A smart detection platform for the state of electrolyte in electrochemical machining of nickel-cobalt alloys

Abstract

This invention discloses an intelligent monitoring platform for the state of electrolytes in the electrochemical processing of nickel-cobalt alloys. The platform includes a sensing system for real-time online acquisition of parameters such as the physicochemical properties, composition content, and electrochemical performance of the electrolyte; a preprocessing unit for data optimization; a verification unit, a time-series recursion unit, and a multi-source data fusion unit for further processing the preprocessed data; a state recognition model for identifying the normal, critical, and abnormal states of the electrolyte; a trend prediction model for predicting future trends in parameter changes; a fault warning model for early warning and root cause analysis of electrolyte faults; and an interactive platform and control system for data application based on model-generated information. The intelligent monitoring platform constructed by this invention achieves high-precision, comprehensive monitoring of the electrolyte state, state trend prediction, automatic fault diagnosis, and data application, improving the accuracy, efficiency, and intelligence level of nickel-cobalt alloy electrochemical processing.

Description

Technical Field

[0001] This invention belongs to the field of electrolyte condition detection technology, specifically, it relates to an intelligent detection platform for the condition of electrolytes in nickel-cobalt alloy electrochemical processing. Background Technology

[0002] Nickel-cobalt alloys, due to their high hardness, excellent corrosion resistance, good magnetic properties, and high-temperature stability, are widely used in high-end manufacturing fields such as aerospace, new energy batteries, precision instruments, and medical devices. Electrochemical machining, as a core process for the efficient and precise manufacturing of nickel-cobalt alloys, has significant advantages such as high machining accuracy, good surface quality, and the ability to integrate machining and surface modification, and has become the mainstream process route for the mass production of nickel-cobalt alloy components. In the electrochemical machining process of nickel-cobalt alloys, the state of the electrolyte is a core factor determining the machining quality, efficiency, and cost. The electrolyte state directly affects the electrochemical reaction rate, the uniformity of metal deposition, the precision of alloy composition control, and the integrity of the workpiece surface. For example, an imbalance in the nickel / cobalt ion concentration ratio will cause the alloy composition to deviate from design requirements; a pH value exceeding the process window will cause hydroxide precipitation, resulting in machining defects; and fluctuations in conductivity directly affect the current efficiency and forming accuracy of electrochemical machining.

[0003] Currently, electrolyte detection in the field of nickel-cobalt alloy electrochemical machining is mainly divided into two categories: offline detection and online detection. Offline detection adopts a detection mode based on manual sampling and laboratory analysis, which lacks on-site real-time detection capabilities and suffers from serious deficiencies in real-time performance, high dependence on manual labor, and inability to reflect dynamic changes in the electrolyte. Current online detection methods overcome the latency bottleneck of offline detection by deploying key parameter sensors at the processing site to collect key parameter data in real time. However, it still has shortcomings such as incomplete parameter coverage and low level of intelligence. Current online detection methods only complete data collection and simple alarms, lacking data optimization, data integration, trend prediction, and fault diagnosis capabilities. The detection and processing processes are disconnected, resulting in the inability to accurately and effectively monitor the electrolyte state, which in turn restricts the accuracy and efficiency of nickel-cobalt alloy electrochemical machining. Summary of the Invention

[0004] To overcome the aforementioned shortcomings in the prior art, this invention provides an electrolyte state detection platform designed for the electrochemical processing characteristics of nickel-cobalt alloys. This platform integrates multiple parameters, features efficient model function settings, and is intelligently online. It enables real-time and accurate sensing of all electrolyte parameters, efficient data processing, intelligent data analysis, state trend prediction, and automatic fault diagnosis. This provides core data support for closed-loop control of the processing process, thereby solving the problems of detection lag, insufficient accuracy, and low level of intelligence in the prior art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A smart detection platform for the state of electrolytes in the electrochemical processing of nickel-cobalt alloys includes a sensing system for real-time online acquisition of parameters such as the physicochemical properties, composition content, and electrochemical performance of the electrolyte; a preprocessing unit for optimizing the data quality of the data acquired by the sensing system; a verification unit for receiving the data processed by the preprocessing unit and calibrating parameters such as ion concentration, conductivity, and pH value; a state recognition model for identifying and judging the normal state, critical state, and abnormal state of the electrolyte based on the output data of the verification unit; a time-series recursion unit for receiving the output data of the verification unit and performing time-series enhancement on the ion concentration and conductivity parameters; and a future time-series recursion unit for the parameters based on the output data of the time-series recursion unit. The system comprises: a trend prediction model for predicting changing trends; a multi-source data fusion unit that receives output data from the verification unit, the state recognition result of the state recognition model, and the time-series predicted trend information of the trend prediction model, and identifies platform faults; a fault warning model that implements early warning and root cause analysis of electrolyte faults based on the output data of the multi-source data fusion unit; an interactive platform that intuitively displays the state recognition result of the state recognition model, the time-series predicted trend information of the trend prediction model, and the fault information of the fault warning model; and a control system that adjusts the electrolyte state according to the output results of the state recognition model, the trend prediction model, and the fault warning model. The state recognition model includes a feature input layer that receives the output data of the verification unit, a feature enhancement layer that uses a convolutional neural network to capture the coupling relationship between parameters, and a state decision layer that integrates random forest and support vector machine decision-making mechanisms to complete classification decisions. The state decision layer includes a deviation quantification submodel to calculate the deviation of the measured values ​​of each parameter from the normal values. The trend prediction model includes a temporal feature input layer that receives the output data of the temporal recursion unit, a temporal feature extraction layer that uses a long short-term memory network to capture the long-range dependencies of parameters at different processing stages, and a prediction output layer that uses a fully connected layer to complete feature mapping. The fault early warning model includes a graded early warning layer, a root cause location layer, and a handling suggestion layer. The graded early warning layer uses a two-factor coupled early warning function to calculate the comprehensive risk value and classify the fault early warning. The root cause location layer determines the root cause of the fault based on feature weights and parameter deviation through a fault root cause contribution function. The handling suggestion layer has a built-in standardized handling suggestion library and automatically generates suggestions adapted to on-site operations based on the early warning level and the root cause of the fault.

[0006] Furthermore, the verification unit includes a conductivity-ion concentration correlation model to calibrate ion concentration and conductivity, and a pH-complexation reaction equilibrium model to calibrate and calculate pH value.

[0007] Furthermore, the deviation quantification sub-model is characterized using the following function:

[0008] In the formula: Indicates the degree of parameter deviation. This represents the real-time detection value. This represents the optimal value of the parameter. This indicates the upper limit of the normal range of the parameter. This indicates the lower limit of the normal range for the parameter.

[0009] Furthermore, the time-series recursion unit is equipped with an ion concentration time-series recursion model that predicts future ion concentration data based on historical data and a conductivity time-series correction model for eliminating temperature and noise interference to obtain accurate conductivity time-series data.

[0010] Furthermore, a prediction error correction function is constructed in the prediction output layer to correct the original prediction value. The prediction error correction function is as follows:

[0011] In the formula: This is the corrected predicted value. These are the original predicted values ​​from the model. for Historical average error at any given time This is the time decay coefficient.

[0012] Furthermore, the multi-source data fusion unit is equipped with a fault risk prediction model and a composite fault superposition model. The fault risk prediction model constructs a risk prediction function based on the parameter change rate output by the trend prediction model to predict the probability of fault occurrence in advance. The composite fault superposition model constructs a superposition function to quantify the overall fault severity for composite faults with multiple parameters that are abnormal at the same time.

[0013] Furthermore, the two-factor coupled early warning function used in the hierarchical early warning layer is as follows:

[0014] In the formula: This indicates the overall risk value for fault warning. Indicates the trend probability weighting coefficient. Indicates the probability of risk in trend prediction. This represents the weighting coefficient for real-time deviation. Indicates the overall deviation of features; The root cause localization layer includes a fault root cause contribution function to quantify the contribution of each abnormal parameter to the fault. The formula for the fault root cause contribution function is as follows:

[0015] In the formula: Indicates the first The fault contribution of each abnormal parameter. Indicates the first Deviation of each parameter Indicates the overall fault deviation. Indicates the first Each parameter has a weighting coefficient.

[0016] Furthermore, the sensing system includes a physicochemical property sensor group and an electrochemical performance monitoring module for detecting the physicochemical properties and electrochemical performance of the electrolyte in the processing tank, a circulating sampling pump for sampling the electrolyte in the processing tank, a storage container for storing the electrolyte sampled by the circulating sampling pump, and a component content detection module for detecting the component content of the electrolyte in the storage container.

[0017] Furthermore, the preprocessing unit performs the following data preprocessing steps on the data collected by the sensing system: identifying and removing abnormal interference values, duplicate data, and data with incorrect formatting from the original data; eliminating high-frequency noise caused by electromagnetic interference, vibration, and electrolyte bubbles in the industrial environment, while retaining the true trend of data changes; completing data missing due to sensing system failures or communication interruptions using an adaptive feature type completion strategy; and eliminating differences in the dimensions and numerical ranges of different features, enabling weighted calculations of features and adapting the coupled logic of deviation quantification and fault early warning models.

[0018] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention realizes real-time online acquisition of multi-dimensional parameters of electrolyte physicochemical properties, component content and electrochemical performance through the deployed sensing system. The preprocessing unit optimizes the data quality of the acquired data, and the verification unit further calibrates the parameters of ion concentration, conductivity and pH value, thereby providing high-quality data input for the state recognition model. The time-series recursion unit enhances the parameters of ion concentration and conductivity based on the data of the verification unit, and provides corresponding time-series data for the trend prediction model. The multi-source data fusion unit receives the output data of the verification unit, the state recognition result of the state recognition model, and the time-series prediction trend information of the trend prediction model and identifies platform faults. The fault warning model then completes the early warning and root cause analysis of electrolyte faults. The interactive middle platform and operation and maintenance system set up in the present invention respectively provide intuitive display and corresponding control of the output information of the state recognition model, trend prediction model and fault warning model. Compared to the current situation where offline detection lacks real-time performance, preliminary online detection has limited coverage of parameters, and has a low level of intelligence, the intelligent detection platform constructed in this invention achieves high-precision all-round monitoring of electrolyte state, state trend prediction, and automatic fault diagnosis through multi-source data detection, data optimization and integration, and efficient model function setting, providing core data support for closed-loop control of the processing process.

[0019] (2) The verification unit of this invention is equipped with a conductivity-ion concentration correlation model to calibrate ion concentration and conductivity, and a pH value-complexation reaction equilibrium model to calibrate and calculate pH value, which further improves the accuracy of data. The state recognition model is trained and learned on the coupling law of all parameters of nickel-cobalt alloy electrolyte through the setting of feature input layer, feature enhancement layer and state decision layer, so as to realize the classification and recognition of multi-dimensional state conditions of electrolyte.

[0020] (3) The time-series recursion unit of the present invention is equipped with an ion concentration time-series recursion model and a conductivity time-series correction model to enhance the ion concentration and conductivity parameters in time series, providing time series data support for the trend prediction model. The trend prediction model uses a long short-term memory network to accurately capture the dynamic change law of electrolyte parameters with processing time, and can predict the risk of parameter imbalance, impurity accumulation exceeding the standard, and additive failure in advance. It breaks the dilemma of passive reminder only after the parameters exceed the standard in the existing technology, transforms passive response into active prediction, and reserves sufficient time for subsequent regulation.

[0021] (4) The multi-source data fusion unit of this invention integrates the output data of the verification unit, the state recognition result of the state recognition model, and the time-series prediction trend information of the trend prediction model to identify platform faults. The fault warning model has the ability to provide early warning and graded prevention and control. The model learns the characteristic information of typical faults of nickel-cobalt alloy electrochemical processing electrolyte through training, which can quickly distinguish the fault category and locate the fault cause. This avoids blind maintenance caused by alarm without location, and greatly shortens the fault handling time. In addition, the handling suggestion layer has a built-in standardized handling suggestion library. Based on the warning level and the root cause of the fault, it automatically generates suggestions that are suitable for on-site operation, further improving the intelligence level of the platform.

[0022] (5) The preprocessing unit of this invention improves the accuracy of parameter data by cleaning, denoising, completing and normalizing the data collected by the sensing system, and provides a high-quality data foundation for the platform. The sensing system is equipped with a group of physicochemical property sensors to collect parameter data of the physicochemical properties of the electrolyte, an electrochemical performance monitoring module to collect parameter data of the electrochemical performance of the electrolyte, and a component content detection module combined with a circulating sampling pump and a storage container to collect data of the component content of the electrolyte. As the front-end core of electrolyte state detection, the sensing system realizes multi-parameter integrated and accurate sensing, and provides an accurate data source for downstream data preprocessing and model algorithms. Attached Figure Description

[0023] Figure 1 This is a schematic diagram illustrating the structural principle of the intelligent electrolyte state detection platform of the present invention.

[0024] Figure 2 This is a schematic diagram of the sensing system of the intelligent electrolyte state detection platform of the present invention.

[0025] Figure 3 This is a schematic diagram of the data processing flow of the preprocessing unit of the intelligent electrolyte state detection platform of the present invention.

[0026] Figure 4 This is a schematic diagram of the interactive platform and control system of the intelligent electrolyte state detection platform of the present invention.

[0027] In the above figures, the component names corresponding to the reference numerals are as follows: 1-Sensing system; 101-Physicochemical property sensor group; 102-Component content detection module; 103-Electrochemical performance monitoring module; 2-Preprocessing unit; 3-Verification unit; 4-Multi-source data fusion unit; 5-Time-series recursion unit; 6-State recognition model; 7-Fault early warning model; 8-Trend prediction model; 9-Interactive platform; 10-Control system; 11-Electrolyte; 12-Processing tank; 13-Circulating sampling pump; 14-Storage container. Detailed Implementation

[0028] The present invention will be further described below with reference to the accompanying drawings and embodiments. The embodiments of the present invention include, but are not limited to, the following embodiments.

[0029] Example like Figures 1 to 4 As shown, this embodiment provides an intelligent detection platform for the state of electrolyte in the electrochemical processing of nickel-cobalt alloys, including a sensing system 1, a preprocessing unit 2, a verification unit 3, a state recognition model 6, a time-series recursive unit 5, a trend prediction model 8, a multi-source data fusion unit 4, a fault early warning model 7, an interactive platform 9, and a control system 10. The sensing system 1 collects parameter information such as the physicochemical properties, component content, and electrochemical performance of the electrolyte 11 in real time online. The preprocessing unit 2 cleans, denoises, completes, and normalizes the data collected by the sensing system 1. The verification unit 3 receives the data processed by the preprocessing unit 2 and calibrates the parameters of ion concentration, conductivity, and pH value. The state recognition model 6 identifies and judges the normal state, critical state, and abnormal state of the electrolyte 11 based on the output data of the verification unit 3. The time-series recursive unit 5... Unit 5 receives the output data from verification unit 3 and performs time-series enhancement on ion concentration and conductivity parameters. Trend prediction model 8 predicts the future trend of parameter changes based on the output data from time-series recursion unit 5. Multi-source data fusion unit 4 receives the output data from verification unit 3, the state recognition result from state recognition model 6, and the time-series prediction trend information from trend prediction model 8, and identifies platform faults. Fault warning model 7 realizes early warning and root cause analysis of electrolyte faults based on the output data from multi-source data fusion unit 4. Interactive platform 9 intuitively displays the state recognition result from state recognition model 6, the time-series prediction trend information from trend prediction model 8, and the fault information from fault warning model 7. Control system 10 adjusts the electrolyte state according to the output results of state recognition model 6, trend prediction model 8, and fault warning model 7.

[0030] The physicochemical properties of the electrolyte 11 in this embodiment reflect the macroscopic physicochemical state of the electrolyte and are fundamental indicators of the process environment for electrochemical machining of nickel-cobalt alloys. They directly affect ion transport rate, electrolyte wettability, and the stability of the processing interface reaction. Specifically, these properties include electrolyte density, temperature, conductivity, and pH value. The electrolyte composition reflects the effective component concentration and harmful impurity level of the electrolyte in the electrochemical machining of nickel-cobalt alloys. It is the essential determinant of electrolyte performance and directly affects the machining accuracy, alloy composition uniformity, and electrolyte lifespan. Specifically, this includes ion concentration, complexing agent content, additive content, and impurity content. Ion concentration refers to the total concentration of nickel ions and cobalt ions. The electrochemical performance of the electrolyte is characterized by current density and current efficiency parameters. The current density parameter is the corrosion current density obtained by fitting the cathodic polarization curve of the nickel-cobalt alloy workpiece in the electrolyte. The current efficiency parameter is the ratio of the actual deposited metal mass to the theoretically calculated value during machining. Electrochemical performance is the actual performance of the electrolyte in the electrochemical machining process of nickel-cobalt alloys and directly affects the workpiece machining quality and machining efficiency.

[0031] In this embodiment, the feature input layer of the state recognition model 6 receives the output data of the verification unit 3. The feature enhancement layer uses a convolutional neural network to capture the coupling correlation between parameters. The convolutional neural network in this embodiment adopts a lightweight design, with only one convolutional layer and one pooling layer, which controls the amount of computation while ensuring feature extraction capability and adapts to real-time detection requirements. The state decision layer integrates the random forest algorithm and the support vector machine algorithm to construct a composite decision mechanism. The random forest algorithm is an ensemble learning method that improves the generalization ability and stability of the model by constructing multiple decision trees and combining their prediction results. The random forest algorithm in this embodiment consists of one hundred decision trees, which are responsible for ranking the importance of input features and completing the preliminary classification. The support vector machine algorithm is a supervised learning algorithm whose core is to find an optimal separating hyperplane to separate data points of different categories to the greatest extent. The support vector machine algorithm in this embodiment optimizes the classification boundary based on the RBF kernel function, and accurately distinguishes the boundary samples of critical state, normal state and abnormal state in the preliminary classification by the random forest algorithm, solving the classification bias problem of single model under small sample and high-dimensional data. The RBF kernel function is a radially symmetric scalar function. RBF stands for Radial Basis. The abbreviation for Function refers to radial basis function. In this embodiment, the state decision layer includes a deviation quantification submodel that calculates the deviation of the measured values ​​of each parameter from the normal values, providing quantitative support for the model's classification results.

[0032] In this embodiment, the time-series feature input layer of the trend prediction model 8 receives the output data from the time-series recursion unit 5. The output data of the time-series recursion unit 5 is a continuous time-series sequence of the past two hours, with a sampling frequency of once per five minutes, for a total of twenty-four time steps. The time-series feature extraction layer uses a long short-term memory network to capture the long-range dependencies of parameters at different processing stages. The prediction output layer uses a fully connected layer to complete feature mapping and output the results, setting eight prediction time steps to correspond to the predicted values ​​of key parameters for the next two hours. The fault warning model 7 in this embodiment includes a graded warning layer, a root cause location layer, and a handling suggestion layer. The graded warning layer uses a two-factor coupled warning function to calculate the comprehensive risk value and grade the fault warning. The root cause location layer determines the fault root cause based on feature weights and parameter deviation using a fault root cause contribution function. The handling suggestion layer has a built-in standardized handling suggestion library and automatically generates suggestions adapted to on-site operations based on the warning level and fault root cause. This embodiment's fault warning model 7 adapts to the suddenness and multi-factor coupling characteristics of faults in electrolyte processing, balancing real-time performance and traceability, achieving both rapid alarm and providing clear suggestions for fault handling.

[0033] In this embodiment, the verification unit 3 is equipped with a conductivity-ion concentration correlation model to calibrate ion concentration and conductivity. The conductivity-ion concentration correlation model is characterized by the following function:

[0034] In the formula: Indicates electrical conductivity. Indicates nickel ion concentration. Indicates the cobalt ion concentration. ~ The model fit coefficients represent the weights of each variable's influence on conductivity. This represents a constant term reflecting the interference of secondary factors on conductivity. Indicates the temperature of the electrolyte; In this embodiment, the verification unit 3 is equipped with a pH-complexation reaction equilibrium model to calibrate and calculate the pH value. The pH-complexation reaction equilibrium model is characterized by the following function:

[0035] In the formula: Indicates the equilibrium degree of the complexation reaction. ~ These represent the model fit coefficients obtained through regression fitting using offline experimental data. The third term representing the pH value of the electrolyte. The term representing the quadratic value of the electrolyte pH is... The term representing the pH value of the electrolyte.

[0036] In this embodiment, the deviation quantification submodel is characterized using the following function:

[0037] In the formula: Indicates the degree of parameter deviation. This represents the real-time detection value. This represents the optimal value of the parameter. This indicates the upper limit of the normal range of the parameter. This indicates the lower limit of the normal range for the parameter.

[0038] In this embodiment, the timing recursion unit 5 is equipped with an ion concentration timing recursion model and a conductivity timing correction model. The ion concentration timing recursion model is expressed as follows:

[0039] In the formula: for Estimated ion concentration at time. for Ion concentration at time, This is the ion mass coefficient. For processing current, For time intervals, This is the total volume of the electrolyte. It is Faraday's constant. It is an ionic oxidation state. For current efficiency; The conductivity time-series correction model is expressed as:

[0040] In the formula: for Time-corrected conductivity for Real-time measured conductivity For temperature coefficient, for Temperature is measured in real time. For standard reference temperature, This is the ion concentration correction factor. for Nickel ion concentration at any given time. for Cobalt ion concentration at any given time.

[0041] In this embodiment, a prediction error correction function is constructed in the prediction output layer to correct the original prediction value. The prediction error correction function is as follows:

[0042] In the formula: This is the corrected predicted value. These are the original predicted values ​​from the model. for Historical average error at any given time This is the time decay coefficient.

[0043] In this embodiment, the multi-source data fusion unit 4 is equipped with a fault risk prediction model and a composite fault superposition model. The fault risk prediction model is represented by the following function:

[0044]

[0045] In the formula: Indicates the probability of a failure occurring. Indicates the risk coefficient. Indicates the rate of change of the parameter. This indicates the upper limit of the normal range of the parameter. This represents the optimal standardized value of the parameter. Indicates the time required for fault escalation. This indicates the deviation threshold for the next level of early warning. Indicates the current deviation. Indicates the rate of change of deviation; The superposition model of complex faults is characterized by the following function:

[0046] In the formula: Indicates the overall deviation of a complex fault. This indicates the maximum deviation of a single parameter. Indicates the weighting coefficient. Indicates the first Deviation of each parameter.

[0047] The two-factor coupled early warning function used in the hierarchical early warning layer of this embodiment is as follows:

[0048] In the formula: This indicates the overall risk value for fault warning. Indicates the trend probability weighting coefficient. Indicates the probability of risk in trend prediction. This represents the weighting coefficient for real-time deviation. Indicates the overall deviation of features; The root cause localization layer includes a fault root cause contribution function to quantify the contribution of each abnormal parameter to the fault. The formula for the fault root cause contribution function is as follows:

[0049] In the formula: Indicates the first The fault contribution of each abnormal parameter. Indicates the first Deviation of each parameter Indicates the overall fault deviation. Indicates the first Each parameter has a weighting coefficient.

[0050] In this embodiment, the sensing system 1 includes a physicochemical property sensor group 101, an electrochemical performance monitoring module 103, a circulating sampling pump 13, a liquid storage container 14, and a component content detection module 102. The physicochemical property sensor group 101 includes a density sensor, a temperature sensor, a conductivity sensor, and a pH sensor. These four types of sensors are selected according to the information in the table below. The electrochemical performance monitoring module 103 focuses on current density and current efficiency monitoring, integrating a three-electrode detection unit and an auxiliary calculation unit to achieve high-precision monitoring. Specifically, the monitoring accuracy is: current density accuracy ≤ ±0.01 A / cm². 2 The current efficiency accuracy is ≤±1%. In this embodiment, the circulating sampling pump 13 samples the electrolyte 11 in the processing tank 12, and the storage container 14 stores the electrolyte 11 sampled by the circulating sampling pump 13. The component content detection module 102 detects the component content of the electrolyte 11 in the storage container 14. In the component content detection module 102, the ion concentration is collected in real time by an ion-selective electrode to collect nickel ion concentration and cobalt ion concentration data. The complexing agent content is quantitatively detected based on the characteristic absorption peak of sodium citrate using an online near-infrared spectroscopy sensor. The additive content is quantitatively detected based on the characteristic absorption wavelength of the additive using an online ultraviolet-visible spectroscopy sensor. The impurity content is detected simultaneously by an online laser particle size analyzer and a conductivity-coupled sensor to detect the total amount of suspended and dissolved impurities. The deployment and connection of the physicochemical property sensor group 101, the electrochemical performance monitoring module 103, the circulating sampling pump 13, the storage container 14, and the component content detection module 102 in this embodiment can be implemented by existing conventional software and hardware designs. The specific software settings and circuit structures will not be described in this embodiment.

[0051]

[0052] In this embodiment, the preprocessing unit 2 cleans, denoises, completes, and normalizes the raw data collected by the sensing system 1 to eliminate on-site interference, data loss, and dimensional differences, outputting standardized, high-quality data. The specific processing steps are as follows: identifying and removing abnormal interference values, duplicate data, and data with incorrect formats from the raw data to prevent invalid data from entering subsequent units or models, which could lead to computational distortion or false alarms; eliminating high-frequency noise caused by electromagnetic interference, vibration, and electrolyte bubbles in the industrial field, while preserving the true trend of data changes; adopting a completion strategy that adapts to the feature type to address data loss caused by sensor failure or communication interruption, avoiding data gaps from affecting model computation; and eliminating differences in the dimensions and numerical ranges of different features, enabling weighted computation of features and adapting the coupling logic of deviation quantification and fault warning models.

[0053] When using this invention, the overall intelligent detection platform is first built. The sensing system 1 detects the electrolyte state parameter data in real time online. The preprocessing unit 2 processes and optimizes the collected raw data. The verification unit 3, the time-series recursion unit 5, and the multi-source data fusion unit 4 further process the data through built-in model functions. The state recognition model 6, the trend prediction model 8, and the fault early warning model 7 complete the state recognition, trend prediction, and fault early warning. The interactive platform 9 in this invention serves as the data application center of the intelligent detection system for the state of the electrolyte in the electrochemical processing of nickel-cobalt alloys. It undertakes the functions of data aggregation, intelligent analysis, visualization, and human-computer interaction. The interactive platform 9 adopts a multi-terminal collaborative display method, including a monitoring screen, a web terminal, and a mobile terminal. The monitoring screen uses a 3D visualized tank model to display the heat map of the distribution of various parameters in real time, and abnormal parameters are highlighted. The web terminal supports viewing multi-dimensional trend analysis charts and parameter drill-down queries. The mobile terminal uses a lightweight monitoring interface and pushes abnormal alarms. The control system 10 of this invention includes a component replenishment unit to regulate ion concentration, complexing agent content, additive content, conductivity, and pH value; a temperature and density adjustment unit to regulate temperature and density; a filtration unit to control impurity content; and a current adjustment unit to regulate current density and current efficiency. The intelligent detection platform constructed by this invention, through multi-source data detection, data optimization and integration, efficient model setting, interactive platform, and control system application, achieves precise and effective monitoring of the electrolyte state, thereby improving the accuracy and efficiency of nickel-cobalt alloy electrochemical processing.

[0054] The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any changes made based on the design principles of the present invention, or any non-creative modifications made thereon, shall fall within the scope of protection of the present invention.

Claims

1. A nickel-cobalt alloy electrochemical machining electrolyte state intelligent detection platform, characterized in that The system includes a sensing system (1) that collects parameter information on the physicochemical properties, component content, and electrochemical performance of the electrolyte (11) in real time online; a preprocessing unit (2) that optimizes the data quality of the data collected by the sensing system (1); a verification unit (3) that receives the data processed by the preprocessing unit (2) and calibrates the parameters of ion concentration, conductivity, and pH value; a state recognition model (6) that identifies and judges the normal state, critical state, and abnormal state of the electrolyte (11) based on the output data of the verification unit (3); a time-series recursion unit (5) that receives the output data of the verification unit (3) and performs time-series enhancement on the parameters of ion concentration and conductivity; and a trend prediction model that predicts the future time-series change trend of the parameters based on the output data of the time-series recursion unit (5). (8), a multi-source data fusion unit (4) that receives the output data of the verification unit (3), the state recognition result of the state recognition model (6), and the time-series prediction trend information of the trend prediction model (8) and identifies platform faults; a fault warning model (7) that realizes early warning and root cause analysis of electrolyte faults based on the output data of the multi-source data fusion unit (4); an interactive platform (9) that intuitively displays the state recognition result of the state recognition model (6), the time-series prediction trend information of the trend prediction model (8), and the fault information of the fault warning model (7); and a control system (10) that adjusts the state of electrolyte according to the output results of the state recognition model (6), the trend prediction model (8), and the fault warning model (7), wherein, The state recognition model (6) includes a feature input layer that receives the output data of the verification unit (3), a feature enhancement layer that uses a convolutional neural network to capture the coupling relationship between parameters, and a state decision layer that integrates random forest and support vector machine decision-making mechanisms to complete classification decisions. The state decision layer is equipped with a deviation quantification sub-model to calculate the deviation of the measured values ​​of each parameter from the normal values. The trend prediction model (8) includes a time-series feature input layer that receives the output data of the time-series recursion unit (5), a time-series feature extraction layer that uses a long short-term memory network to capture the long-range dependence of parameters at different processing stages, and a prediction output layer that uses a fully connected layer to complete feature mapping. The fault early warning model (7) includes a graded early warning layer, a root cause location layer, and a handling suggestion layer. The graded early warning layer uses a two-factor coupled early warning function to calculate the comprehensive risk value and classify the fault early warning. The root cause location layer determines the root cause of the fault based on the feature weight and parameter deviation degree through the fault root cause contribution function. The handling suggestion layer has a built-in standardized handling suggestion library and automatically generates suggestions that are suitable for on-site operations based on the early warning level and the root cause of the fault.

2. The platform for intelligent detection of the state of an electrochemical processing electrolyte for a nickel-cobalt alloy according to claim 1, characterized in that, The verification unit (3) is equipped with a conductivity-ion concentration correlation model to calibrate ion concentration and conductivity, and the verification unit (3) is equipped with a pH value-complexation reaction equilibrium model to calibrate and calculate pH value.

3. The platform for intelligent detection of the state of an electrochemical processing electrolyte for a nickel-cobalt alloy according to claim 2, characterized in that, The deviation quantification sub-model is characterized by the following function: ； In the formula: represents a parameter deviation degree, represents a real-time detection value, represents a parameter optimal value, represents an upper limit of a parameter normal range, represents a lower limit of a parameter normal range.

4. The intelligent detection platform for the state of electrolyte in electrochemical machining of nickel-cobalt alloys according to claim 2, characterized in that, The timing recursion unit (5) is equipped with an ion concentration timing recursion model based on historical data to predict future ion concentration data and a conductivity timing correction model for eliminating temperature and noise interference to obtain accurate conductivity timing data.

5. The intelligent detection platform for the state of electrolyte in electrochemical machining of nickel-cobalt alloys according to claim 4, characterized in that, The prediction output layer constructs a prediction error correction function to correct the original prediction value. The prediction error correction function is as follows: ； In the formula: This is the corrected predicted value. These are the original predicted values ​​from the model. for Historical average error at any given time This is the time decay coefficient.

6. The intelligent detection platform for the state of electrolyte in electrochemical machining of nickel-cobalt alloys according to claim 5, characterized in that, The multi-source data fusion unit (4) is equipped with a fault risk prediction model and a composite fault superposition model. The fault risk prediction model is based on the parameter change rate output by the trend prediction model and constructs a risk prediction function to predict the probability of fault occurrence in advance. The composite fault superposition model is for composite faults with multiple parameters that are abnormal at the same time. It constructs a superposition function to quantify the overall fault severity.

7. The intelligent detection platform for the state of electrolyte in electrochemical machining of nickel-cobalt alloys according to claim 6, characterized in that, The two-factor coupled early warning function used in the hierarchical early warning layer is as follows: ； In the formula: This indicates the overall risk value for fault warning. Indicates the trend probability weighting coefficient. Indicates the probability of risk in trend prediction. This represents the weighting coefficient for real-time deviation. Indicates the overall deviation of features; The root cause localization layer includes a fault root cause contribution function to quantify the contribution of each abnormal parameter to the fault. The formula for the fault root cause contribution function is as follows: ； In the formula: Indicates the first The fault contribution of each abnormal parameter. Indicates the first Deviation of each parameter Indicates the overall fault deviation. Indicates the first Each parameter has a weighting coefficient.

8. The intelligent detection platform for the state of electrolyte in electrochemical machining of nickel-cobalt alloys according to claim 1, characterized in that: The sensing system (1) includes a physicochemical property sensor group (101) and an electrochemical performance monitoring module (103) for detecting the physicochemical properties and electrochemical performance of the electrolyte (11) in the processing tank (12), a circulating sampling pump (13) for sampling the electrolyte (11) in the processing tank (12), a storage container (14) for storing the electrolyte (11) taken by the circulating sampling pump (13), and a component content detection module (102) for detecting the component content of the electrolyte (11) in the storage container (14).

9. The intelligent detection platform for the state of electrolyte in electrochemical machining of nickel-cobalt alloys according to any one of claims 1-8, characterized in that: The preprocessing unit (2) performs the following data preprocessing steps on the data collected by the sensing system (1): identifying and removing abnormal interference values, duplicate data and data with incorrect format in the original data; eliminating high-frequency noise caused by electromagnetic interference, vibration and electrolyte bubbles in the industrial field, and retaining the true trend of data change; and using an adaptive feature type completion strategy to complete the data for data loss caused by sensing system failure and communication interruption. Eliminate the differences in the dimensions and numerical ranges of different features, enable weighted operations on features, and adapt to the coupled logic of deviation quantification and fault early warning models.

Images