Rare earth metal electrolysis system and electrolysis furnace thereof

Abstract

The invention discloses a rare earth metal electrolysis system and an electrolysis furnace thereof, and relates to the technical field of metallurgy, the rare earth metal electrolysis system is characterized by comprising a potential data acquisition and preprocessing module, a dendritic crystal feature extraction and quantification module, a dendritic crystal identification and intelligent determination module and an inverse pulse control and dynamic regulation module, in the electrolysis process, potential distribution original data of the surface of a rare earth metal cathode are detected and collected in real time, the obtained original data are preprocessed, and a preprocessed potential data set is established based on a time sequence. According to the method, intelligent, real-time and accurate inhibition of dendritic crystals in the rare earth metal electrolysis process is achieved, short circuit, splashing and electrode ablation are effectively prevented, the compactness and purity of metal deposition are improved, the current efficiency is remarkably improved, and compared with a traditional scheme, the method has the advantages of being high in real-time performance, good in inhibition effect, small in interference and high in adaptability, and the method is suitable for popularization and application. And efficient and safe preparation of the rare earth metal is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of metallurgy, in particular to a rare earth metal electrolysis system and an electrolytic furnace thereof. BACKGROUND

[0002] The rare earth metal electrolysis system refers to a complete technical system and process flow for extracting rare earth metal elements from rare earth compounds (such as rare earth chlorides, rare earth fluorides, etc.) by electrolysis. Rare earth metals are usually difficult to be directly prepared by conventional metallurgical means (such as carbon thermal reduction), so electrolytic technologies such as molten salt electrolysis or electroslag electrolysis are generally used. The system mainly includes an electrolytic cell, an electrolytic electrode, a molten salt system (as an electrolyte), a power supply device, and auxiliary devices such as temperature control and atmosphere control. In the electrolysis process, rare earth ions are reduced to rare earth metals by obtaining electrons at the cathode, while the anode usually releases chlorine gas or other gas byproducts. The core of the rare earth metal electrolysis system is the efficient reduction and deposition of rare earth ions in a high-temperature molten salt environment, which requires good electrochemical stability, electrical conductivity, and high-purity control capability, and is widely used in metallurgy, material preparation, and high-end electronic, magnetic, optical, and other rare earth application fields.

[0003] In actual electrolytic production, rare earth metals usually need to be reduced at high current density, mainly to meet the needs of high yield, rapid nucleation, and grain control. Specifically, when the electrolytic production faces the process goals of improving the metal production per unit time, shortening the electrolysis cycle, or forming rare earth metals with specific morphology and organization structure, high current density operation is often required. High current density can significantly accelerate the electrochemical reduction rate on the cathode surface, promote rapid nucleation and deposition of rare earth ions, reduce the long residence time of rare earth in high-temperature molten salt, and reduce the risk of secondary pollution or oxidation. At the same time, in the preparation of some special rare earth metals or rare earth alloys, high current density can obtain fine-grained, dense, or specially oriented deposited metal layers to meet the requirements of subsequent precision machining or functional material preparation. However, high current density operation is also accompanied by problems such as insufficient ion migration rate and high overpotential, which can easily induce dendrite growth and abnormal cathode deposition, so in industrial practice, a balance and optimization between high current density and dendrite suppression measures are often needed.

[0004] The prior art has the following disadvantages: in the rare earth metal electrolysis process, especially under high current density operating conditions, due to the significant enhancement of the local electric field intensity on the cathode surface, the electrochemical reaction rate is nonlinearly improved, and the reduction and deposition process of rare earth ions on the cathode interface tends to be unstable, and the metal ions tend to grow preferentially in the electric field concentration area. Such abnormal deposition behavior often promotes the directional acceleration of metal crystals along a certain direction, forming needle-like, dendritic or burr-like metal dendritic structures. Once the dendrites are formed, not only will they further enhance the local electric field, causing the dendrites to continue to expand, but they may also pierce the electrolytic separation structure between the anode and the cathode, inducing serious accidents such as electrolytic cell short circuit, molten salt splashing or electrode ablation, while seriously affecting the density and purity of the rare earth metal deposition layer, resulting in a significant reduction in current efficiency.

[0005] The above information disclosed in the background section is only intended to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0006] The purpose of the present application is to provide a rare earth metal electrolysis system and its electrolytic furnace, which realizes intelligent and real-time precise inhibition of dendrite growth in the rare earth metal electrolysis process, effectively prevents safety risks such as short circuit, molten salt splashing and electrode ablation caused by dendrites, and significantly improves the density, uniformity and purity of metal deposition. Through dynamic monitoring of potential distribution characteristics and intelligent recognition of machine learning, dendrite growth can be actively and directionally controlled, overcoming the shortcomings of existing experience-dependent or fixed pulse strategies, with the advantages of strong real-time performance, good inhibition effect, small process interference, strong self-adaptation ability, etc. It provides a reliable guarantee for the efficient and safe preparation of rare earth metals to solve the problems in the background technology.

[0007] In order to achieve the above purpose, the present application provides the following technical scheme: a rare earth metal electrolysis system, comprising a potential data acquisition and preprocessing module, a dendrite feature extraction and quantification module, a dendrite recognition and intelligent judgment module, and an inverse pulse control and dynamic adjustment module: The potential data acquisition and preprocessing module detects and acquires the potential distribution raw data of the rare earth metal cathode surface in real time during the electrolysis process, pre-processes the acquired raw data, and establishes a pre-processed potential data set based on time series; The dendrite feature extraction and quantification module extracts key indicators reflecting the formation of dendrites on the cathode surface from the data set through feature engineering technology, and quantifies the dendrite formation process on the cathode surface through comprehensive analysis of the extracted key indicators; The dendrite recognition and intelligent judgment module takes the dendrite growth quantification index as the input feature and inputs it into the pre-trained machine learning model, and intelligently recognizes the dendrite formation in the current electrolysis process through the machine learning model; The reverse pulse control and dynamic adjustment module, when the evaluation result shows that dendrite growth occurs on the cathode surface, based on real-time growth feedback, applies a short reverse pulse current, and dynamically adjusts the frequency, intensity and duty cycle of the pulse according to the dendrite growth state, so that the rare earth metal crystal nucleus in the local protruding area is re-oxidized and dissolved into the molten salt system, and the metal deposition is locally redistributed.

[0008] Preferably, the potential distribution raw data of the rare earth metal cathode surface is detected and collected in real time, the obtained raw data is preprocessed, and the specific steps of establishing the preprocessed potential data set based on time sequence are as follows: Firstly, the potential distribution raw data of each measuring point on the cathode surface is detected and collected in real time by the high-temperature corrosion-resistant potential sensor arranged in the electrolytic cell, and the spatial distribution and time-varying original signal sequence of the potential are obtained; Secondly, the collected raw potential data is preprocessed to ensure the stability and accuracy of the data; Finally, the preprocessed potential data set is constructed in the format of time sequence based on the time continuity and spatial distribution characteristics of the potential data.

[0009] Preferably, the key indicators reflecting the dendrite formation on the cathode surface are extracted from the data set by feature engineering technology, the extracted indicators include the oscillation degree of the cathode surface potential in a very short time and the change of the singular spectrum characteristic width of the real-time potential distribution data, the oscillation degree of the cathode surface potential in a very short time and the change of the singular spectrum characteristic width of the real-time potential distribution data are analyzed in the detection window to generate the potential gradient transient singular change factor and the dynamic potential singular spectrum width factor respectively, and the dendrite formation process on the cathode surface is quantified by the potential gradient transient singular change factor and the dynamic potential singular spectrum width factor.

[0010] Preferably, the oscillation degree of the cathode surface potential in a very short time is analyzed in the detection window to generate the potential gradient transient singular change factor, and the specific steps are as follows: Firstly, the potential distribution raw data collected in real time on the cathode surface is discretized to construct a two-dimensional potential distribution matrix , wherein and respectively represent the spatial coordinate points of the cathode surface in the horizontal and vertical directions, is the instantaneous potential value at the corresponding position, based on the potential matrix, for each sampling position a local neighborhood is defined around the center , the maximum and minimum values of the potential in the neighborhood are extracted, and the local potential range is calculated, and the calculation expression is as follows: , wherein is the local potential range, is the the maximum value of the measured potential value among all covered points, is the minimum value of the measured potential value among all covered points, is a local neighborhood region; After the extraction of the local range is completed, by quantifying the abnormality of the global range distribution, the potential gradient transient anomaly factor is defined, and the calculation expression is as follows: , wherein is the potential gradient transient anomaly factor, is the median of all in the detection window, is the set of all sampling points in the detection window, is the abnormal amplification index.

[0011] Preferably, the specific steps of analyzing the change of the singular spectrum characteristic width of the real-time potential distribution data under the detection window to generate the dynamic potential singular spectrum width factor are as follows: Select the actual potential distribution signal in a detection window as , wherein represents the coordinate of the cathode surface along the spatial position, and for , the scale segmentation method is used to divide the cathode surface into multiple segmentation intervals, and the local fluctuation energy density in each interval region is calculated, and the calculation expression is as follows: , wherein is the local fluctuation energy of the th interval, is the th sub-interval, is the potential gradient enhancement factor; A local singular index is constructed using the normalized energy density, and the calculation expression is as follows: , wherein is the local singular index of the th interval, is the spatial scale of the th sub-interval; After obtaining the local singular index of all intervals , draw the singular spectrum, and calculate the dynamic potential singular spectrum width factor, and the calculation expression is as follows: , wherein is the maximum local singular index appearing in the singular spectrum, is the minimum local singular index appearing in the singular spectrum, is the dynamic potential singular spectrum width factor.

[0012] Preferably, the dendrite growth quantification indicators, the potential gradient transient anomaly factor and the dynamic potential singularity spectrum width factor, are taken as input features and input into the pre-trained machine learning model, and a dendrite generation risk coefficient is generated by the machine learning model, and the dendrite formation in the current electrolysis process is intelligently identified by the dendrite generation risk coefficient.

[0013] Preferably, the dendrite generation risk coefficient generated when the dendrite formation in the current electrolysis process is intelligently identified by the pre-trained machine learning model is compared and analyzed with the pre-set dendrite generation risk coefficient reference threshold value, and it is judged whether there is dendrite formation in the current electrolysis process, and the division steps are as follows: If the dendrite generation risk coefficient is greater than the pre-set dendrite generation risk coefficient reference threshold value, it is judged that there is dendrite formation in the current electrolysis process; if the dendrite generation risk coefficient is less than or equal to the pre-set dendrite generation risk coefficient reference threshold value, it is judged that there is no dendrite formation in the current electrolysis process.

[0014] Preferably, when the evaluation result shows that dendrite growth appears on the cathode surface, a short-time reverse pulse current is applied, and the specific steps of dynamically adjusting the frequency, intensity and duty cycle of the pulse according to the dendrite growth state are as follows: When the evaluation result shows that dendrite growth appears on the cathode surface, immediately start the reverse pulse control, and dynamically calculate the initial reverse pulse parameters based on the deviation of the dendrite generation risk coefficient from the dendrite generation risk coefficient reference threshold value, and the calculation expression is as follows: , wherein, is the reverse pulse current intensity, is the basic reverse pulse current intensity, is the dendrite generation risk coefficient, is the dendrite generation risk coefficient reference threshold value, is the amplitude modulation coefficient, is the nonlinear gain index; After the initial pulse amplitude is calculated, the duty cycle and the reverse pulse frequency of the reverse pulse are further adaptively determined according to the risk degree, so that the pulse control takes into account the suppression intensity and the electrolysis stability, and the calculation expression is as follows: , wherein, is the reverse pulse duty cycle, is the basic duty cycle, is the duty cycle dynamic adjustment coefficient, is the reverse pulse frequency, is the basic reverse pulse frequency, is the frequency dynamic adjustment coefficient; Based on the calculated pulse parameters, the pulse effect is applied to the cathode surface, the dendritic crystal nucleus of the local protrusion is removed, and the dynamic metal ion redistribution is used to correct the deposition layer morphology, and the calculation expression is as follows: , wherein, is the inverse pulse energy input factor, is the inverse pulse current intensity, is the local metal ion concentration redistribution increment.

[0015] In the above technical solution, the technical effects and advantages provided by the present application are as follows: The present application realizes intelligent, real-time and accurate inhibition of dendrite formation in the rare earth metal electrolysis process, effectively avoids the safety risks such as short circuit, molten salt splashing and electrode ablation caused by abnormal growth of dendrites, and improves the density and uniformity of the metal deposition layer, significantly improves the purity and current efficiency of the rare earth metal product. Compared with the existing traditional scheme which relies on experience or simple pulse control, the present application can actively, directionally and adjustably control the dendrite growth process based on the dynamic monitoring of the cathode surface potential distribution characteristics and the intelligent recognition of machine learning, has the beneficial effects of strong real-time, good inhibition effect, small interference to the electrolysis process and strong self-adaptability, and provides reliable technical support for efficient, safe and high-quality preparation of rare earth metals. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments or prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0017] Figure 1 The present application provides a rare earth metal electrolysis system and a method flowchart of the electrolytic furnace thereof. DETAILED DESCRIPTION

[0018] Example implementations will now be described with reference to the drawings; however, it should be understood that the example implementations can be implemented in many different forms and should not be construed as being limited to the examples set forth herein; rather, these example implementations are provided so that the disclosure will be more thorough and complete, and to fully convey the concept of the example implementations to those skilled in the art.

[0019] The present application provides a rare earth metal electrolysis system as shown in Figure 1 , comprising a potential data acquisition and preprocessing module, a dendrite feature extraction and quantification module, a dendrite recognition and intelligent judgment module, and an inverse pulse control and dynamic adjustment module: The potential data acquisition and preprocessing module detects and acquires the potential distribution raw data of the rare earth metal cathode surface in real time during the electrolysis process, pre-processes the acquired raw data, and establishes a pre-processed potential data set based on time series; Since the growth of dendrites is often accompanied by abnormal enhancement of local electric field, high-temperature potential sensing array or techniques such as electrochemical impedance spectroscopy (EIS), cathode surface micro-area potential scanning can be used to realize high-precision and real-time monitoring of the cathode surface potential distribution. The core role of this step is to capture the small changes of the electrode surface space potential, provide the "precursor signal" of dendrite formation, and provide basic data support for subsequent data processing and dendrite recognition.

[0020] Considering the possible noise interference (such as temperature fluctuation, power ripple, bubble disturbance, etc.) in the electrolysis environment, signal processing algorithms (wavelet denoising, adaptive filtering, local outlier detection) are used to improve the accuracy and effectiveness of the data. The goal of preprocessing is to ensure that the subsequent analysis is based on a high-quality potential distribution data set, avoid false positives or false negatives due to noise or false signals, and ensure the reliability of dendrite detection.

[0021] The specific steps of real-time detection and acquisition of potential distribution raw data of the rare earth metal cathode surface, preprocessing of the acquired raw data, and establishment of a pre-processed potential data set based on time series are as follows: First, through the high-temperature corrosion-resistant potential sensor arranged in the electrolytic cell, the potential distribution raw data of each measuring point on the cathode surface is detected and continuously collected in real time, and the spatial distribution and time-varying original signal sequence of the potential are obtained. Second, the collected raw potential data is pre-processed, mainly including signal denoising (filtering), outlier rejection (rejecting signal burrs caused by bubbles, splashes, etc.), signal smoothing (eliminating high-frequency noise), data normalization (unifying dimension), etc. to ensure the stability and accuracy of the data. Finally, based on the time continuity and spatial distribution characteristics of the potential data, a pre-processed potential data set is constructed in the format of time series, i.e. the potential changes of each measuring point on the cathode surface are organized in time sequence to form a multi-dimensional data set that meets the requirements of subsequent feature extraction and modeling analysis, serving as a basic data source for subsequent dendrite feature extraction and growth determination.

[0022] The dendrite feature extraction and quantification module extracts key indicators reflecting the formation of dendrites on the cathode surface from the data set through feature engineering technology, and comprehensively analyzes the extracted key indicators to quantify the dendrite formation process on the cathode surface. The key indicators reflecting the dendrite formation on the cathode surface are extracted from the data set by feature engineering technology, including the extent of oscillation of the cathode surface potential in an extremely short time and the width change of the singular spectrum characteristics (such as multifractal spectrum) of real-time potential distribution data. The extent of oscillation of the cathode surface potential in an extremely short time and the width change of the singular spectrum characteristics (such as multifractal spectrum) of real-time potential distribution data are comprehensively analyzed under the detection window to generate the potential gradient transient anomaly factor and the dynamic potential singular spectrum width factor, respectively, and the cathode surface dendrite formation process is quantified by the potential gradient transient anomaly factor and the dynamic potential singular spectrum width factor.

[0023] The serious oscillation of the cathode surface potential in an extremely short time can usually be used as one of the important criteria for the rare earth metal cathode to generate dendrites. The essential reason is that the germination and initial growth of dendrites will cause significant disturbance of local potential distribution. After the rare earth ions are preferentially reduced into metal nuclei on the local high electric field area of the cathode surface, the formation of the nuclei will exhibit a sharp local potential mutation in the potential field. Due to the insufficient supply of ions under the condition of high current density, the mutation is prone to show the characteristics of rapid oscillation in a short time with the continuous adsorption and deposition of the nuclei. The potential oscillation reflects the dramatic fluctuation of the local electrochemical reaction rate, which shows the abnormal dynamic instability of the deposition process on the cathode surface, and this is one of the typical electrochemical signal characteristics in the dendrite growth process. In addition, the oscillation process is accompanied by the change of dendrite morphology, local electric field distortion and dynamic fluctuation of interface impedance, which further aggravates the short-time abnormal fluctuation of the cathode potential. Therefore, when the cathode surface potential oscillates severely in an extremely short time, it can usually be used as a direct potential signal characteristic to determine that the dendrites are growing rapidly, which has high diagnostic value.

[0024] The specific steps of analyzing the extent of oscillation of the cathode surface potential in an extremely short time under the detection window to generate the potential gradient transient anomaly factor are as follows: Firstly, the real-time collected potential distribution data of the cathode surface are discretized to construct a two-dimensional potential distribution matrix , wherein and represent the spatial coordinate points of the cathode surface in the horizontal and vertical directions, respectively, is the instantaneous potential value at the corresponding position, based on the potential matrix, for each sampling position a local neighborhood is defined as the center (such as a 3×3 or 5×5 point array as a neighborhood), the maximum and minimum values of the potential in the neighborhood are extracted, and the local potential range is calculated, and the calculation expression is as follows: , wherein is the local potential range, which represents the center The intensity of the potential distribution fluctuation in the neighborhood of the center, reflecting the intensity of the potential in the local region, is the maximum value of the measured potential value among all the points covered, is the minimum value of the measured potential value among all the points covered, is the local neighborhood region, indicating the local neighborhood set containing several surrounding points centered on the point, The purpose of this step is to directly measure the degree of sharp fluctuation of the potential of the local region of the bright electrode surface. The greater the potential difference, the more intense the potential mutation in the neighborhood, which is often closely related to the abnormal deposition of local metal ions and the phenomenon of dendrite germination. By calculating the of all positions through global scanning, , the basic atlas of the cathode surface potential gradient distribution can be obtained.

[0025] After the extraction of the local difference is completed, the abnormality of the global difference distribution is quantified, and the potential gradient transient anomaly factor is defined, and the calculation expression is as follows: , wherein, is the potential gradient transient anomaly factor, is the median of all in the detection window, is the set of all sampling points in the detection window, is an abnormal amplification index, which is a real number greater than 1, and the commonly selected value is 1.5-2.

[0026] The core of this step is to comprehensively measure the deviation of each local difference value in the global potential distribution from the global median. The greater the deviation, the higher the factor value, indicating that there is a sharp and concentrated potential anomaly on the cathode surface, i.e. it is likely that rapid growth of dendrites is occurring. By amplifying the contribution of the local high difference region in the form of an index, the dendrite growth signal can be more prominent in the overall features, facilitating subsequent dendrite determination and inverse pulse control.

[0027] ​​​The potential gradient transient abnormality factor is generated by analyzing the extent of oscillation of the cathode surface potential in an extremely short time within the detection window. The generation process of dendrites is often accompanied by distortion of local electric field and violent fluctuation of potential. When dendrite germination occurs on the surface of the rare earth metal cathode, the local potential distribution will change dramatically in a very short time due to the rapid growth of metal nuclei, which is manifested as potential mutation, gradient increase, and even repeated oscillation, directly leading to a significant increase in the value of the potential gradient transient abnormality factor. Therefore, within the detection window, the greater the potential gradient transient abnormality factor, the more likely dendrite germination or rapid growth occurs on the cathode surface; otherwise, if the factor remains at a low level, it indicates that the potential distribution on the cathode surface is stable, the electrochemical reaction process is stable, and the cathode does not grow dendrites.

[0028] The widening of the singular spectrum characteristics (such as multifractal spectrum) of real-time potential distribution data can be used as an important signal of dendrite generation on the rare earth metal cathode. The reason is that during the growth of dendrites, the cathode surface electric field distribution and potential distribution will change from the original relatively stable, multi-scale self-similar structure to a complex structure dominated by local protrusions, which will significantly enhance the multifractal characteristics of the potential signal and cause the width of the singular spectrum (i.e., the local fractal dimension distribution) to widen. Specifically, the germination and local rapid growth of dendrites will form an abnormal concentration area of potential on the cathode surface, break the original stable potential field, cause the local potential fluctuation in some areas to intensify dramatically, form a multi-scale, strongly nonlinear fluctuation distribution pattern, and further lead to the widening of the singular spectrum width, reflecting the more complex local heterogeneity of the potential distribution. Since the singular spectrum width is highly sensitive to sudden and local structural changes, when the singular spectrum width widens significantly, it can be considered that dendrites have germinated or even entered the rapid growth stage on the cathode surface, which is a highly reliable characterization index.

[0029] The specific steps for generating the dynamic potential singular spectrum width factor by analyzing the width change of the singular spectrum characteristics (such as multifractal spectrum) of real-time potential distribution data within the detection window are as follows: Select the actual potential distribution signal within a detection window as , where represents the coordinate of the cathode surface along the spatial position, and The scale segmentation method is used to divide the cathode surface into multiple segmentation intervals, calculate the local fluctuation energy density in each interval, and the calculation expression is as follows: , where is the local fluctuation energy of the th interval, is the th sub-interval, is the potential gradient enhancement factor, which controls the sensitivity to high-intensity local potential gradient, and is usually taken as ; The local singular exponent is constructed by using the normalized energy density, and the calculation expression is as follows: , wherein is the local singular exponent of the interval, reflecting the roughness of the potential distribution of the interval, is the spatial scale of the sub-interval; By real-time monitoring and spatial segmentation of the cathode surface potential distribution, the spatial variation characteristics of the potential in each local region are accurately captured, reflecting the real dynamics of the micro potential field of the cathode surface. Further, by calculating the local fluctuation energy and the singular exponent, high-quality basic data and characterization capabilities are laid for the subsequent extraction of dendrite growth characteristics and fractal spectrum analysis.

[0030] After obtaining the local singular exponent of all intervals , the singular spectrum is drawn, and the dynamic potential singular spectrum width factor is calculated, and the calculation expression is as follows: , wherein is the maximum local singular exponent appearing in the singular spectrum, is the minimum local singular exponent appearing in the singular spectrum, is the dynamic potential singular spectrum width factor.

[0031] The singular spectrum is a function curve used to describe the multifractal characteristics of signals or fields (such as potential distribution) in space or time, which reflects the complexity and scale dependence of different local regions in the signal. Specifically, the singular spectrum draws the relationship curve of by analyzing the distribution of local singular exponent (i.e. local fractal dimension or roughness) in the entire signal, wherein represents the local singular exponent. The role of drawing the singular spectrum is to intuitively reveal the "roughness" or "mutation" degree of each local region in the potential distribution at different scales, to judge whether the system is in a single, stable fractal state, or whether there are multiple fractal phenomena due to local abnormalities (such as dendrite germination). When there are abnormal peaks or mutations in the local region of the potential distribution (typical characteristics of dendrites), the singular spectrum will be significantly widened or distorted, therefore, the singular spectrum can be used as a sensitive diagnostic tool for cathode dendrite germination and growth in rare earth metal electrolysis process, and early warning and quantitative analysis of dendrite growth can be realized.

[0032] Extracting multi-scale singular features from real-time potential distribution data that can sensitively characterize dendrite germination, reflecting the local unevenness and complexity of the cathode surface potential distribution through singular exponent distribution. By constructing and quantifying the dynamic potential singular spectrum width factor, the real-time, accurate and quantitative description of the dendrite growth process is realized.

[0033] The change of the singular spectrum characteristics (such as multifractal spectrum) width of real-time potential distribution data is analyzed under the detection window to generate the dynamic potential singular spectrum width factor. During the dendrite generation process, the potential distribution on the cathode surface will change from a relatively uniform and stable state to a complex and uneven structure with local high potential protrusions. This process will significantly change the fractal and singular spectrum characteristics of the potential signal, which is manifested as an increase in the singular spectrum width. The greater the potential singular spectrum width factor, the more abnormal fluctuations and local high field regions exist in the potential distribution, which is the typical signal of dendrite germination and growth on the cathode surface. Therefore, when the dynamic potential singular spectrum width factor continuously increases under the monitoring window, it can be considered that dendrite growth is occurring on the cathode surface; on the contrary, when the factor remains at a low level and has no obvious fluctuation, it usually indicates that the potential distribution on the cathode surface is stable and no dendrite formation occurs.

[0034] The dendrite recognition and intelligent judgment module takes the dendrite growth quantitative index as the input feature and inputs it into the pre-trained machine learning model to intelligently identify the dendrite formation in the current electrolysis process through the machine learning model. The dendrite growth quantitative index potential gradient transient singular change factor and the dynamic potential singular spectrum width factor are taken as input features and input into the pre-trained machine learning model to generate a dendrite generation risk coefficient through the machine learning model, and the dendrite formation in the current electrolysis process is intelligently identified through the dendrite generation risk coefficient.

[0035] The pre-trained machine learning model described in the present solution refers to an electrolysis process intelligent discrimination model that is constructed and trained offline based on a large amount of potential distribution data and actual dendrite growth accumulated from historical rare earth metal electrolysis experiments or industrial processes. The establishment process of the model includes: first, a large amount of rare earth metal potential distribution data under different electrolysis conditions and the corresponding actual dendrite growth (which can be verified by subsequent microscopic morphology analysis or online imaging) are collected to form a labeled training data set; then, a variety of key feature indicators (such as the potential gradient transient anomaly factor and the dynamic potential singularity spectrum width factor in the present solution) that can reflect the dendrite growth behavior are extracted from the original potential data through feature engineering technology, and these features are used as model inputs to construct a classification or regression type machine learning model, such as support vector machine (SVM), random forest (RF), gradient boosting tree (GBDT), or convolutional neural network (CNN), etc. Then, the training data is used to train, cross-validate and optimize the hyperparameters of the selected model, and finally a discrimination model that can accurately identify the dendrite germination and growth state is obtained. At this time, the model is called a pre-trained model and can be deployed in a real-time monitoring and control system of the electrolysis process.

[0036] In the actual electrolysis process, the system extracts the feature factors (such as the potential gradient transient anomaly factor and the dynamic potential singularity spectrum width factor) reflecting the dendrite growth from the current potential distribution data as input features, inputs them into the above trained machine learning model, and automatically outputs the dendrite generation risk coefficient, i.e. the risk degree of dendrite germination or abnormal deposition on the current cathode surface. Since the model has learned the nonlinear and complex relationship between the dendrite generation features and the potential feature changes based on a large amount of actual data, compared with traditional rule determination or simple threshold method, it can realize intelligent identification with higher accuracy, stronger robustness and better generalization ability, and accurately judge the dendrite growth state in the current electrolysis process. Based on the risk coefficient, the system can further trigger the subsequent reverse pulse suppression and pulse parameter dynamic regulation to form a real-time and closed-loop dendrite growth control system, significantly improving the stability and product quality of the electrolysis process. In the present solution, the core role of the machine learning model is to accurately complete the dendrite risk prediction and determination using its powerful nonlinear modeling and feature extraction capability, and to provide intelligent decision basis for dynamic regulation.

[0037] The machine learning model is not limited here, and machine learning models that can comprehensively analyze the potential gradient transient anomaly factor and the dynamic potential singularity spectrum width factor to generate a dendrite generation risk coefficient are all acceptable. To achieve the technical solution of the present invention, the present invention provides a specific implementation method: dendrite generation risk coefficient The generation formula is as follows: , where and Potential gradient transient odd change factor and dynamic potential singular spectrum width factor The preset scaling factor of and Both are greater than 0.

[0038] The preset proportional coefficient refers to the dendrite formation risk coefficient ( ) in the calculation formula, the two characteristic factors (i.e., the transient odd change factor of the potential gradient and dynamic potential singular spectrum width factor ) are respectively given weights or importance coefficients in comprehensive analysis, that is, and Since different characteristics have different importance in reflecting the risk of dendrite initiation and growth, it is necessary to determine their importance in advance through artificial or empirical, experimental, and modeling methods. The relative proportion in the calculation is called the preset scale factor.

[0039] In short, the role of the preset proportional coefficient is to control the influence of each characteristic factor on the final risk coefficient. For example, if it is found in the actual rare earth electrolysis process that More sensitive and reliable to dendrite formation, it can be set appropriately , that is, giving Higher weight; on the contrary, if It can better reflect the growth trend of dendrites, so it can be increased This coefficient is usually derived from statistical analysis of historical electrolysis data, expert experience or machine learning training results. It is a key step in achieving quantitative and reasonable synthesis of dendrite risk coefficients, which helps to improve Prediction accuracy and adaptability to dendrite growth status.

[0040] It can be seen from the dendrite formation risk coefficient that the larger the potential gradient transient singular change factor generated after analyzing the degree of oscillation of the cathode surface potential in an extremely short time under the detection window, and the larger the dynamic potential singular spectrum width factor generated after analyzing the width change of the singular spectrum characteristics (such as the multifractal spectrum) of the real-time potential distribution data under the detection window, the greater the dendrite formation risk coefficient formed when the pre-trained machine learning model is used to intelligently identify the dendrite growth in the current electrolysis process, indicating that the probability of dendrite formation on the cathode surface in the current electrolysis process is greater, and vice versa, the smaller the probability of dendrite formation on the cathode surface in the current electrolysis process.

[0041] The dendrite formation risk coefficient generated by intelligently identifying the dendrite formation in the current electrolysis process through the pre-trained machine learning model is compared and analyzed with the pre-set dendrite formation risk coefficient reference threshold to determine whether dendrite formation exists in the current electrolysis process, and the steps are as follows: If the dendrite formation risk coefficient is greater than the pre-set dendrite formation risk coefficient reference threshold, it is determined that dendrite formation exists in the current electrolysis process; if the dendrite formation risk coefficient is less than or equal to the pre-set dendrite formation risk coefficient reference threshold, it is determined that dendrite formation does not exist in the current electrolysis process.

[0042] The reverse pulse control and dynamic adjustment module, when the evaluation result shows that dendrite growth appears on the cathode surface, based on real-time growth feedback, applies a short reverse pulse current, and dynamically adjusts the frequency, intensity and duty cycle of the pulse according to the dendrite growth state, so that the rare earth metal crystal nucleus in the local protruding area is re-oxidized and dissolved into the molten salt system, and the metal deposition is locally redistributed; Active intervention and fine control are performed on the identified cathode surface dendrite growth process to prevent dendrite from continuing to grow, expand and lose control, and ultimately optimize the deposition behavior and suppress dendrites in the rare earth metal electrolysis process. When the system determines that dendrites have germinated or entered the growth stage through the previous potential distribution analysis and machine learning evaluation, a short reverse pulse current is immediately applied based on the real-time monitored dendrite growth state (such as growth speed, range, position), and the rare earth metal crystal nucleus in the local protruding area on the cathode surface loses electrons and undergoes electrochemical re-oxidation through the reverse pulse current, converting into rare earth ions and re-dissolving into the molten salt system, thereby eliminating the formed dendrite structure.

[0043] In addition, in order to make the reverse pulse more efficiently and accurately act on the dendrite, the system will dynamically adjust the frequency, intensity and duty cycle of the reverse pulse according to the specific growth state of the dendrite, to ensure that the reverse pulse can efficiently dissolve the dendrite without damaging the normal deposition process and large-area deposition layer or reducing the metal deposition efficiency. Through dynamic optimization of pulse parameters, local and precise intervention of the micro-deposition behavior on the cathode surface can be achieved, effectively preventing the continuous growth and polarization deterioration of dendrites, and reducing potential risks such as short circuit, molten salt splashing and insufficient density of the deposition layer.

[0044] More importantly, after removing the local dendrites through the short reverse pulse, the metal ions in the local area are redistributed to a more uniform deposition area, forming a stable, flat and dense rare earth metal deposition layer, which helps to improve product purity and mechanical properties, while ensuring the long-term stability and high current efficiency of the electrolysis process. Therefore, this step is not only a simple dendrite removal process, but also a key technical link to realize dendrite identification, control, elimination and deposition reconstruction in the rare earth metal electrolysis process.

[0045] When the evaluation result shows that dendrite growth occurs on the cathode surface, short-time reverse pulse current is applied, and the specific steps of dynamically adjusting the frequency, intensity and duty cycle of the pulse according to the dendrite growth state are as follows: When the evaluation result shows that dendrite growth occurs on the cathode surface, immediately start the reverse pulse control, and dynamically calculate the initial reverse pulse parameters based on the deviation of the dendrite generation risk coefficient from the reference threshold value The calculation expression is as follows: , wherein, is the reverse pulse current intensity, is the basic reverse pulse current intensity, is the dendrite generation risk coefficient, is the dendrite generation risk coefficient reference threshold value, is the amplitude modulation coefficient, which controls the sensitivity of the reverse pulse current intensity to the deviation between and , and is the nonlinear gain index, which makes the reverse pulse intensity exhibit exponential nonlinear enhancement with the increase of the risk coefficient; This step dynamically adjusts the reverse pulse current intensity according to the degree that is greater than the threshold value, and amplifies the pulse intensity in the high-risk case by to ensure that the higher the dendrite risk, the stronger the inhibition.

[0046] After calculating the initial pulse amplitude , further adaptively determine the duty cycle and frequency of the reverse pulse according to the risk degree, so that the pulse control takes into account the inhibition intensity and electrolytic stability, and the calculation expression is as follows: , wherein, is the reverse pulse duty cycle, which represents the proportion of the on-time of the reverse pulse current in the pulse period, and the value range is , is the basic duty cycle, which is the initial duty cycle when the dendrite risk coefficient just reaches the reference threshold value (i.e. ), is the duty cycle dynamic adjustment coefficient, which controls the sensitivity of the duty cycle to the risk deviation , is the reverse pulse frequency, which represents the application frequency of the reverse pulse current, i.e. the action periodicity of the reverse pulse, is the basic reverse pulse frequency, which is the default reverse pulse frequency when , is the frequency dynamic adjustment coefficient, which controls the response intensity of the reverse pulse frequency to the risk deviation; According to the real-time change of the dendrite generation risk coefficient, the duty cycle and the frequency of the reverse pulse are dynamically adjusted, so that the reverse pulse can be adaptively matched with the growth state of the dendrite. Through this step, flexible control of the dendrite growth can be realized, and it is ensured that the reverse pulse can effectively inhibit the dendrite under different risk levels, and avoid interference with the normal metal deposition process.

[0047] Based on the calculated pulse parameters, the pulse is applied to the cathode surface to remove the dendrite nucleus of the local protrusion, and the dynamic metal ion redistribution is used to correct the deposition layer morphology, and the calculation expression is as follows: , wherein, is the reverse pulse energy input factor, which describes the energy input ability of the reverse pulse to the local dendrite growth area of the cathode per unit time, is the reverse pulse current intensity, which refers to the reverse current intensity applied to the cathode surface in the reverse pulse stage, is the local metal ion concentration redistribution increment, which represents the effective redistribution amount of the rare earth metal ion concentration in the local molten salt of the cathode per unit time after the reverse pulse action.

[0048] Through the dynamic reverse pulse control based on the dendrite risk coefficient, the dendrite nucleus that has germinated or preliminarily grown on the cathode surface is effectively removed to prevent its continuous growth. The redistribution of local metal ions is driven by pulse energy input, so that the deposition process tends to be uniform and stable again, and the density and deposition quality of the rare earth metal deposition layer are ensured.

[0049] The present application realizes intelligent, real-time and accurate inhibition of dendrite formation in the electrolysis process of rare earth metals, effectively avoids safety risks such as short circuit, molten salt splashing and electrode ablation caused by abnormal growth of dendrites, and at the same time improves the density and uniformity of the metal deposition layer, significantly improves the purity and current efficiency of rare earth metal products. Compared with the existing traditional scheme which relies on experience or simple pulse control, the present application can actively, directionally and adjustably control the dendrite growth process based on the dynamic monitoring of the potential distribution characteristics of the cathode surface and the intelligent identification of machine learning, has the beneficial effects of strong real-time, good inhibition effect, small interference to the electrolysis process and strong adaptability, and provides reliable technical support for efficient, safe and high-quality preparation of rare earth metals.

[0050] The above formulas are all de-dimensioned to calculate their numerical values, the formula is obtained by software simulation of a large amount of data to reflect the current real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0051] It is apparent that for the person skilled in the art, many modifications and changes can be suggested without departing from the scope of the application, and it is intended to encompass these modifications and changes as fall within the scope of the application.

[0052] It should be noted that, in the present document, relational terms are used solely to estab- lish a correspondence between particular entities and other entities, not necessarily a direct or causal relationship. For example, without necessarily implying a causal relationship, an entity A is said to be used by entity B, if entity B uses entity A. In other words, an entity A is said to be used by entity B, if entity B uses entity A, without necessarily implying a causal relationship between the two entities. Also, the use of the term "comprising" or "comprises" or "comprised of" or "comprised as" or "comprising as" is intended to mean that other elements can also be present in addition to those identified, and not to exclude additional, unrecited elements. Further, the use of the term "including" or "includes" or "included" is intended to mean that other elements can also be present in addition to those identified, and not to exclude additional, unrecited elements. The use of the term "including" or "includes" or "included" is intended to mean that other elements can also be present in addition to those identified, and not to exclude additional, unrecited elements. In other words, the use of the term "including" or "includes" or "included" is intended to mean that other elements can also be present in addition to those identified, and not to exclude additional, unrecited elements.

[0053] It should be understood that the sequence of the processes described above in the various embodiments of the present application does not mean the order of execution, and the execution order of the processes should be determined according to their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0054] Those skilled in the art can clearly understand that the units and algorithm steps of the examples described in combination with the embodiments disclosed in the present document can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art 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.

[0055] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the system, device and unit described above can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0056] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiments according to actual needs.

[0057] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit.

[0058] The above merely describes some exemplary embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0059] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that the described embodiments can be modified in various ways without departing from the spirit and scope of the present application for those skilled in the art. Therefore, the above drawings and descriptions are illustrative in nature and should not be understood as limiting the protection scope of the claims of the present application.

Claims

1. A rare earth metal electrolysis system, characterized in that: It includes potential data acquisition and preprocessing module, dendrite feature extraction and quantification module, dendrite identification and intelligent judgment module, and reverse pulse control and dynamic adjustment module: The potential data acquisition and preprocessing module detects and collects the raw data of the potential distribution on the surface of the rare earth metal cathode in real time during the electrolysis process, preprocesses the acquired raw data, and establishes a preprocessed potential data set based on the time series; The dendrite feature extraction and quantification module uses feature engineering technology to extract key indicators reflecting the formation of dendrites on the cathode surface from the data set, conducts a comprehensive analysis of the extracted key indicators, and quantifies the dendrite formation process on the cathode surface; The dendrite identification and intelligent judgment module uses the dendrite growth quantitative indicators as input features and inputs them into a pre-trained machine learning model. The machine learning model then intelligently identifies dendrite formation during the current electrolysis process. The reverse pulse control and dynamic adjustment module applies a short reverse pulse current based on real-time growth feedback when the evaluation results show that dendrite growth occurs on the cathode surface. At the same time, the frequency, intensity and duty cycle of the pulse are dynamically adjusted according to the dendrite growth state, so that the rare earth metal crystal nuclei in the local protrusion area are reoxidized and dissolved back into the molten salt system, causing local redistribution of metal deposition.

2. The rare earth metal electrolysis system according to claim 1, characterized in that: The specific steps for real-time detection and acquisition of raw data on the potential distribution of the rare earth metal cathode surface, preprocessing the acquired raw data, and establishing a preprocessed potential data set based on the time series are as follows: First, a high-temperature, corrosion-resistant potential sensor placed in the electrolytic cell detects and continuously collects raw data on the potential distribution at each measuring point on the cathode surface in real time, obtaining the spatial distribution of the potential and the original signal sequence that changes over time. Secondly, the collected raw potential data is preprocessed to ensure the stability and accuracy of the data; Finally, based on the temporal continuity and spatial distribution characteristics of the potential data, a preprocessed potential dataset is constructed in the format of a time series.

3. The rare earth metal electrolysis system according to claim 1, characterized in that: Through feature engineering technology, key indicators reflecting the formation of cathode surface dendrites are extracted from the data set. The extracted indicators include the degree of oscillation of the cathode surface potential in a very short time and the change in the width of the singular spectrum characteristics of the real-time potential distribution data. The degree of oscillation of the cathode surface potential in a very short time and the change in the width of the singular spectrum characteristics of the real-time potential distribution data are comprehensively analyzed under the detection window to generate the potential gradient transient singular change factor and the dynamic potential singular spectrum width factor, respectively. The potential gradient transient singular change factor and the dynamic potential singular spectrum width factor are used to quantify the cathode surface dendrite formation process.

4. The rare earth metal electrolysis system according to claim 3, characterized in that: The specific steps for analyzing the oscillation degree of cathode surface potential in a very short time under the detection window to generate the potential gradient transient erratic factor are as follows: First, the potential distribution raw data collected in real time on the cathode surface is discretized to construct a two-dimensional potential distribution matrix ,in and Represent the spatial coordinate points of the cathode surface in the horizontal and vertical directions, is the instantaneous potential value at the corresponding position. Based on the potential matrix, for each sampling position Define a local neighborhood for the center , extract the maximum and minimum values ​​of the potential in the neighborhood, and calculate the local potential range. The calculation expression is as follows: , where is the local potential difference, is The maximum value of the potential measured among all the covered points, is The minimum value of the potential measured among all the covered points, is the local neighborhood area; After completing the extraction of the local range, the abnormality of the global range distribution is quantified to define the transient odd change factor of the potential gradient. The calculation expression is as follows: , where is the transient odd change factor of the potential gradient, All the the median of is the set of all sampling points in the detection window, is the anomaly amplification index.

5. The rare earth metal electrolysis system according to claim 3, characterized in that: The specific steps for analyzing the singular spectrum characteristic width change of real-time potential distribution data in the detection window to generate the dynamic potential singular spectrum width factor are as follows: The actual potential distribution signal within a detection window is selected and calibrated as ,in Represents the coordinates of the cathode surface along the space position, The scale segmentation method is used to divide the cathode surface into multiple segmentation intervals, and the local fluctuation energy density in each interval is calculated. The calculation expression is as follows: , where It is The local fluctuation energy of the interval, It is subintervals, is the potential gradient enhancement factor; The local singular index is constructed using the normalized energy density, and the calculation expression is as follows: , where It is The local singular exponent of the interval, It is The spatial scale of the subintervals; Get the local singular indices of all intervals After that, draw the singular spectrum and calculate the dynamic potential singular spectrum width factor. The calculation expression is as follows: , where is the largest local singular index appearing in the singular spectrum, The smallest local singular index appearing in the singular spectrum, is the dynamic potential singular spectrum width factor.

6. The rare earth metal electrolysis system according to claim 3, characterized in that: The dendrite growth quantitative indicators, the potential gradient transient singular change factor and the dynamic potential singular spectrum width factor, are used as input features and input into a pre-trained machine learning model. The dendrite formation risk coefficient is generated by the machine learning model, and the dendrite formation in the current electrolysis process is intelligently identified by the dendrite formation risk coefficient.

7. The rare earth metal electrolysis system according to claim 6, characterized in that: The dendrite formation risk coefficient generated by intelligently identifying dendrite formation in the current electrolysis process using a pre-trained machine learning model is compared with a pre-set dendrite formation risk coefficient reference threshold to determine whether dendrite formation exists in the current electrolysis process. The steps are as follows: If the dendrite formation risk coefficient is greater than the pre-set dendrite formation risk coefficient reference threshold, it is determined that dendrite formation exists in the current electrolysis process; if the dendrite formation risk coefficient is less than or equal to the pre-set dendrite formation risk coefficient reference threshold, it is determined that no dendrite formation exists in the current electrolysis process.

8. The rare earth metal electrolysis system according to claim 1, characterized in that: When the evaluation results show that dendrite growth occurs on the cathode surface, a short reverse pulse current is applied, and the frequency, intensity, and duty cycle of the pulse are dynamically adjusted according to the dendrite growth state. The specific steps are as follows: When the evaluation results show that dendrite growth occurs on the cathode surface, reverse pulse control is immediately started and a risk factor is generated based on the dendrite growth. The initial reverse pulse parameter is dynamically calculated based on the degree of deviation from the reference threshold of the dendrite formation risk factor. The calculation expression is as follows: , where is the reverse pulse current intensity, is the basic reverse pulse current intensity, is the dendrite formation risk factor, is the reference threshold of the dendrite formation risk coefficient, is the amplitude modulation coefficient, is the nonlinear gain exponent; After calculating the initial pulse amplitude After that, the duty cycle and frequency of the reverse pulse are further determined adaptively according to the risk level, so that the pulse control takes into account both the suppression strength and the electrolysis stability. The calculation expression is as follows; , where is the reverse pulse duty cycle, is the base duty cycle, is the duty cycle dynamic adjustment coefficient, is the reverse pulse frequency, is the basic reverse pulse frequency, is the frequency dynamic adjustment coefficient; Based on the calculated pulse parameters, a pulse is applied to the cathode surface to remove the locally protruding dendrite cores and modify the deposited layer morphology using dynamic metal ion redistribution. The calculation expression is as follows: , where is the reverse pulse energy input factor, is the reverse pulse current intensity, is the local metal ion concentration redistribution increment.

9. A rare earth metal electrolysis furnace comprising the rare earth metal electrolysis system according to any one of claims 1 to 8.

Images