device state associated clusters, that is, the device failure cluster and the device normal cluster.

[0081] In the present embodiment, weights can be assigned according to the influence of device runtime length and parameter state on failure (e.g., time difference value weight 0.3, parameter distance weight 0.7), and the fusion distance is 0.3x time difference value + 0.7x parameter distance, which comprehensively reflects the dual influence of time decay and parameter offset.

[0082] In the present embodiment, the local density is the number of other fusion distances within a certain threshold (e.g., 0.1) around each fusion distance, and the more the number, the higher the density; after calculating the local density of all fusion distances, the density device parameter sequence is sorted from high to low, and the device with high density corresponds to the common state of the device, which is placed at the front end of the sequence.

[0083] In the present implementation, the same dimension principal component refers to the principal component at the same position in the dimension reduction vector (for example, the first and second principal components are included in the dimension reduction vector of all time periods); the variance contribution rate of each principal component in different time periods is calculated, and then the average value is obtained, for example, the variance contribution rates of the first principal component in three time periods are 40%, 42% and 38%, and the average value is 40%; the average value of the second principal component is 35%, and the average value of the variance contribution rate of the same dimension principal component is obtained.

[0084] In the present implementation, the contribution rate weight matrix is a diagonal matrix, the diagonal elements are the average values of the variance contribution rates of the principal components, and the non-diagonal elements are 0, for example, the average value of the first principal component is 40% and the average value of the second principal component is 35%, and the weight matrix is [[0.4, 0], [0, 0.35]], which highlights the influence of the principal component with high variance contribution rate on the equipment state.

[0085] In the present implementation, the dimension reduction equipment parameter vector is multiplied by the contribution rate weight matrix to obtain a weighted equipment parameter vector, for example, after the dimension reduction vector [0.8, 0.6] is multiplied by the weight matrix, the weighted vector is [0.8*0.4, 0.6*0.35] = [0.32, 0.21], and the characteristics of the core principal component are emphasized.

[0086] In the present implementation, the Manhattan distance (corresponding to the sum of absolute values of dimension difference) between the weighted vectors in different time periods is calculated as the dimension reduction equipment parameter difference degree, for example, the weighted vector in time period 1 is [0.32, 0.21] and the weighted vector in time period 2 is [0.35, 0.23], and the difference degree is |0.32-0.35| + |0.21-0.23| = 0.039.

[0087] In the present implementation, the density penalty factor is set according to the density equipment parameter sequence (the higher the density, the smaller the penalty factor, for example, the penalty factor for the first 20% of the density is 0.8); the dimension reduction equipment parameter difference degree is multiplied by the penalty factor to obtain the metric distance, for example, the difference degree is 0.039 and the penalty factor is 0.8, and the metric distance is 0.0312, which weakens the difference degree weight in the high-density area and highlights the abnormal characteristics in the low-density area.

[0088] In the present implementation, the DBSCAN clustering algorithm is used, the neighborhood radius (for example, 0.05) and the minimum number of points (for example, 3) are set, the difference degrees with similar metric distances are clustered into multiple clusters, and each cluster corresponds to a type of equipment state.

[0089] In the present implementation, the center of each cluster (the mean of the intra-cluster difference degree) and the density feature are calculated, and the device is about to fail cluster has a large fusion distance (large time difference and large parameter offset) corresponding to the center of the cluster, and a low local density; the device normal cluster has a small fusion distance corresponding to the center of the cluster, and a high local density; by comparing the cluster features, the cluster with a large fusion distance and a low density is marked as a device about to fail cluster, and the cluster with a small fusion distance and a high density is marked as a device normal cluster.

[0090] In S240, the real-time device parameters are obtained, and when it is determined that the real-time device parameters belong to the device about to fail cluster, the real-time device parameters are adjusted to the device normal cluster, and prompt information that the device parameters are being adjusted is sent.

[0091] In the present implementation, the parameters such as electrolyte temperature, current density, and stirring speed are collected in real time by sensors, and are converted into a reduced dimension device parameter vector according to a principal component analysis method of historical parameters; the distance between the vector and the center of each device state associated cluster is calculated, and if it belongs to the device about to fail cluster, the real-time parameters are adjusted according to the parameter range of the normal cluster (for example, the electrolyte temperature is reduced from 45°C to 40°C, and the current density is adjusted from 300A / m² to 250A / m²); at the same time, prompt information is sent through the device alarm system or the operation interface to ensure that the operator knows the adjustment state in time.

[0092] Through the present implementation, the difference degree is first optimized through weighted processing, and then adjusted in combination with the density penalty, so that the metric distance considers the influence of the principal component weight and reflects the local density difference. Such a metric distance has higher discrimination, can more accurately distinguish different device states, makes the device state associated cluster obtained by clustering more accurate, and thus more reliably identifies the device about to fail cluster and the normal cluster, and improves the accuracy of device state judgment.

[0093] Through the present implementation, different principal components are given corresponding weights through the mean of the same dimension principal component variance contribution rate, so that the weight distribution is consistent with the actual influence degree of the principal component on the device state. After such processing, the calculation of the reduced dimension device parameter difference degree can better reflect the change of the key principal component, and the accuracy of the parameter difference degree is improved, providing a more reliable basis for subsequent clustering.

[0094] Through the present implementation, the difference of the principal component with a high variance contribution rate is highlighted through weighting, so that the parameter difference of different time periods is more consistent with the real change of the device state. Such improved difference degree can effectively avoid the interference of irrelevant or less influential parameter changes, more accurately capture the difference of the device state, and improve the reliability of subsequent analysis.

[0095] Figure 7 A flowchart of a fourth metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in FIG. 4. Figure 7As shown, in some implementations, in S230, the determination of the metric distance corresponding to each of the plurality of dimensionality reduction device parameter differences based on the plurality of dimensionality reduction device parameter difference degrees and the density device parameter sequence includes S231-S232, which are described below.

[0096] S231, determine a density penalty factor corresponding to each of the plurality of historical device parameters based on the density device parameter sequence. The greater the density of the device parameter in the density device parameter sequence, the smaller the density penalty factor.

[0097] Figure 8 The fourth metal indium purification control method based on historical data analysis provided by the embodiment of the present application has a working flowchart as shown in Figure 8 As shown, in the present implementation, the density penalty factor corresponding to each historical device parameter can be determined based on the density device parameter sequence, wherein the density device parameter sequence is a sequence sorted from high to low according to the local density of the historical device parameters, and the local density reflects the degree of aggregation of the same or similar device state parameters.

[0098] In the present implementation, the density penalty factor is set according to the rule that the greater the density, the smaller the factor. This processing can make the historical parameters in the high-density area (usually corresponding to the common state of normal operation of the device) have a lower weight in subsequent calculations, avoiding excessive interference with the identification of abnormal states.

[0099] For example, assume that the local density of the first 30% of the historical device parameters in the density device parameter sequence is the largest, corresponding to the normal state of stable operation of the device under full load, and the density penalty factor thereof can be set to 0.1; the local density of the middle 40% of the parameters is medium, corresponding to the state of the device under light load or small parameter fluctuations, and the penalty factor thereof is set to 0.5; the local density of the last 30% of the parameters is the smallest, corresponding to the abnormal state of sudden load change or slight wear of the device components, and the penalty factor thereof is set to 1.0, clearly reflecting the rule that the greater the density, the smaller the penalty factor.

[0100] S232, determine the product of the dimensionality reduction device parameter difference degree and the density penalty factor corresponding to each of the plurality of historical device parameters as the metric distance corresponding to each of the plurality of dimensionality reduction device parameter differences.

[0101] In the present embodiment, the dimensionality reduction equipment parameter difference degree corresponding to each historical equipment parameter can be multiplied by the density penalty factor to obtain the metric distance corresponding to the parameter. The dimensionality reduction equipment parameter difference degree is the distance between the weighted equipment parameter vectors of different time periods after weighted processing by the contribution rate weight matrix (the contribution rate weight matrix is constructed based on the mean of the variance contribution rate of the same dimension principal component, and the influence degree of the principal components such as electrolytic equipment temperature and electrolytic cell vibration parameters on the equipment state has been considered); the density penalty factor further adjusts the weight of different density regions, and the combination of the two allows the metric distance to retain the key differences of the principal components and distinguish the density characteristics of normal and abnormal states.

[0102] It should be noted that the dimensionality reduction equipment parameter difference degree focuses on the core difference of the principal component (such as the deviation of the electrolyte concentration from the normal range, the fluctuation of the current density, etc.), and the density penalty factor suppresses the non-key difference of the high-density normal region. Such a product calculation can make the parameter difference of the low-density abnormal region more prominent and avoid the dense parameters in the normal region from masking the abnormal points that are about to fail.

[0103] Exemplarily, the dimensionality reduction equipment parameter difference degree of a certain historical equipment parameter is 0.7 (the principal component analysis shows that the electrolyte concentration is slightly lower than the normal threshold, and the current density is slightly higher), and the corresponding density penalty factor is 0.2 (the parameter is located at the edge of the normal region with a larger density), and the metric distance is 0.7*0.2=0.14; the dimensionality reduction equipment parameter difference degree of another historical parameter is 0.5 (the electrolytic equipment temperature is 8°C higher than normal, and the electrolytic cell vibration parameter exceeds the design threshold), and the corresponding density penalty factor is 1.0 (the parameter is located in a region with very small density that is about to fail), and the metric distance is 0.5*1.0=0.5. Through this calculation, the metric distance of the abnormal parameter is significantly greater than that of the parameter in the normal region, and it is easier to be accurately classified into the equipment about to fail cluster in the subsequent clustering.

[0104] Through the present embodiment, the difference degree corresponding to the high-density and more representative historical equipment parameter is adjusted by a smaller penalty factor, the difference degree weight of the normal parameter in the high-density region is reduced, the distinguishing ability of the metric distance to the equipment state is improved, and the misjudgment of the normal parameter as an abnormal parameter is reduced. The obtained equipment state associated cluster is more accurate, the distinguishing precision of the equipment about to fail cluster and the equipment normal cluster is improved, and the state misjudgment is reduced.

[0105] Through the present embodiment, when the real-time parameter belongs to the equipment about to fail cluster, the target of adjusting to the equipment normal cluster is more clear, because the equipment normal cluster is determined based on more accurate clustering results, and is closer to the real normal equipment state. This adjustment improves the pertinence and effectiveness of the real-time equipment parameter adjustment, reduces the metal indium purification quality fluctuation caused by improper parameter adjustment, and ensures the stability of the purification process.

[0106] In some implementations, the first target principal component includes electrolyte concentration and electrolyte temperature. The second target principal component includes cell voltage and current density.

[0107] In this implementation, the electrolyte concentration and electrolyte temperature, which directly reflect the chemical properties of the electrolyte, can be classified as the first target principal component. The electrolyte concentration represents the content of solutes such as indium ions in the electrolyte, determining the solubility and migration ability of indium ions. The electrolyte temperature affects the ion diffusion rate and the thermodynamic equilibrium of the electrolysis reaction, both of which together constitute the core environmental basis of the indium purification chemical reaction. For example, if the electrolyte concentration is higher than 320 g / L, crystallization may occur due to oversaturation of solutes, blocking the ion migration channel. If the electrolyte temperature is lower than 22℃, the ion diffusion rate decreases by about 15%, which will delay the electrolysis reaction process.

[0108] In this implementation, the cell voltage and current density, which directly control the electrochemical reaction, can be classified as the second target principal component. The cell voltage provides the necessary potential difference for the electrolysis reaction, determining the direction and rate of the electrochemical reaction. The current density reflects the current load per unit electrode area, directly affecting the rate and uniformity of indium deposition. Both of them together constitute the key parameters of the indium purification electric driving process. For example, if the cell voltage exceeds 1.9V, the electrolysis of water will be triggered, producing hydrogen gas and consuming electric energy, while reducing the purity of indium deposition. If the current density is higher than 210 A / m², the indium deposition layer will become loose due to the excessive rate, and the impurity embedding amount will increase by about 8%.

[0109] Through this implementation, the reduced equipment parameter vector more accurately represents the real state of the purification, reduces the interference of irrelevant variables, and improves the explanation ability of principal component analysis on the purification process. The reduced parameter vector more comprehensively reflects the state of key indicators, allowing subsequent cluster analysis to more accurately identify the equipment state that affects the purification effect and reduce the misjudgment caused by incomplete coverage of key parameters.

[0110] Figure 9 A flowchart of a fifth metal indium purification control method based on historical data analysis provided by an embodiment of the present application is shown in FIG. 5. As shown in FIG. 5, in some implementations, the method further includes S310 to S330, which are described in detail below. Figure 9

[0111] S310, a plurality of historical recovery parameters of a metal indium recovery raw material pretreatment link are obtained. The historical recovery parameters include leaching agent concentration, leaching agent temperature, indium concentration after purification, and impurity content.

[0112] Figure 10 A workflow diagram of the fifth metal indium purification control method based on historical data analysis provided by an embodiment of the present application is shown in FIG. 5. As shown in FIG. 5, in some implementations, the method further includes S310 to S330, which are described in detail below. Figure 10 ​As shown, in the present embodiment, multiple historical recovery parameters of the indium metal recovery raw material pretreatment link can be obtained, which are directly related to the raw material pretreatment quality, and the historical recovery parameters include leaching agent concentration, leaching agent temperature, indium concentration after purification, and impurity content.

[0113] In the present embodiment, the leaching agent concentration refers to the concentration of the reagent (such as sulfuric acid concentration) for dissolving indium in the recovery raw material, which affects the dissolution rate of indium; the leaching agent temperature affects the activity of the leaching agent and the diffusion rate of indium ions; the indium concentration after purification is the effective concentration of indium in the raw material after pretreatment, which reflects the effect of indium extraction; and the impurity content is the proportion of impurities remaining after pretreatment, which determines the difficulty of subsequent purification.

[0114] For example, a batch of recovery raw material uses sulfuric acid as the leaching agent, and the historical recovery parameters are leaching agent concentration 150 g / L, leaching agent temperature 60°C, indium concentration after purification 20 g / L, and impurity content 0.5%.

[0115] S320, performing principal component analysis dimension reduction on the multiple historical recovery parameters to determine target recovery parameter principal components with a cumulative variance contribution rate greater than or equal to a preset cumulative variance contribution rate. A dimension-reduced recovery parameter vector corresponding to the target recovery parameter principal components in different time periods is constructed. The multiple dimension-reduced recovery parameter vectors are clustered to obtain multiple recovery parameter association clusters. Multiple device state association clusters are obtained. In the multiple recovery parameter association clusters, the best recovery parameter association cluster corresponding to the device state association cluster is determined.

[0116] In the present embodiment, the multiple historical recovery parameters can be subjected to principal component analysis dimension reduction. The historical recovery parameters are first standardized to eliminate dimensional differences, then a parameter covariance matrix is calculated, eigenvalues and eigenvectors are solved, the eigenvalues are sorted from large to small and the cumulative variance contribution rate is calculated, and when the cumulative variance contribution rate reaches a preset value (such as 85%), the corresponding eigenvector is selected as the target recovery parameter principal component.

[0117] For example, after calculating four historical recovery parameters, the cumulative variance contribution rate of the first two principal components reaches 88%, which exceeds the preset 85%, and therefore the first two principal components are determined as the target recovery parameter principal components.

[0118] In the present embodiment, a dimension-reduced recovery parameter vector corresponding to the target recovery parameter principal components in different time periods can be constructed. The historical recovery parameters in each time period are projected onto the target principal components to obtain a dimension-reduced vector for that time period (each element corresponds to a principal component score).

[0119] For example, the parameters of a certain time period are leaching agent concentration 150 g / L, temperature 60°C, indium concentration after purification 20 g / L, and impurity content 0.5%. After projection onto the first two principal components, the dimension-reduced recovery parameter vector is [1.2, 0.8].

[0120] In the present implementation, the multiple dimension reduction recovery parameter vectors can be clustered, similar dimension reduction vectors are divided into the same cluster by using K-means algorithm, and multiple recovery parameter associated clusters are obtained, and each cluster represents a type of pretreatment parameter combination and raw material quality state.

[0121] For example, after K-means clustering of the dimension reduction vectors of 50 time periods, 3 clusters are obtained: cluster 1 corresponds to high leaching agent concentration, high temperature, high indium after purification, and low impurity; cluster 2 is a medium parameter; and cluster 3 is a low parameter and high impurity.

[0122] In the present implementation, multiple device state associated clusters (such as a normal device cluster and an about-to-fail cluster obtained by clustering the purification device parameters) can be obtained, and the best recovery parameter associated cluster corresponding to the device state associated cluster is determined in the recovery parameter associated cluster, that is, the recovery parameter cluster that can make the subsequent purification effect (purity, yield) optimal under a certain device state.

[0123] For example, the best recovery parameter associated cluster corresponding to the normal device cluster is cluster 1, because the raw material pretreatment quality is good and matches the high requirements of normal device operation.

[0124] In S330, the current recovery parameter and the current device parameter are obtained, the current device state associated cluster to which the current device parameter belongs is determined, and the best recovery parameter associated cluster corresponding to the current device state associated cluster is determined. When the current recovery parameter deviates from the best recovery parameter associated cluster, the current recovery parameter is adjusted to the recovery parameter corresponding to the best recovery parameter associated cluster.

[0125] In the present implementation, the current recovery parameter and the current device parameter can be obtained. The current recovery parameter is a real-time parameter in the pretreatment link (such as the current leaching agent concentration 140 g / L, the temperature 55°C, the indium after purification 18 g / L, and the impurity 0.6%); and the current device parameter is a real-time operation parameter of the purification device (such as the electrolyte concentration and the current density).

[0126] In the present implementation, the current device state associated cluster to which the current device parameter belongs can be determined. After dimension reduction of the current device parameter, the distance between the current device parameter and the center of each device state cluster is calculated, and the cluster closest to the center is the current cluster; and then according to the preset corresponding relationship, the best recovery parameter cluster corresponding to the current device state cluster is found.

[0127] For example, after dimension reduction of the current device parameter, the center of the normal device cluster is closest to the center, and the best recovery parameter cluster corresponding to the normal device cluster is cluster 1.

[0128] In the present implementation, when the current recovery parameter deviates from the best recovery parameter associated cluster, the current recovery parameter is adjusted to the parameter corresponding to the best cluster. The deviation is determined by calculating the distance between the dimension reduction vector of the current recovery parameter and the center of the best cluster. If the distance exceeds a threshold value (such as 0.5), it is determined that the deviation exists. When adjusting, the current value is corrected according to the parameter range of the best cluster.

[0129] For example, the current dimension reduction vector and the cluster 1 center distance is 0.7 (over the threshold value 0.5), the leaching agent concentration needs to be increased from 140 g / L to 150 g / L, the temperature needs to be increased from 55°C to 60°C, so that the purified indium reaches 20 g / L, the impurities are reduced to 0.5%, and the parameters of cluster 1 are returned to the parameter range.

[0130] By the present implementation, the raw material pretreatment recovery parameters and the subsequent equipment operation parameters are linked, the whole process from raw material pretreatment to equipment operation is cooperatively controlled, the limitations of single link adjustment are avoided, the stability and efficiency of the whole process of indium purification are improved, and the purity and yield of the final indium product are ensured.

[0131] By the present implementation, by establishing the association between the pretreatment recovery parameters and the equipment state, the recovery parameter adjustment is matched with the equipment state demand, the poor pretreatment effect caused by parameter mismatch is avoided, the cooperativity of the pretreatment link and the subsequent purification equipment is improved, and high-quality raw materials are provided for subsequent purification; the key parameters of the pretreatment link are focused on, the best parameter range is accurately positioned through dimension reduction clustering, random adjustment of the pretreatment parameters is avoided, the indium concentration after purification is effectively improved, the impurity content is reduced, and the quality of the raw material pretreatment is optimized.

[0132] Figure 11 A flowchart of a sixth metal indium purification control method based on historical data analysis provided by the embodiments of the present application is shown in Figure 11 As shown in the above S320, in some implementations, in the multiple recovery parameter associated clusters, the best recovery parameter associated cluster corresponding to the current equipment state associated cluster is determined, including S321 to S322, which are specifically described as follows.

[0133] S321, the center equipment state parameters corresponding to the multiple equipment state associated clusters are determined, and the center recovery parameters corresponding to the multiple recovery parameter associated clusters are determined. The center equipment state parameters corresponding to the multiple equipment state associated clusters and the center recovery parameters corresponding to the multiple recovery parameter associated clusters are associated to a unified feature space, and the center parameter distance between the center equipment state parameters and the center recovery parameters in the unified feature space is determined.

[0134] In the present implementation, statistical analysis can be performed on all the equipment state parameters in each equipment state associated cluster, the mean or median of each dimension parameter in the cluster is calculated, and the center equipment state parameter representing the core feature of the cluster is obtained; at the same time, the same statistical analysis is performed on all the recovery parameters in each recovery parameter associated cluster, the mean or median of each dimension parameter is calculated, and the center recovery parameter representing the core feature of the recovery parameter associated cluster is obtained. The center parameter condenses the overall features of the parameters in the cluster, avoids the interference of single parameter fluctuation on the cluster state, and more accurately reflects the true attributes of the cluster.

[0135] For example, a certain device state associated cluster contains 100 sets of device state parameters (each set contains electrolyte concentration, temperature, cell voltage, current density), and the center device state parameters are obtained by calculating the mean value of each dimension, which are electrolyte concentration 180 g / L, temperature 35℃, cell voltage 3.2 V, and current density 150 A / m². A certain recovery parameter associated cluster contains 80 sets of recovery parameters (each set contains leaching agent concentration, temperature, indium concentration after purification, and impurity content), and the center recovery parameters are obtained by calculating the mean value of each dimension, which are leaching agent concentration 195 g / L, temperature 58℃, indium concentration after purification 5.2 g / L, and impurity content 0.09%.

[0136] In the present implementation, the center device state parameters and the center recovery parameters can be converted to a unified feature space by a standardization method. The dimensions of the two types of parameters are first standardized (such as mapping the parameter values to the 0-1 interval) to eliminate the differences in dimensions and numerical ranges, and then the standardized parameters are mapped to the same dimensional feature space.

[0137] For example, the center device state parameters after standardization are [0.6, 0.5, 0.7, 0.4], and the center recovery parameters after standardization are [0.55, 0.6, 0.7, 0.3], and the Euclidean distance is calculated to be 0.15, which is the center parameter distance between the two.

[0138] S322, in the unified feature space, the current center device state parameters corresponding to the current device state associated cluster are determined, and all center parameter distances corresponding to the current center device state parameters are obtained. Among all the center parameter distances, the minimum center parameter distance is determined, and the recovery parameter associated cluster corresponding to the minimum center parameter distance is determined as the best recovery parameter associated cluster.

[0139] In the present implementation, the center parameter distance between the center device state parameters and the center recovery parameters is calculated in the unified feature space by methods such as Euclidean distance or cosine distance, which directly reflects the adaptation degree of the two.

[0140] In the present implementation, after determining the current device state associated cluster to which the current device parameters belong, the center device state parameters of the cluster are directly called as the current center device state parameters. In the unified feature space, the center parameter distances between the current center device state parameters and the center recovery parameters of all recovery parameter associated clusters are determined, forming a set of distance values, which reflect the adaptation of the current device state and each recovery parameter associated cluster.

[0141] For example, the current device state associated cluster is the device normal cluster, the center device state parameter after standardization is [0.6, 0.5, 0.7, 0.4], and the center recovery parameter of the three recovery parameter associated clusters after standardization is [0.55, 0.6, 0.7, 0.3], [0.7, 0.4, 0.6, 0.5], and [0.5, 0.7, 0.8, 0.2] respectively. The distance values are 0.15, 0.2, and 0.25 respectively.

[0142] In the present implementation, all center parameter distances corresponding to the current center device state parameter are compared, and the recovery parameter associated cluster corresponding to the smallest distance value is determined as the best recovery parameter associated cluster, because the core features of the recovery parameter associated cluster and the current device state are the most matched and the adaptation degree is the highest. Therefore, the recovery parameter associated cluster is determined as the best recovery parameter associated cluster.

[0143] For example, the distance values corresponding to the current center device state parameter are 0.15, 0.2, and 0.25, and the recovery parameter associated cluster corresponding to the smallest distance 0.15 is the best, which can provide the most adaptive recovery parameter range for the current device state.

[0144] Through the present implementation, the determination method of the center parameter distance based on the unified feature space provides a precise basis for the linkage adjustment of the device state and the recovery parameter, avoids the blindness of the recovery parameter adjustment, improves the synergistic effect of the indium recovery raw material pretreatment link and the purification link, and guarantees the stability of the purification process and the indium recovery efficiency.

[0145] In some implementations, when the current recovery parameter deviates from the best recovery parameter associated cluster in S330, the current recovery parameter is adjusted to the recovery parameter corresponding to the best recovery parameter associated cluster, including S331 to S332, which will be described below.

[0146] S331, determine the center recovery parameter of the best recovery parameter associated cluster, and determine the distance between the center recovery parameter and the current recovery parameter as the recovery parameter distance.

[0147] In the present implementation, when the current recovery parameter deviating from the best recovery parameter associated cluster needs to be adjusted, the core feature of the best recovery parameter associated cluster is first determined, which is the center recovery parameter. It is the statistical core value of all historical recovery parameters in the cluster, which can represent the optimal state of the recovery parameter of the cluster. Then, the center recovery parameter and the current recovery parameter are mapped to the unified feature space, and the quantitative difference between the two is calculated, which is the recovery parameter distance.

[0148] Exemplarily, if the center recovery parameter of the optimal recovery parameter associated cluster is a leaching agent concentration of 200 g / L, a leaching agent temperature of 60°C, an indium concentration after purification of 50 g / L, and an impurity content of 0.1%, and the current recovery parameter is a leaching agent concentration of 180 g / L, a leaching agent temperature of 55°C, an indium concentration after purification of 45 g / L, and an impurity content of 0.3%, the Euclidean distance of the two in the unified feature space is calculated, and the recovery parameter distance is 7.

[0149] In S332, when the recovery parameter distance is greater than or equal to the preset recovery parameter distance, it is determined that the current recovery parameter deviates from the optimal recovery parameter associated cluster, and the current recovery parameter is adjusted to the recovery parameter corresponding to the optimal recovery parameter associated cluster.

[0150] In the present embodiment, the preset recovery parameter distance can be set as a quantitative standard for determining deviation. When the recovery parameter distance reaches or exceeds the threshold, it means that the difference between the current recovery parameter and the optimal state is too large. At this time, the current recovery parameter is adjusted to the parameter corresponding to the optimal recovery parameter associated cluster, and specifically, the center recovery parameter can be taken as the target to ensure that the adjusted parameter fits the optimal state.

[0151] Exemplarily, if the preset recovery parameter distance is 5 and the current recovery parameter distance is 7 (exceeding the threshold), it is determined that there is deviation. At this time, the leaching agent concentration is increased from 180 g / L to 200 g / L, the leaching agent temperature is heated from 55°C to 60°C, the indium concentration after purification is increased to 50 g / L by optimizing the purification process, and the impurity content is reduced to 0.1%, so that the current recovery parameter matches the center parameter of the optimal recovery parameter associated cluster.

[0152] By the present embodiment, the accuracy of the recovery parameter deviation determination is improved, the misjudgment or omission caused by subjective judgment is avoided, it is ensured that only the recovery parameter that deviates from the optimal state is adjusted, unnecessary parameter changes are reduced, the target of the recovery parameter adjustment is more clear, the adjusted recovery parameter is closer to the optimal state, and the processing effect of the indium recovery raw material pretreatment link is effectively improved.

[0153] The present application also provides a metal indium purification control system based on historical data analysis.

[0154] Figure 12 A logic structure diagram of a metal indium purification control system based on historical data analysis provided by the present application is shown in Figure 12As shown, the system 1 of the embodiment includes a processing unit 11, a storage unit 12 and a transceiver unit 13, the processing unit 11 is configured to process data, the storage unit 12 is configured to store data, and the transceiver unit 13 is configured to transceive data, and the processing unit 11, the storage unit 12 and the transceiver unit 13 cooperate to implement the method described above. The beneficial effects of the embodiments of the present application have been described in the above method, which will not be repeated here.

[0155] It should be noted that the information interaction, execution process and the like between the above-mentioned devices / units are based on the same concept as the method embodiments of the present application, and the specific functions and the technical effects brought by them can be referred to the method embodiments part, which will not be repeated here.

[0156] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit, and the integrated unit can be realized in the form of hardware or software. In addition, the specific name of each functional unit and module is only for easy distinction, and does not limit the protection scope of the present application. The specific working process of the unit and module in the above system can refer to the corresponding process in the above method embodiments, which will not be repeated here.

[0157] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the present application can implement all or part of the processes in the above-mentioned embodiment methods through a computer program to instruct relevant hardware to complete, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms. The computer readable medium at least includes any entity or device capable of carrying the computer program code to the photographing device / terminal equipment, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium. For example, U disk, mobile hard disk, magnetic disk or optical disk, etc. In some jurisdictions, according to legislation and patent practice, the computer readable medium can not be an electrical carrier signal and a telecommunication signal.

[0158] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0159] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0160] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division, and actual implementation can have another division manner. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0161] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may also be distributed to multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0162] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for controlling the purification of metallic indium based on historical data analysis, characterized in that, The method includes: Multiple historical equipment parameters for indium metal purification were obtained at different time periods. The different time periods corresponded to different total equipment runtimes. The historical equipment parameters included electrolyte temperature, electrolyte concentration, cell voltage, current density, electrolysis equipment temperature, electrolysis cell vibration parameters, stirring parameters, indium deposition rate, and indium deposition purity. Principal component analysis (PCA) is performed on multiple historical equipment parameters across different time periods to reduce dimensionality. The PCA identifies principal components of target equipment parameters whose cumulative variance contribution rate is greater than or equal to a preset cumulative variance contribution rate. Dimensionally reduced equipment parameter vectors corresponding to the principal components of the target equipment parameters across different time periods are constructed. Benchmark dimensionally reduced equipment parameter vectors corresponding to the baseline total equipment runtime and normal equipment state are obtained. The difference between the total equipment runtime and the benchmark total equipment runtime is determined as the total equipment runtime difference value. The distance between the benchmark dimensionally reduced equipment parameter vector and the dimensionally reduced equipment parameter vector is determined as the dimensionality reduced equipment parameter distance. The total equipment runtime difference and the distance of the reduced-dimensional equipment parameters within different time periods are weighted and fused to form the fusion distance for different time periods. Local density is determined based on the fusion distance and sorted from high to low local density to form the density equipment parameter sequence. The metric distances corresponding to the multiple reduced-dimensional equipment parameter distances are determined based on the multiple reduced-dimensional equipment parameter distances and the density equipment parameter sequence. The multiple reduced-dimensional equipment parameter distances are clustered according to the metric distances to obtain multiple equipment state association clusters. Finally, the clusters of equipment about to fail and the clusters of equipment in normal condition are identified within the multiple equipment state association clusters. If the real-time device parameters are obtained and it is determined that the real-time device parameters belong to the cluster of devices that are about to fail, the real-time device parameters are adjusted to the normal device cluster, and a prompt message indicating that the device parameters are being adjusted is issued.

2. The method according to claim 1, characterized in that, Based on multiple distances between dimensionality-reduced device parameters and density device parameter sequences, determine the metric distances corresponding to the distances between multiple dimensionality-reduced device parameters, including: Based on the density equipment parameter sequence, determine the density penalty factor corresponding to multiple historical equipment parameters; where the larger the density of the equipment parameter in the density equipment parameter sequence, the smaller the density penalty factor. The product of the distance between the dimensionality-reduced device parameters and the density penalty factor corresponding to multiple historical device parameters is determined as the metric distance corresponding to the distance between the multiple dimensionality-reduced device parameters.

3. The method according to claim 2, characterized in that, The first objective principal component includes the temperature of the electrolysis equipment and the vibration parameters of the electrolysis cell; the second objective principal component includes the electrolyte concentration and the current density.

4. The method according to claim 3, characterized in that, The method further includes: Multiple historical equipment parameters for indium metal purification were obtained at different time periods. The different time periods corresponded to different total equipment runtimes. The historical equipment parameters included electrolyte temperature, electrolyte concentration, cell voltage, current density, stirring parameters, deposition rate, and deposition purity. Principal component analysis (PCA) is used to reduce the dimensionality of multiple historical equipment parameters within different time periods. Target principal components with a cumulative variance contribution rate greater than or equal to a preset contribution rate are identified. Dimensionally reduced equipment parameter vectors corresponding to the target principal components within different time periods are constructed. Benchmark dimensionally reduced equipment parameter vectors corresponding to the benchmark total equipment runtime and normal equipment state are obtained. The difference between the total equipment runtime and the benchmark total equipment runtime is determined as the total equipment runtime difference value. The distance between the benchmark dimensionally reduced equipment parameter vector and the dimensionally reduced equipment parameter vector is determined as the dimensionality reduced equipment parameter distance. The method involves weighted fusion of the total equipment runtime differences across different time periods and the distances between dimensionality-reduced equipment parameters, resulting in a fusion distance corresponding to multiple historical equipment parameters. Based on this fusion distance, the local density of each fusion distance is determined and sorted from high to low to form a density equipment parameter sequence. The mean of the variance contribution rate of the principal components in the same dimension of the dimensionality-reduced equipment parameter vector is determined. A contribution rate weight matrix is ​​constructed based on the mean of the variance contribution rate of the principal components in the same dimension. Based on the dimensionality-reduced equipment parameter vectors and the contribution rate weight matrix, weighted equipment parameter vectors corresponding to different time periods are determined. The distances between the weighted equipment parameter vectors corresponding to different time periods are determined as the dimensionality-reduced equipment parameter differences corresponding to multiple historical equipment parameters. Based on the dimensionality-reduced equipment parameter differences and the density equipment parameter sequence, a metric distance corresponding to the dimensionality-reduced equipment parameter differences is determined. The dimensionality-reduced equipment parameter differences are clustered according to the metric distance to obtain multiple equipment state association clusters. Finally, clusters of equipment about to fail and equipment in normal condition are identified within these multiple equipment state association clusters. If the real-time device parameters are obtained and it is determined that the real-time device parameters belong to the cluster of devices that are about to fail, the real-time device parameters are adjusted to the normal device cluster, and a prompt message indicating that the device parameters are being adjusted is issued.

5. The method according to claim 4, characterized in that, Based on the parameter differences of multiple dimensionality reduction devices and the sequence of density device parameters, determine the metric distance corresponding to the parameter differences of multiple dimensionality reduction devices, including: Based on the density equipment parameter sequence, determine the density penalty factor corresponding to multiple historical equipment parameters; where the larger the density of the equipment parameter in the density equipment parameter sequence, the smaller the density penalty factor. The product of the dimensionality difference of the reduced-dimensional device parameters and the density penalty factor corresponding to multiple historical device parameters is determined as the metric distance corresponding to the dimensionality difference of multiple reduced-dimensional device parameters.

6. The method according to claim 5, characterized in that, The first objective principal component includes electrolyte concentration and electrolyte temperature; the second objective principal component includes cell voltage and current density.

7. The method according to claim 6, characterized in that, The method further includes: Multiple historical recovery parameters were obtained from the pretreatment stage of indium metal recovery raw materials; these historical recovery parameters included leaching agent concentration, leaching agent temperature, indium concentration after purification, and impurity content. Principal component analysis is performed on multiple historical recovery parameters to reduce dimensionality and identify the principal components of the target recovery parameters whose cumulative variance contribution rate is greater than or equal to the preset cumulative variance contribution rate. Dimensionally reduced recovery parameter vectors corresponding to the principal components of the target recovery parameters in different time periods are constructed. Multiple dimensionally reduced recovery parameter vectors are clustered to obtain multiple recovery parameter association clusters. Multiple equipment status association clusters are obtained. Among the multiple recovery parameter association clusters, the optimal recovery parameter association cluster corresponding to the equipment status association cluster is determined. Obtain the current recovery parameters and current device parameters; determine the current device status association cluster to which the current device parameters belong, and determine the optimal recovery parameter association cluster corresponding to the current device status association cluster; when the current recovery parameters deviate from the optimal recovery parameter association cluster, adjust the current recovery parameters to the recovery parameters corresponding to the optimal recovery parameter association cluster.

8. The method according to claim 7, characterized in that, Among multiple recovery parameter association clusters, determine the optimal recovery parameter association cluster corresponding to the current equipment status association cluster, including: Determine the central device status parameters corresponding to multiple device status association clusters, and determine the central recovery parameters corresponding to multiple recovery parameter association clusters; associate the central device status parameters corresponding to multiple device status association clusters and the central recovery parameters corresponding to multiple recovery parameter association clusters to a unified feature space, and determine the central parameter distance between the central device status parameters and the central recovery parameters within the unified feature space; In the unified feature space, determine the current central device state parameters corresponding to the current device state association cluster, and obtain all central parameter distances corresponding to the current central device state parameters; among all central parameter distances, determine the minimum central parameter distance, and determine the recovery parameter association cluster corresponding to the minimum central parameter distance, as the optimal recovery parameter association cluster.

9. The method according to claim 8, characterized in that, When the current recovery parameters deviate from the optimal recovery parameter association cluster, adjust the current recovery parameters to the recovery parameters corresponding to the optimal recovery parameter association cluster, including: Determine the central recovery parameter of the optimal recovery parameter associated cluster, and determine the distance between the central recovery parameter and the current recovery parameter, as the recovery parameter distance; When the distance of the recovery parameter is greater than or equal to the preset distance of the recovery parameter, it is determined that the current recovery parameter deviates from the optimal recovery parameter association cluster, and the current recovery parameter is adjusted to the recovery parameter corresponding to the optimal recovery parameter association cluster.

10. A control system for the purification of metallic indium based on historical data analysis, characterized in that, Includes units for implementing the method of any one of claims 1 to 9.

Images