A brine tank water addition automatic control system and method

Abstract

The application discloses a kind of brine tank water quantity automatic control system and method, it is related to the field of automation control technology, method includes: real-time acquisition brine level, conductivity, temperature and flow signal;Real-time concentration and liquid level data are obtained by composite filtering and temperature compensation correction;Control instruction is calculated using improved PID algorithm combined with fuzzy logic;Drive actuator to adjust water inflow;Closed-loop feedback mechanism is constructed to realize dynamic balance.The system is composed of sensing module, data preprocessing module, decision control module, execution driving module and closed-loop feedback module.The application aims to solve the problems of logic simplification, dynamic compensation loss and low control precision in traditional control mode, through multi-parameter deep coupling and self-tuning control strategy, precise perception and intelligent adjustment of brine tank operating conditions are realized, the stability and production safety of brine concentration are significantly improved, and energy consumption and resource waste are effectively reduced.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, and in particular to an automatic control system and method for adding water to a salt tank. Background Technology

[0002] Salt baths are core equipment in processes such as salt and nitrate production, crystallization separation, and mineral salt dissolution. Their operational stability directly affects production continuity and product quality. Furthermore, in the field of mechanical heat treatment, the nitrate bath used in isothermal quenching processes, as a container for molten salt containing water, directly affects the quenching quality of the workpiece due to the stability of its water content and liquid level. Precise control of the water content relies on automatic adjustment of water addition. Accurate control of the water addition to the salt bath is a key means to maintain a constant solution concentration, stable liquid level, and balanced salt dissolution rate.

[0003] Traditional brine tank water addition control technology has several drawbacks: First, the control logic is overly simplified, relying heavily on manual experience or single liquid level threshold triggers, making it difficult to handle the dynamic balance requirements of complex salt systems. Second, existing technologies are mostly based on static physicochemical models for proportioning calculations, lacking effective real-time closed-loop compensation mechanisms when facing dynamic disturbances such as nonlinear changes in solubility caused by drastic fluctuations in ambient temperature during continuous production and real-time drift of sensor liquid levels. Third, the system does not pay enough attention to key process parameters specific to salt production, failing to achieve deep coupling control of core indicators such as solution saturation, real-time concentration, and salt layer thickness. Fourth, the control system architecture has limited ability to fuse and process multi-source sensor data, resulting in poor response accuracy and adaptability when handling specific brine tank conditions, making it difficult to achieve fine-tuning and easily causing problems such as large fluctuations in solution concentration, unstable product quality, and water waste. Specifically, the aforementioned defects manifest in the following technical challenges in actual salt tank operation that are difficult to solve through simple technology aggregation: real-time distortion of sensor signals—high salt mist and easily crystallizing environments cause conductivity drift and liquid level jumps, and traditional filtering cannot distinguish between signal anomalies and changes in operating conditions; dynamic uncertainty of dissolved salt—the thickness of the bottom salt layer changes in real time with the load, and traditional solutions ignore this factor, leading to a mismatch between water addition and salt dissolution requirements; difficulty in multi-objective coordinated control of liquid level and concentration—the response speeds of the two are significantly different and strongly coupled, and traditional decoupling strategies are ineffective; lack of large lag feedforward compensation—there is a pure lag of several minutes to tens of minutes from water addition to concentration change, and traditional feedback control will inevitably produce periodic fluctuations.

[0004] To address the aforementioned shortcomings, this invention proposes an automatic control system and method for adding water to a salt tank. Summary of the Invention

[0005] The purpose of this invention is to provide an automatic control system and method for adding water to a salt tank, thereby solving the complex technical problems of existing technologies in salt tank operations characterized by high salt content, easy crystallization, large time lag, and strong coupling of multiple parameters. These problems include the inability to effectively handle real-time distortion of sensor signals, dynamic uncertainty of dissolved salt, difficulties in multi-objective coordinated control of liquid level and concentration, and the lack of feedforward compensation for large time lag. To achieve the above objective, this invention provides the following technical solution:

[0006] An automatic control system for adding water to a brine tank, the system comprising:

[0007] The sensing module is used to synchronously acquire raw physical signals, including solution level, conductivity, solution temperature, and real-time inlet water flow rate, through a sensor array deployed inside the brine tank and on the inlet water pipeline; the raw physical signals are converted into standard current signals using a signal transmitter, and the standard current signals are converted into digital signals for transmission through an analog acquisition module; the sensing module also includes a pressure transmitter installed at the bottom of the brine tank to acquire the total pressure at the bottom of the tank.

[0008] The data preprocessing module is used to remove random noise and outliers from the original physical signal using a composite filtering algorithm, and to construct a temperature compensation model based on the nonlinear functional relationship between solution conductivity and temperature. It also uses real-time acquired solution temperature to adjust and correct the conductivity measurements, obtaining real-time salt concentration data and standardized liquid level data that reflect the true state inside the salt tank. Furthermore, the data preprocessing module is used to calculate the real-time solution density based on the total pressure at the bottom of the tank obtained by the pressure transmitter and the liquid level height obtained by the ultrasonic level gauge, thereby estimating the thickness of the undissolved solid salt layer inside the salt tank.

[0009] The decision control module compares the real-time salt concentration data and the standardized liquid level data with preset target concentration and target liquid level values, respectively, to calculate the concentration deviation and liquid level deviation. Using a PID control algorithm with parameter self-tuning capabilities combined with fuzzy logic reasoning, the control logic is divided into multiple intervals based on the magnitude of the deviation. Within each interval, the corresponding control coefficient is retrieved, and the control output u(t) is calculated according to the following formula:

[0010]

[0011] Where t represents the sequence number of the current control cycle (t=0,1,2,…). The control period is 1 second in this embodiment; E(t) represents the concentration deviation or liquid level deviation. , , The proportional, integral, and derivative coefficients are dynamically adjusted through fuzzy logic reasoning. The control output u(t) is converted into a 4-20mA standard signal after range conversion and output as a drive command to the execution drive module to adjust the opening of the electric regulating valve or the frequency of the variable frequency water pump, thereby achieving precise control of the inlet water flow.

[0012] The control logic is divided into three intervals: when the deviation is greater than the first threshold, an open-loop high-flow water addition mode is adopted, and the actuator is fully open to quickly reduce the deviation; when the deviation is between the first and second thresholds, the fuzzy logic controller is activated, and the PID coefficients are retrieved from the preset fuzzy rule base according to the deviation and its rate of change; when the deviation is less than the second threshold, the fine adjustment mode is switched to eliminate static error through the cumulative effect of the integral term; the decision control module also dynamically adjusts the target water addition value and the salt dissolution rate prediction according to the calculated salt layer thickness, and automatically reduces the rising slope of the target water addition value when the salt layer thickness is lower than the preset threshold.

[0013] The execution drive module is used to generate a drive signal of corresponding intensity according to the drive command, and adjust the opening of the electric regulating valve or the operating frequency of the variable frequency water pump on the water inlet pipeline through the analog output module to adjust the water inlet flow rate; the execution drive module also includes an anti-vibration unit, which is used to monitor the rate of change of the action direction of the drive command, and automatically increase the control dead zone and smooth the drive command when the rate of change of the action direction exceeds a preset threshold to avoid frequent valve vibration;

[0014] The closed-loop feedback module is used to monitor the changing trend of parameters inside the salt tank in real time during the water addition process, and adjust the control weights according to the parameter regression rate until the liquid level and concentration reach the preset equilibrium dead zone range.

[0015] A bidirectional coupling architecture is constructed between the data preprocessing module and the decision control module: the signal characteristic parameters output by the data preprocessing module are fed back to the decision control module in real time for dynamic adjustment of control parameters; the control output quantity output by the decision control module is synchronously fed back to the data preprocessing module for dynamic adjustment of filtering parameters; the bidirectional coupling architecture is used to realize deep information interaction between the perception layer and the control layer, forming a closed-loop linkage mechanism that adapts to operating conditions.

[0016] Preferably, the decision control module further includes a feedforward prediction unit, which is used to predict the trend of salt tank concentration change based on historical operating data using a long short-term memory neural network, and convert the prediction result into a feedforward compensation term superimposed on the PID control output, so as to intervene in water addition in advance when the concentration decrease trend appears, and overcome the physical lag in the salt dissolution process.

[0017] Preferably, the system also includes a safety protection module for real-time monitoring of sensor signal integrity. When a signal value exceeds a preset normal current range threshold, it is determined to be a sensor malfunction, automatically switching to manual safety mode and blocking the water addition command. When the ultrasonic level gauge signal is abnormal, it automatically switches to the level calculation mode based on the pressure transmitter. The salt tank is also equipped with a mechanical high-level float switch independent of the control system, which forcibly cuts off the water supply when the level exceeds the safety limit height.

[0018] Preferably, the system further includes a human-machine interface and a data communication module. The human-machine interface uses an industrial-grade touchscreen to display the salt tank's level curve, concentration trend, valve status, and historical alarm records in real time. The data communication module supports industrial Ethernet protocol or serial communication protocol, enabling the uploading of all process parameters to the factory's production execution system for remote monitoring and data aggregation.

[0019] Preferably, when the system is applied to continuous production conditions, it automatically records the concentration recovery time after each water addition cycle and uses a long short-term memory neural network to model the dissolution load of the brine tank. Based on historical operating data for preset cycles, the system can predict the concentration change trend within a predetermined time period, thereby achieving feedforward control and intervening in water addition in advance when the concentration decline trend begins to appear, reducing the concentration fluctuation amplitude.

[0020] An automatic control method for water addition to a salt tank includes the following steps:

[0021] Step 1: Real-time acquisition of multi-source operating parameters inside the salt tank: A sensor array deployed inside the salt tank and on the inlet water pipeline synchronously acquires raw physical signals including solution level, conductivity, solution temperature, and real-time inlet water flow rate; the total pressure at the bottom of the tank is obtained through a pressure transmitter installed at the bottom of the salt tank, and combined with the liquid level height obtained by the ultrasonic level gauge for subsequent calculation of the real-time solution density and the thickness of the undissolved solid salt layer; the raw physical signals are converted into standard electrical signals for transmission.

[0022] Step 2: Data preprocessing and fusion of multi-source operating parameters: A composite filtering algorithm is used to remove random noise and outliers from the original physical signal, and the conductivity signal is corrected based on a temperature compensation model to obtain real-time salt concentration data and standardized liquid level data that reflect the true state of the salt tank; the real-time density of the solution is calculated based on the total pressure at the bottom of the tank and the liquid level height obtained by the pressure transmitter, and then the thickness of the undissolved solid salt layer in the salt tank is deduced.

[0023] Step 3: Calculate control deviation and decision control output: Compare real-time salt concentration data and standardized liquid level data with preset target concentration and target liquid level values ​​respectively to calculate concentration deviation and liquid level deviation; Utilize a PID control algorithm with parameter self-tuning function combined with fuzzy logic reasoning to divide the control logic into multiple intervals according to the magnitude of the deviation, and retrieve the corresponding control coefficients in different intervals to calculate and output a numerical control output for adjusting the water inlet drive mechanism; Dynamically adjust the target water addition value and salt dissolution rate prediction based on the calculated salt layer thickness; When the salt layer thickness is lower than the preset threshold, automatically reduce the upward slope of the target water addition value;

[0024] Step 4: Drive the actuator to perform precise water addition: The controller outputs a drive signal of corresponding strength according to the drive command generated after the control output quantity is converted through range conversion, and adjusts the opening of the electric regulating valve on the water inlet pipe or the operating frequency of the variable frequency water pump to regulate the water flow rate into the salt tank; during the drive process, the rate of change of the action direction of the drive command is monitored. When the rate of change of the action direction exceeds the preset threshold, the control dead zone is automatically increased and the drive command is smoothed.

[0025] Step 5: Construct a closed-loop feedback and verification mechanism: Monitor the changing trend of parameters inside the brine tank in real time during the water addition process, adjust the control weights according to the parameter regression rate, until the liquid level and concentration reach the preset equilibrium dead zone range, and complete a complete automatic water addition control cycle;

[0026] In this process, the signal characteristic parameters output in step 2 are fed back to step 3 in real time for dynamic adjustment of control parameters, and the control output quantity output in step 3 is fed back to step 2 synchronously for dynamic adjustment of filter parameters, forming a two-way coupled architecture.

[0027] Preferably, the sensor array in step 1 includes an ultrasonic level gauge, a four-electrode conductivity sensor, a platinum resistance temperature sensor, and an electromagnetic flow meter. The ultrasonic level gauge is mounted on a measuring bracket at the top of the brine tank; its measuring range is adapted to the depth of the brine tank, its measurement blind zone is within a preset allowable range, and its output signal is a preset standard current signal. The four-electrode conductivity sensor uses graphite electrodes with anti-fouling properties, and its measuring range covers a preset conductivity range, enabling long-term continuous monitoring in high-salinity environments. The platinum resistance temperature sensor uses a three-wire connection to eliminate measurement errors caused by wire resistance, achieving a preset accuracy threshold. The electromagnetic flow meter is installed on the inlet main pipeline and measures flow rate by sensing the induced electromotive force generated when the conductive medium flows through a magnetic field; it features no obstructing components and low pressure loss.

[0028] Preferably, the composite filtering algorithm in step 2 combines median filtering and weighted moving average filtering. The system first sorts a preset number of continuously collected sampling points and takes the median value to eliminate instantaneous liquid level jumps caused by stirring or material input within the salt tank. Then, it performs a sliding window weighted average on the sequence after median filtering. The weighting coefficient decays exponentially with the time distance of the sampling points, thus maintaining good dynamic response characteristics while ensuring signal smoothness. The temperature compensation model is based on the nonlinear functional relationship between solution conductivity and temperature. Since the conductivity of the salt solution increases with increasing temperature, the system uses a preset reference temperature as a standard and converts the conductivity measurement value using the real-time collected solution temperature. The compensation coefficient is preset according to the characteristics of different salt systems to ensure that the conversion error of salt concentration is within a preset allowable range within the preset operating temperature range.

[0029] Preferably, the improved PID control algorithm in step 3 has a parameter self-tuning function. The controller divides the control logic into multiple intervals according to the magnitude of the deviation. When the deviation is greater than a preset first threshold, the system adopts an open-loop high-flow-rate water injection mode, with the actuator fully open to quickly reduce the deviation; when the deviation is between the first and second thresholds, the system activates the fuzzy logic controller, which retrieves the optimal PID coefficients from a preset fuzzy rule base based on the deviation and its rate of change; when the deviation is less than the second threshold, the system switches to a fine-tuning mode, eliminating static errors through the cumulative effect of the integral term to ensure control accuracy.

[0030] Preferably, the method further includes a feedforward control step: the system automatically records the concentration recovery time after each water addition cycle and uses a long short-term memory neural network to model the dissolution load of the salt tank; based on the historical preset cycle operation data, the system predicts the concentration change trend in the future predetermined time period, intervenes in water addition in advance when the concentration decline trend is initially apparent, and converts the prediction result into a feedforward compensation term to be introduced into the control loop.

[0031] Preferably, the output frequency of the control output is set within a preset operating frequency range to ensure smooth operation of the actuator. The controller uses a programmable logic controller as its core processing unit, which integrates a high-speed pulse output module and an analog output module.

[0032] Preferably, in step 4, the electric regulating valve has a percentage flow characteristic, and its actuator includes a position feedback potentiometer, which can feed back the actual valve opening signal to the controller in real time. The controller compares the commanded opening with the feedback opening to achieve closed-loop control of the valve position, and the positioning accuracy reaches the preset target accuracy. The variable frequency water pump receives a preset voltage signal through the frequency converter to achieve stepless adjustment of the speed within the preset speed range.

[0033] Preferably, the method further includes anomaly monitoring and safety protection functions: the system monitors the signal integrity of the sensor in real time, and once the signal value is detected to exceed the preset normal current range threshold, it is determined to be a sensor disconnection or short circuit fault. At this time, the system automatically switches to manual safety mode and blocks the water addition command; when the ultrasonic level gauge signal is abnormal, it automatically switches to the level calculation mode based on the pressure transmitter; at the same time, the salt tank is equipped with a mechanical high level float switch independent of the control system. When the liquid level exceeds the safety limit height, the power supply of the water inlet pump is forcibly cut off and the electric regulating valve is closed to prevent the tank from overflowing.

[0034] Preferably, the drive logic of the actuator includes an anti-oscillation algorithm. When the actuator frequently changes its direction of motion within a short period of time, the system automatically increases the control dead zone to avoid increased mechanical wear and energy consumption caused by frequent valve vibration.

[0035] Preferably, the present invention can also be used in the quenching process. By adding a quenching mode to the system, the water concentration increases and the liquid level decreases due to the violent evaporation of water during the quenching process. The present invention can also periodically add water through the system according to the quenching requirements.

[0036] Preferably, the present invention is also applicable to the automatic water replenishment control of isothermal quenching nitrate baths. By using a high-frequency impedance spectroscopy sensor or conductivity sensor to detect the trace water content in the nitrate melt in real time, and in conjunction with an immersion water supply pipe and a temperature safety interlock mechanism, the water content is precisely controlled within the target range (e.g., 0.5%~4.0%), thereby effectively stabilizing the quenching cooling rate and improving the consistency of heat treatment quality.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] This invention achieves precise control through deep coupling of multiple parameters and overcomes the limitations of traditional single-level liquid control. By integrating multiple sensors, a multi-dimensional operating condition sensing system is constructed. Combined with a temperature compensation model and density calculation logic, the system can acquire real-time, accurate salt concentration data, solving the interference of ambient temperature fluctuations on measurement accuracy. Through the deep fusion of proportional-integral-differential algorithms and fuzzy logic, coordinated adjustment of concentration deviation and liquid level deviation is achieved, keeping the physicochemical balance inside the brine tank within an extremely narrow fluctuation range. This significantly improves the quality stability of the finished brine, and the concentration fluctuation rate is significantly reduced compared to traditional methods.

[0039] This invention possesses a high degree of dynamic process adaptability and intelligence. Through the self-tuning control strategy, it can automatically identify the operating characteristics under different production loads. By dividing the control interval, it achieves seamless switching from rapid water replenishment to fine adjustment. The introduced long short-term memory neural network prediction model endows the system with feedforward control capabilities, enabling it to predict process trends based on historical data and transform from passive response to active adjustment. This intelligent control architecture effectively solves the control lag problem caused by fluctuations in raw salt composition, environmental interference, and equipment drift during continuous production. The system response time is significantly shortened, greatly improving the automation efficiency of the production line.

[0040] This invention significantly improves the safety and resource utilization of the system. By constructing multiple safety protection mechanisms, it eliminates production accidents such as overflow, dry tank, and concentration runaway from both hardware and software dimensions. At the same time, precise water addition control avoids energy waste and increased load on subsequent evaporation processes due to over-dilution, realizing refined management of water resources. While reducing the intensity of manual labor, it effectively reduces the system's operating energy consumption, providing core technical support for the green and intelligent transformation of the salt chemical industry.

[0041] This invention enhances the system's stability and long-term operational capability. By employing sensors with anti-pollution properties and flow meters without obstruction components, it solves the problems of sensor vulnerability and drift in high-salt and easily crystallizing environments. The application of a composite filtering algorithm effectively shields complex electromagnetic interference in industrial settings, ensuring the purity of control signals. The combination of anti-vibration algorithms and variable frequency drive technology reduces mechanical fatigue of actuators, extends the service life of valves and pumps, lowers system maintenance costs, and ensures long-term stable operation of the production equipment in harsh industrial environments.

[0042] This invention addresses the problem of real-time sensor signal distortion by deeply coupling a composite filtering algorithm with a temperature compensation model and dynamically adjusting the filtering parameters based on the control output, achieving accurate identification and effective removal of abnormal signals. For the problem of dynamic uncertainty in dissolved salt, it achieves online soft measurement of salt layer thickness for the first time through a combination of a pressure transmitter and an ultrasonic level gauge, dynamically adjusting the target water addition value and the predicted salt dissolution rate accordingly. For the problem of multi-objective coordinated control of liquid level and concentration, it constructs a coordinated control strategy adaptable to different operating conditions through multi-interval division and dynamic adjustment of deviation priority using a fuzzy PID algorithm. For the problem of missing large-lag feedforward compensation, it uses an LSTM neural network to predict concentration change trends in advance, incorporating the pure lag element into the control loop, achieving a paradigm shift from "deviation response" to "trend prediction."

[0043] The various technical means employed in this invention generate a synergistic effect that goes beyond simple superposition. Field measurements show that introducing fuzzy PID control alone can reduce concentration fluctuations from ±3% to ±1.5%; introducing LSTM feedforward prediction alone can reduce fluctuations to ±1.8%; introducing salt layer thickness compensation alone can reduce fluctuations to ±2.0%; and combining all three further reduces concentration fluctuations to within ±0.5%, far below the expected linear superposition of the individual effects of each technique (theoretical linear superposition value is ±0.7%), demonstrating the non-obviousness of the technical solution of this invention. Furthermore, this invention, through the combination of a pressure transmitter and an ultrasonic level gauge, achieves online soft measurement of salt layer thickness for the first time, solving the long-standing technical blind spot in the salt chemical industry of "inability to know the amount of undissolved salt in the tank in real time," thus improving salt dissolution efficiency by 12%. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the overall technical solution architecture of an automatic control system for water addition in a salt tank proposed in this invention.

[0045] Figure 2 This is a schematic diagram of the core principle framework of the multi-source working condition perception and fuzzy self-tuning algorithm in this invention;

[0046] Figure 3 This is a closed-loop feedback and logic flow diagram of the automatic control process for adding water to the salt tank in this invention. Detailed Implementation

[0047] The features and exemplary embodiments of various aspects of the present invention will now be described in detail. To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely intended to explain the present invention and not to limit the present invention. For those skilled in the art, the present invention can be practiced without some of these specific details. The following description of the embodiments is merely to provide a better understanding of the present invention by illustrating examples of the invention.

[0048] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0049] In the embodiments of the present invention, the same reference numerals denote the same components, and for the sake of brevity, detailed descriptions of the same components are omitted in different embodiments. It should be understood that the thickness, length, width, and other dimensions of various components in the embodiments of the present invention shown in the accompanying drawings, as well as the overall thickness, length, width, and other dimensions of the integrated device, are merely illustrative and should not constitute any limitation on the present invention; the term "multiple" in the present invention refers to two or more (including two).

[0050] Example 1: Basic Scheme

[0051] This embodiment provides an automatic control system for the water addition to a salt tank, deployed in the salt dissolving section of a vacuum salt production line with an annual output of 300,000 tons. See also... Figure 1 The system consists of a perception module, a data preprocessing module, a decision control module, an execution drive module, a closed-loop feedback module, and a human-computer interaction interface and data communication module.

[0052] The hardware configuration of the sensing module includes a sensor array deployed inside the salt tank and on the inlet water pipe, as detailed below:

[0053] Liquid level monitoring employs a high-frequency ultrasonic level gauge with a transducer frequency of 40kHz, a measuring range of 0–6 meters, and a measurement accuracy of ±2mm. Its measurement blind zone is within a preset allowable range. The level gauge is installed on a flange bracket at the top of the salt tank and transmits the liquid level height to the controller in real time via a 4–20mA current signal.

[0054] Concentration monitoring utilizes a four-electrode conductivity sensor. The electrodes are made of 316L stainless steel with a special anti-corrosion coating. The electrode constant is set to 10.0, and the measurement range covers 0–500 mS / cm, enabling long-term continuous monitoring in high-salinity environments. The sensor is equipped with a signal transmitter and features automatic range switching.

[0055] Temperature monitoring uses a Class A platinum resistance temperature sensor, encapsulated in a stainless steel protective tube, and connected to the temperature transmitter via a three-wire connection to eliminate wire resistance errors caused by long-distance transmission. The measurement accuracy reaches the preset accuracy threshold, and the measurement range is 0~100℃.

[0056] Flow monitoring utilizes an electromagnetic flow meter, installed on the main inlet water pipeline. The lining material is polytetrafluoroethylene (PTFE), and the electrode material is tantalum, providing excellent corrosion resistance. The flow range ratio reaches 100:1. This flow meter measures flow by sensing the induced electromotive force generated when a conductive medium flows through a magnetic field. It has no internal flow-obstructing components, resulting in minimal pressure loss.

[0057] The raw physical signals output by the aforementioned sensors are converted into 4-20mA standard current signals by a signal transmitter, and then converted into digital signals by an analog acquisition module before being transmitted to the controller.

[0058] In the workflow of the data preprocessing module, the controller adopts an industrial-grade programmable logic controller with a built-in 32-bit high-speed processor, and runs a composite filtering algorithm and a temperature compensation model.

[0059] The composite filtering algorithm combines median filtering and weighted moving average filtering. Its parameter design is based on the spectral analysis of the salt tank operating condition signal: the median filtering window length N=10. This value is determined by collecting the liquid level signal under normal operating conditions of the salt tank (sampling frequency 1Hz) and analyzing its spectral characteristics through Fourier transform. Spectral analysis shows that the liquid level jump frequency caused by stirring eddies is approximately 0.1~0.3Hz, while the instantaneous jump caused by material input exhibits high-frequency components above 0.5Hz. With a window length N=10, instantaneous jumps lasting 1~2 seconds can be effectively eliminated, while retaining the true liquid level change trend below 0.05Hz. The weighted moving average filtering uses exponentially decaying weights, with weight coefficients... Where λ=0.3, and i is the time distance between sampling points (i=0 is the current point, i=1 is the previous point, and so on). This value of λ was determined through step response experiments: a step water addition signal was applied to the salt tank, and the filtering effect was tested at λ=0.1, 0.2, 0.3, 0.4, and 0.5 respectively. The final set value of λ=0.3, which has the shortest adjustment time and the smallest overshoot, was taken as the final set value, achieving the optimal balance between smoothness and response speed.

[0060] The temperature compensation model uses a standard reference temperature of 25℃ as a benchmark and calculates the conductivity value by adjusting the real-time collected solution temperature. The compensation coefficient is determined through large-sample salt dissolution experiments based on the characteristics of different salt systems.

[0061] Within the operating temperature range of 10℃ to 60℃, standard concentration saline solutions (concentration range of 5% to 30%) were prepared at 5℃ intervals, and the conductivity values ​​at each temperature were measured using a high-precision laboratory conductivity meter. The conductivity-temperature characteristic curves of sodium chloride, potassium chloride, and sodium sulfate systems were fitted using the least squares method to determine the compensation coefficient α for each system.

[0062] For the sodium chloride system, the fitted value was determined to be α = 0.021 (i.e., 2.1% / ℃), and the concentration conversion error was ≤ ±0.2% in the range of 10~60℃. For the potassium chloride system, the fitted value was determined to be α = 0.020. For the sodium sulfate system, since there is a phase transition point at 32.4℃, a piecewise compensation model was adopted. The experiment determined that α = 0.018 when t ≥ 32.4℃ and α = 0.016 when t < 32.4℃, ensuring that the conversion error near the phase transition point is ≤ ±0.3%.

[0063] The aforementioned compensation coefficients and segmented model parameters are stored in the controller's non-volatile memory, forming a dedicated salt compensation database. Operators can automatically retrieve this database after selecting the salt type through the human-machine interface.

[0064] The module ultimately outputs real-time salt concentration data and standardized liquid level data that reflect the true state inside the salt tank.

[0065] An adaptive linkage mechanism based on operating conditions is established between the data preprocessing module and the decision control module:

[0066] The initialization process of the bidirectional coupled architecture is as follows: after system power-on, the data preprocessing module and the decision control module load their respective default parameters and run the first control cycle with these default parameters. From the second control cycle onwards, the bidirectional coupled architecture officially takes effect: the data preprocessing module sends signal characteristic parameters to the decision control module for adjusting control parameters, and the decision control module sends control outputs to the data preprocessing module for adjusting filter parameters. Upon system reset or restart, the above initialization process is automatically re-executed.

[0067] When the temperature compensation model detects drastic fluctuations in the solution temperature (temperature difference rate exceeding 5℃ / min), the system determines that the salt tank is in an unsteady state and automatically sends a gain scheduling signal to the decision control module to temporarily increase the proportional coefficient of the fuzzy PID by 20%~30% to enhance the system response speed.

[0068] When the composite filtering algorithm detects high-frequency pulsations (frequency higher than 0.5Hz) in the liquid level signal, the system determines that there is stirring eddy or material impact, and automatically sends a filtering synchronization signal to the decision control module to temporarily reduce the differential coefficient of the fuzzy PID by 50% to avoid misjudging high-frequency noise as a trend change and over-adjusting.

[0069] The control output from the decision control module, in addition to driving the actuator, is synchronously fed back to the data preprocessing module to dynamically adjust the window length of the weighted moving average filter: when the rate of change of the control output exceeds 10% / s, the filter window is shortened to improve response sensitivity; when the control output tends to stabilize (rate of change is less than 2% / s), the filter window is extended to enhance anti-interference capability.

[0070] See Figure 2 In the control logic of the decision control module, the decision control module compares the real-time salt concentration data and standardized liquid level data with the preset target concentration value and target liquid level value, respectively, calculates the concentration deviation and liquid level deviation, and uses an improved PID algorithm with parameter self-tuning function combined with fuzzy logic reasoning to divide the control logic into multiple intervals according to the magnitude of the deviation, retrieves the corresponding control coefficient in different intervals, and outputs the control output quantity.

[0071] Open-loop high-flow-rate water filling mode: When the concentration deviation or liquid level deviation exceeds a preset first threshold (e.g., 20%), the system enters the open-loop high-flow-rate water filling mode. In this mode, the controller no longer relies on the closed-loop regulation of the PID algorithm, but directly outputs saturation control commands: the electric regulating valve opening is forcibly set to 95%–100%, and the frequency of the variable frequency water pump is forcibly increased to 50Hz (rated frequency). At the same time, the system starts a timer to record the duration of continuous open-loop control. If the open-loop control continues for more than 120 seconds and the deviation still has not dropped below the first threshold, the system determines that there is an abnormal operating condition (such as blockage of the inlet pipe or abnormal dissolution of raw salt), automatically triggers an alarm, and switches to the safety protection mode.

[0072] The open-loop mode exit condition is as follows: both the concentration deviation and the liquid level deviation drop below the first threshold, and the liquid level rise rate tends to stabilize (the change rate is less than 5% / s for three consecutive sampling cycles). After the condition is met, the system smoothly switches to fuzzy logic control mode.

[0073] The fuzzy logic controller is activated when the deviation falls between a first threshold (20%) and a second threshold (5%). The fuzzy inference mechanism is as follows:

[0074] Input variable: Deviation E. First, the concentration deviation and liquid level deviation are normalized separately: the concentration deviation is divided by the allowable concentration fluctuation range (±5%), and the liquid level deviation is divided by the allowable liquid level fluctuation range (±0.2m) to obtain dimensionless normalized deviation values; then, the larger of the two absolute values ​​is taken as E. The universe of discourse of E is [-20%, 20%], quantized to 7 fuzzy levels: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large.

[0075] Deviation change rate This refers to the change in deviation E per unit time, reflecting the trend of deviation change. The calculation formula is: ,in The sampling period is 1 second in this embodiment. A value greater than 0 indicates that the deviation is widening. <0 indicates that the deviation is decreasing. The domain range is [-10% / s, 10% / s], quantized to 5 fuzzy levels: negative large, negative small, zero, positive small, and positive large.

[0076] Output variables:

[0077] proportional coefficient correction amount Used to adjust the scaling factor , ,in This is the initial scaling factor (taken as 1.2).

[0078] Integral coefficient correction Used to adjust the integral coefficient. , ,in The initial integration coefficient is 0.08.

[0079] Differential coefficient correction Δ Used to adjust the differential coefficients , ,in The initial differential coefficient is 0.3.

[0080] The initial PID coefficients were determined using the critical proportional gain method combined with a salt bath step response experiment.

[0081] Experimental procedure: Control the salt tank level at approximately 50% of the target value, apply a step opening signal with an amplitude of 10% to the electric regulating valve, and record the response curves of the level and concentration. Analysis of the response curves yields: pure lag time τ≈15 seconds, inertial time constant T≈45 seconds, and amplification factor K≈0.8 (% concentration change / 10% opening).

[0082] According to the Ziegler-Nichols tuning rule, the initial PID parameters are calculated as follows:

[0083]

[0084] in, The inertial time constant, in seconds. This is the magnification factor, expressed as % / 10% of the opening. The time delay is the pure time lag in seconds, and 1.2 is the Ziegler-Nichols tuning factor. However, during field commissioning, it was found that this value was too large, leading to severe overshoot. After three field experiments and optimizations, the final value was... Revised to 1.2, with corresponding adjustments. =0.08、 =0.3. The corrected parameters, in the step response test, showed a settling time of 120 seconds and an overshoot of 8%, meeting the process requirements.

[0085] Each output variable is quantized to 5 fuzzy levels: negative large, negative small, zero, positive small, and positive large.

[0086] Fuzzy rule base: The system stores 35 fuzzy rules, covering 7 levels of deviation E and the rate of change of deviation. All combinations of the 5 levels. Each rule simultaneously determines... , Δ The direction of change. The rule design follows the logic below, which is based on the mechanism analysis and simulation optimization of the salt tank process:

[0087] By applying water addition step signals of different amplitudes to the salt tank, the response curves of liquid level and concentration were measured, and the inertial time constant T≈45 seconds, pure lag time τ≈15 seconds, and amplification factor K varied with salt layer thickness between 0.6 and 1.2.

[0088] Based on the above characteristics, after coarsely tuning the initial PID parameters using the critical proportional gain method, the fuzzy rules are optimized offline in the Matlab / Simulink environment using a genetic algorithm. The optimization objective function is the weighted sum of the integral absolute error (IAE) and the control energy consumption (valve actuation amplitude): After 500 generations of evolution, it converged to the current rule table. This rule table has been verified in field operation. In the concentration step response test, the settling time was shortened from 180 seconds in the traditional PID to 120 seconds, and the overshoot was reduced from 15% to 8%.

[0089] proportionality coefficient The adjustment logic is as follows: When the deviation E is large (positive, positive, negative, or negative), Taking a positive value that is large or small enhances the proportional effect and speeds up the response; when the deviation E is small (small positive or small negative) and the rate of change of the deviation is large... When in the same direction as the deviation, Using a smaller or larger negative value weakens the proportional effect and prevents overshoot; when the deviation E is close to zero, Take zero or a small negative value to maintain stability.

[0090] Integral coefficient The adjustment logic is as follows: when the deviation E is large, Taking a larger or smaller negative value weakens the integral effect and prevents integral saturation; when the deviation E is close to zero... Taking a small or large positive value enhances the integral effect and eliminates steady-state error; when the rate of change of deviation... When the deviation E is opposite, the integral effect should be appropriately enhanced to accelerate the regression.

[0091] Differential coefficient Δ Adjustment logic: When the rate of change of deviation When it is in the same direction as the deviation E (e.g., the deviation is positive and still increasing). Taking a positive value (large or small) enhances the differential action and suppresses inertia; when the rate of change of the deviation... When the deviation E is opposite (e.g., the deviation is large but decreasing), Δ Taking a larger or smaller negative value weakens the differential effect and avoids excessive damping; when the deviation is close to zero and the rate of change is small, Δ Use zero or a negative value to reduce the impact of noise.

[0092] The following are some example rules (the complete rule base follows the logic described above):

[0093] In the aforementioned fuzzy inference mechanism, the domain of deviation E is not fixed but dynamically adjusted based on the signal quality assessment index fed back by the data preprocessing module. When the signal quality assessment index falls below a threshold (indicating sensor contamination or abnormal interference), the system automatically expands the domain of deviation E from ±20% to ±25% and adjusts the deviation change rate accordingly. The domain range is expanded from ±10% / s to ±15% / s to enhance the control algorithm's tolerance to uncertainty; when the signal quality is restored, the domain range is automatically restored to the original setting.

[0094] Output quantization value: The centroid method is used to convert the fuzzy output into precise PID coefficient correction values. This method weights the output value of each rule according to the activation degree of that rule, and then takes the weighted average as the final output. For example, the calculation formula is:

[0095]

[0096] in, The activation degree of the i-th rule is determined by the input variable deviation E and the rate of change of deviation. The membership degrees are multiplied together to obtain the membership degree, i.e. ; For the corresponding rule output Quantization value. The denominator is the sum of the activations of all rules, which serves as a normalization function to ensure that the output value is within a reasonable range. The quantization values ​​corresponding to each fuzziness level are: -0.6 for large negative values, -0.3 for small negative values, 0 for zero, 0.3 for small positive values, and 0.6 for large positive values.

[0097] Similarly, it can be calculated and Their quantization ranges are [-0.04, 0.04] and [-0.15, 0.15], respectively.

[0098] Calculation and use of PID coefficients:

[0099] Defuzzification yields the correction amount , , Then, the system calculates the PID coefficients for the current moment using the following formula:

[0100] , ,

[0101] in, , , The preset initial PID coefficients are 1.2, 0.08, and 0.3, respectively.

[0102] Then, the positional PID control algorithm is applied to calculate the control output u(t):

[0103]

[0104] Where t represents the sequence number of the current control cycle (t=0,1,2,…). The control period is 1 second in this embodiment; E(t) is the deviation value at the current moment. After range conversion, the control output u(t) is converted into a 4-20mA standard signal and output to the execution drive module to adjust the opening of the electric regulating valve or the frequency of the variable frequency water pump, thereby achieving precise control of the inlet water flow.

[0105] In fine-tuning mode, the system switches to fine-tuning mode when the deviation is less than the second threshold (5%). In this mode, the control strategy focuses on eliminating static error and preventing overshoot, specifically including:

[0106] Integral separation: when the absolute value of the deviation When it is less than 2%, the integral term participates normally in the control calculation, and the residual steady error is eliminated by utilizing the integral accumulation effect; when When the integral is greater than 2%, the integral term output is forcibly set to zero, i.e. It is controlled only by the proportional and derivative terms. The purpose of this strategy is to prevent the integral term from accumulating excessively and causing system overshoot when the deviation is large, and to activate the integral term to eliminate steady-state error when the deviation is close to the target value.

[0107] Integral Limiting: The system imposes upper and lower limits on the output value of the integral term. Let the output of the integral term be:

[0108]

[0109] in, E(j) is the integral coefficient, which is dynamically adjusted by the fuzzy logic controller; E(j) is the deviation value of the j-th sampling period, and Δt is the sampling period (1 second in this embodiment);

[0110] The actual integral term used is:

[0111]

[0112] in, , , To control the maximum output range (corresponding to 20mA output), this strategy prevents the integral term from accumulating too much and avoids the system remaining in a saturated state for a long time when the deviation is reversed.

[0113] Differential-first approach: This approach uses a differential-first structure, performing differential operations only on the measured value and not on changes in the setpoint, thus avoiding control shocks caused by sudden changes in the setpoint. The differential term expression is:

[0114]

[0115] in, The differential coefficient is dynamically adjusted by the fuzzy logic controller; C(t) is the measured concentration value; C(t−1) is the measured concentration value of the previous sampling period; Δt is the sampling period (1 second in this embodiment).

[0116] The purpose of this strategy is to ensure that when the operator modifies the target concentration or target liquid level, the change in the set value will not cause a sudden change in the differential term, thus avoiding drastic fluctuations in the control output.

[0117] Control dead zone: When |E(t)| < 0.5% and the duration exceeds 30 seconds, the system enters a sleep monitoring state, stops outputting adjustment commands, and only maintains the current actuator opening. When |E(t)| ≥ 0.5%, the system immediately wakes up and resumes adjustment. The purpose of this strategy is to avoid frequent actuator movements when the deviation is already very small, thereby reducing mechanical wear and energy consumption.

[0118] The mode switching mechanism employs a shockless switching strategy between the three control modes to ensure continuous control output during the switching process.

[0119] Switching from open-loop mode to fuzzy control mode: At the switching moment, the initial value of the integral term of the fuzzy controller is set to the steady-state control quantity corresponding to the current opening degree.

[0120] Switching from fuzzy control mode to fine adjustment mode: At the time of switching, the integral term of the fine adjustment mode inherits the current value of the integral term of the fuzzy controller.

[0121] Fine adjustment mode switches to fuzzy control mode: When the deviation expands again to more than 5% due to disturbance, the system automatically switches back to fuzzy control mode and clears the derivative term to zero.

[0122] The controller maintains a mode state machine internally, which records the current control mode in real time and automatically transitions the state based on the deviation and its rate of change.

[0123] The output frequency of the control output is set within a preset operating frequency range to ensure smooth operation of the actuator. The controller integrates a high-speed pulse output module and an analog output module.

[0124] In the execution mode of the drive module, the drive module generates a drive signal of corresponding strength according to the drive command, and adjusts the opening of the electric regulating valve on the water inlet pipe or the operating frequency of the variable frequency water pump through the analog output module to achieve linear intervention on the water flow rate entering the salt tank.

[0125] Electric control valve: Utilizing an equal percentage flow characteristic valve core and equipped with an intelligent actuator, it supports 4-20mA signal adjustment and features position feedback. The controller achieves closed-loop control of the valve position by comparing the commanded opening degree with the feedback opening degree, achieving a preset target positioning accuracy.

[0126] Variable frequency water pump: Driven by a 15kW three-phase asynchronous motor and equipped with a vector control type frequency converter, it receives a preset voltage signal and adjusts the pump speed according to the command frequency, achieving stepless speed adjustment within the preset speed range. The operating status parameters of the frequency converter are fed back to the controller in real time via a communication bus.

[0127] In the initial stage of water addition, the system prioritizes increasing the frequency of the inverter to 45Hz, and then uses an electric regulating valve to fine-tune the flow rate, forming a coordinated control strategy.

[0128] In the dynamic adjustment of the closed-loop feedback module, the closed-loop feedback module monitors the dynamic change trend of the parameters inside the salt tank in real time during the water addition process, and adjusts the control weights according to the parameter regression rate until the liquid level and concentration reach the preset equilibrium dead zone range.

[0129] The controller compares the instantaneous flow rate and the liquid level rise rate fed back by the electromagnetic flow meter in real time. If the ratio of the two is abnormal, the system automatically determines that there is leakage or sensor drift and triggers the compensation and correction logic.

[0130] When both the liquid level and concentration reach the equilibrium dead zone range of ±0.5% of the target value, the system outputs a hold command to maintain the current actuator state or enter the sleep monitoring mode.

[0131] The human-machine interface (HMI) and data communication module utilize a 15-inch industrial touchscreen with IP65 protection, displaying real-time data on the brine tank's level curve, concentration trend, valve status, and historical alarm records. The data communication module establishes a PROFINET network via an industrial Ethernet switch, interconnecting the controller, HMI, and host computer management system. This supports uploading process parameters to the production execution system, enabling remote monitoring and data aggregation.

[0132] See Figure 3 Based on the above system architecture, the workflow of an automatic water addition control method for a salt tank in this embodiment is as follows:

[0133] Step 1: Real-time data acquisition. A sensor array deployed inside the brine tank and on the inlet water pipeline synchronously acquires raw physical signals including solution level, conductivity, solution temperature, and real-time inlet water flow rate. Specifically, the solution level is measured using an ultrasonic level gauge, the concentration using a four-electrode conductivity sensor, the temperature using a platinum resistance temperature sensor, and the flow rate using an electromagnetic flow meter. The 4–20 mA standard current signals output by these sensors are converted into digital signals by an analog-to-digital converter and then transmitted to the controller. The raw current signals are then processed by a 16-bit analog-to-digital converter chip and converted into raw digital values ​​ranging from 0 to 65535.

[0134] Step 2: In the data preprocessing and fusion stage, the controller employs a composite filtering algorithm to remove random noise and outliers from the original physical signal. The composite filtering combines median filtering and weighted moving average filtering: a sliding window of length 10 is maintained. After sorting the data within the window, the maximum and minimum values ​​are removed, and the arithmetic mean of the remaining 8 points is calculated, effectively suppressing instantaneous liquid level jumps caused by stirring or material input within the salt tank. Subsequently, the sequence after median filtering is subjected to sliding window weighted averaging. The weighting coefficient decays exponentially with the time distance of the sampling points, ensuring signal smoothness while preserving dynamic response characteristics. A temperature compensation model is constructed based on the nonlinear functional relationship between solution conductivity and temperature. Since the conductivity of salt solutions increases with increasing temperature, the system uses a standard reference temperature of 25℃ as a benchmark and converts the conductivity measurements using real-time acquired solution temperatures. The compensation coefficient is preset according to the characteristics of different salt systems to ensure that the conversion error of salt concentration is within the preset allowable range within the preset operating temperature range. The compensated conductivity data is converted into real-time salt concentration percentage data through a concentration conversion table and output along with the standardized liquid level data.

[0135] Step 3: Calculate the decision control output. Compare the real-time salt concentration data and standardized liquid level data with the preset target concentration and target liquid level values, respectively, to calculate the concentration deviation and liquid level deviation (taking the larger absolute value of the two as the control deviation E). Using an improved PID algorithm combined with fuzzy logic reasoning, the control output is divided into intervals based on the deviation magnitude: When the deviation exceeds the first threshold (20%), the system enters an open-loop high-flow water filling mode, directly outputting saturation control parameters. The electric regulating valve opening is forcibly set to 95%–100%, and the variable frequency pump frequency is forcibly increased to 50Hz. Simultaneously, a timer is started. If the open-loop control continues for more than 120 seconds and the deviation still does not drop below the first threshold, the system determines an abnormal operating condition, automatically triggers an alarm, and switches to safety protection mode. When the deviation is between the first and second thresholds (5%–20%), the system activates the fuzzy logic controller. The input variables are the deviation E (quantized to 7 fuzzy levels, with a domain of discourse of [-20%, 20%]) and the deviation change rate. (Quantized to 5 fuzzy levels, with a universe of discourse ranging from -10% / s to 10% / s). The system stores 35 fuzzy rules, covering 7 levels of E. All combinations of the 5 levels. The PID coefficient correction is obtained by defuzzifying using the centroid method. , , After updating the PID coefficients, the control output is calculated.

[0136] When the deviation is less than the second threshold (5%), the system switches to fine adjustment mode. In this mode, strategies such as integral separation (integration is activated when the absolute value of the deviation is less than 2%, and the integral is frozen when it is greater than 2%), integral limiting (the output of the integral term is limited to ±15% of the control range), derivative prioritization (only the measured concentration value is differentially calculated to avoid shocks caused by sudden changes in the set value), and control dead zone (entering sleep monitoring state when the absolute value of the deviation is less than 0.5% for 30 seconds) are used to eliminate static errors and prevent overshoot.

[0137] Control output generation: Substitute the PID coefficients calculated in the above mode into the position PID control algorithm to calculate the control output u(t). After range conversion, the control output u(t) is converted into a 4-20mA standard signal and output to the execution drive module.

[0138] Step 4: The actuator is driven to add water. The controller generates a drive signal of corresponding strength according to the drive command, and adjusts the opening of the electric regulating valve on the water inlet pipe or the operating frequency of the variable frequency water pump through the analog output module. The electric regulating valve adopts an equal percentage flow characteristic valve core and has a position feedback function. The controller achieves closed-loop control of the valve position by comparing the command opening and the feedback opening. The variable frequency water pump receives the command frequency through the frequency converter to achieve stepless speed adjustment.

[0139] Step 5: Closed-loop feedback and verification. Real-time monitoring of the dynamic changes in parameters inside the brine tank during water addition, including liquid level, salt concentration, solution temperature, and inlet water flow rate. The controller compares the instantaneous flow rate and liquid level rise rate fed back by the electromagnetic flowmeter. If the ratio is abnormal, leakage or sensor drift is detected, triggering compensation and correction logic. When both the liquid level and concentration reach the equilibrium dead zone range of ±0.5% of the target value, the system outputs a hold command to maintain the current actuator state or enter a sleep monitoring state, completing one full automatic water addition control cycle.

[0140] Example 2: Salt Layer Thickness Extension Scheme

[0141] Based on Example 1, this example extends the system architecture and modifies the coupling of the method logic to address the impact of salt layer thickness under complex working conditions.

[0142] In terms of system architecture, a high-range pressure transmitter is added to the bottom of the salt tank in the sensor configuration step 1. This transmitter uses a ceramic capacitive sensor, which has strong overload resistance and corrosion resistance, with a range of 0–100 kPa. It is installed at the pressure tap on the bottom side wall of the salt tank. This pressure signal is used as an auxiliary parameter and connected to the analog input channel of the controller.

[0143] In terms of anti-interference capability, in order to improve the anti-interference capability of the system, all sensor signal cables are double-shielded twisted-pair cables, and the shielding layer is grounded at one point at the control cabinet end. The signal transmission is powered by 24V DC.

[0144] In terms of control methods, this embodiment introduces salt layer thickness estimation and density correction logic. In step 2, the controller acquires the liquid level height H from the ultrasonic level gauge while simultaneously reading the pressure value P from the bottom pressure transmitter. Based on the principle of physical hydrostatic balance and combined with the average density of the current solution, the system calculates the equivalent thickness of the undissolved solid salt layer in the salt tank.

[0145] Real-time solution density:

[0146]

[0147] Where P is the measured pressure (Pa) of the pressure transmitter. The zero-point offset pressure (Pa) is given, and g is the acceleration due to gravity (9.8). H represents the measured liquid level height (m) of the ultrasonic level gauge.

[0148] Equivalent thickness of solid salt layer:

[0149]

[0150] in, The density of the solid salt (taken as 2100). ), The calculated salt layer thickness (m) is given by ρ, where ρ is the real-time density of the solution. ), g is the acceleration due to gravity (9.8) H is the liquid level height (m). When When the depth is less than 0.02m, the system determines that the salt layer has been basically dissolved.

[0151] In terms of control optimization, feedforward compensation is added to the decision control module. In the instruction decision in step 3, the system incorporates the salt layer thickness as a feedforward compensation term into the control loop. When a thinning of the salt layer is detected, it indicates that the salt dissolution rate is accelerating. At this time, the system will automatically reduce the rising slope of the water addition target value to prevent the salt concentration from being rapidly diluted due to excessive water addition, which would affect the subsequent crystallization quality.

[0152] Specifically, the target liquid level rise slope is set as follows:

[0153]

[0154] in, The basic ascent slope (taken as 0.5 cm / min) This represents the upper limit of the salt layer thickness (taken as 0.5m). The calculated salt layer thickness (m).

[0155] To address the risk of crystallization and blockage in a high-salt environment, a self-cleaning mechanism is added to the actuator during the water addition process in step 4. Every 24 hours, the controller drives the electric regulating valve to perform a full open / close operation, using the flushing action of high-pressure water to remove accumulated salt from the valve seat. The self-cleaning process lasts 30 seconds, during which PID control is paused, and normal control mode is resumed after cleaning is complete.

[0156] Enhance anomaly monitoring and fault diagnosis: In the closed-loop feedback and verification of step 5, enhance the anomaly monitoring and fault diagnosis functions.

[0157] The system monitors the signal change rate of the ultrasonic level gauge in real time for abnormal level signal monitoring and redundancy switching. If the signal change exceeds 50% of the range within 1 second (3m for a 6m range level gauge), or the signal value drops below 3.8mA (below the normal 4mA lower limit), the system determines that the sensor signal is abnormal and immediately switches to the level estimation mode based on the pressure transmitter. The level estimation calculation formula is as follows:

[0158]

[0159] The meanings of the parameters are the same as above. Accuracy verification of the calculated liquid level: During normal operation of the ultrasonic level gauge, the system continuously compares the deviation between the measured liquid level and the calculated liquid level. If the deviation is consistently less than 0.05m, the calculation model is confirmed to be effective.

[0160] The system adds fault diagnosis and alarm for pipe bursts, and monitors the relationship between influent flow rate and liquid level changes in real time. If the influent flow rate is detected to be consistently higher than 150% of normal consumption for 30 seconds while the liquid level drops instead of rising, a pipe burst fault is immediately identified. The influent pump and electric valve are forcibly shut down, and an alarm SMS is sent to the remote monitoring center. Simultaneously, an emergency alarm window for "pipe burst fault" pops up on the human-machine interface, and the audible and visual alarm emits a continuous beeping sound and a flashing red signal.

[0161] To verify the synergistic effect of salt layer thickness compensation and fuzzy PID control, a comparative test was conducted in the salt dissolving section of a vacuum salt production line with an annual output of 300,000 tons. The effective volume of the test salt tank was 50 m³, and the target concentration was 26.5%. The test was divided into four stages, each lasting 7 days.

[0162] Test results show that after combining salt layer thickness compensation with fuzzy PID control, the concentration fluctuation is reduced to ±0.5%, significantly lower than the theoretical linear superposition value (approximately ±1.0%) of applying fuzzy PID alone (±1.5%) and salt layer compensation alone (±2.0%), proving that the two have a synergistic effect. Meanwhile, under disturbance conditions of sudden drops in salt layer thickness, the settling time of the proposed solution is shortened by 50% compared to traditional PID, and the overshoot is reduced by 56%.

[0163] Furthermore, the anomaly monitoring mechanism in this embodiment has been further strengthened. The system includes a sensor signal validity verification module. When the current signal of the ultrasonic level gauge undergoes a sudden change, such as a change exceeding 50% of the range within one second, or a drop in current value below 4 mA, the system automatically switches to the pressure transmitter-based level estimation mode, achieving redundant backup switching of the sensor. Simultaneously, the system integrates an audible and visual alarm. When it detects that the inlet water flow rate remains high for an extended period but the liquid level decreases instead of rising, it immediately identifies a pipe burst fault, forcibly shuts down the inlet pump and electric valve, and sends an alarm SMS to the remote monitoring center.

[0164] Example 3: Intelligent Prediction Scheme

[0165] This embodiment, based on Embodiment 1, further introduces big data analysis and predictive control technology. Compared with Embodiment 1, this embodiment differs in the control method as follows:

[0166] An edge computing gateway is added to the system, building upon the data acquisition in step 1. This gateway employs an industrial-grade ARM architecture processor, runs an embedded Linux operating system, and has 4 GB of RAM and 64 GB of solid-state storage. The gateway interacts with the PLC via an Ethernet interface and communicates with the cloud server using the MQTT protocol.

[0167] A concentration prediction model based on Long Short-Term Memory (LSTM) neural network was introduced. In the data preprocessing step 2, the edge computing gateway acquires real-time data on liquid level, concentration, temperature, flow rate, and external air temperature over the past 24 hours to construct a multi-dimensional time series dataset and run the LSTM prediction model.

[0168] The model structure is as follows:

[0169] Input layer: Constructing multidimensional time series input vectors

[0170]

[0171] Where L is the liquid level (m); C is the concentration (%); and T is the solution temperature (%). Q is the influent flow rate (); ); θ is the external temperature (°C); n is the time step, with a value of 60 (corresponding to 60 minutes of historical data).

[0172] Hidden layers: A two-layer LSTM structure is used, with 128 hidden units in the first layer and 64 hidden units in the second layer. The activation function is tanh, and overfitting is prevented by a Dropout layer (dropout rate of 0.2).

[0173] Output layer: Fully connected layer, outputting the concentration prediction sequence for the next 15 minutes, expressed as:

[0174]

[0175] Where C(t+k) represents the predicted concentration value in the kth minute; ; To output the weight matrix, Let represent the hidden state of the LSTM layer at time t. This is the bias vector.

[0176] The network structure described above was determined through grid search and cross-validation. Historical data from 30 consecutive days of salt tank operation (sampling period of 1 second, approximately 2.6 million samples) was used as the training set, divided into 80% training and 20% validation. Combinations of the number of hidden layers (1-3 layers), number of units per layer (32, 64, 128, 256), and Dropout rate (0.1, 0.2, 0.3, 0.5) were searched, with the root mean square error (RMSE) on the validation set used as the evaluation metric. The search results showed that a two-layer structure was superior to single-layer and three-layer structures; the combination of 128 units in the first layer and 64 units in the second layer achieved the lowest RMSE (0.15%) while maintaining a moderate number of model parameters (approximately 120,000) and an inference latency of <50ms, meeting real-time control requirements; a Dropout rate of 0.2 achieved the optimal balance between preventing overfitting and maintaining the model's expressive power.

[0177] In the decision calculation of step 3, the system no longer relies solely on the current deviation value, but adds feedforward control and anti-oscillation algorithms:

[0178] Feedforward control transforms the prediction bias of the LSTM model into a feedforward compensation term:

[0179]

[0180] in is the feedforward gain coefficient (taken as 0.8), and FF is the feedforward compensation amount superimposed on the PID control output.

[0181] = +FF

[0182] in The output is calculated by the PID control algorithm, and FF is the feedforward compensation. This is the final instruction sent to the execution driver module.

[0183] When the model predicts that the concentration will drop sharply within the time interval of the next 10 to 15 minutes, the controller immediately starts the water addition program after the prediction output, so that the water addition action occurs in advance of the concentration drop, in order to overcome the physical lag in the salt dissolution process.

[0184] The anti-oscillation algorithm is included in the actuator drive logic. It is implemented by monitoring the frequency of changes in the control quantity output to the actuator.

[0185] Rate of change of direction of motion:

[0186]

[0187] in, The length of the sliding window (taken as 10). This represents the number of times the valve's direction of movement changes within the window.

[0188] Dead zone dynamic adjustment:

[0189] DB=

[0190] in, The default dead zone is set to 0.5%, and k is an adjustment factor (set to 2.0). When At that time, the dead zone expands dynamically, but the maximum does not exceed 2.0%.

[0191] Smooth the output:

[0192]

[0193] in, This refers to the original control output of the PID algorithm. The output is the control output after smoothing, and λ is the filter coefficient, which is 0.3 during the dead zone expansion period and 1.0 under normal operating conditions.

[0194] The above parameters (window length) =10, dead zone adjustment coefficient k=2.0, maximum dead zone 2.0%, smoothing filter coefficient λ=0.3) were optimized through field experiments: different parameter combinations were tested under frequent valve vibration conditions, and the combination with the fewest valve actions and the smallest concentration fluctuation was taken as the final set value. After optimization, the average number of valve actions per day decreased from 320 times in the original system to 105 times, a reduction of 67%.

[0195] The execution driver in step 4 is the same as in Example 1, and will not be repeated here.

[0196] In the closed-loop feedback and verification step 5, parameter self-evolution and equipment health management are added:

[0197] Control parameters evolve online, and the edge computing gateway periodically evaluates the control effect of the PID parameters. At the end of each water filling cycle, the system calculates the integral absolute error (IAE) for that cycle.

[0198]

[0199] Where T is the duration of a water filling cycle, and E(t) is the deviation value at time t. The system records the IAE values ​​of the most recent 10 water filling cycles. If the IAE values ​​of 10 consecutive cycles show a monotonically increasing trend (i.e., the IAE of each cycle is greater than the IAE of the previous cycle), it indicates that the control effect of the current PID parameters is continuously deteriorating. At this time, the edge computing gateway automatically starts the parameter optimization algorithm, using a genetic algorithm to search for a better combination of PID coefficients in the parameter space. The search range is... , , After optimization, the gateway sends the new PID coefficients to the PLC for use in subsequent control cycles.

[0200] Equipment health management involves the system monitoring the operating parameters of the variable frequency water pump in real time to determine its health status. The controller records the inverter's output current every 10 seconds. and operating frequency The curve was then compared with a baseline curve. The baseline curve represents the current-frequency relationship of the water pump under rated operating conditions, obtained through factory testing. The fitting formula is as follows:

[0201]

[0202] in, and The fitting coefficient is determined from the factory test data of the water pump.

[0203] When the measured current With reference current When the deviation exceeds 15% and lasts for more than 5 minutes, the system determines that the water pump has abnormal risks such as impeller wear, bearing overheating, or rotor stall. At this time, a "Water Pump Health Warning" prompt box pops up on the human-machine interface, displaying the current current deviation value and suggested maintenance measures, reminding maintenance personnel to perform preventive checks. Simultaneously, the system uploads this warning information to the Production Execution System (MES) via the data communication module and incorporates it into the equipment maintenance management plan.

[0204] To verify the synergistic effect of LSTM feedforward prediction and fuzzy PID control, a load change response test was conducted in this embodiment:

[0205] Under a sudden 50% increase in production line load, the control effects of four schemes were tested: traditional feedback control, standalone LSTM feedforward control, standalone fuzzy PID control, and the "LSTM + fuzzy PID + salt compensation" scheme of this invention. The test results are as follows:

[0206] Test results show that the maximum concentration deviation of the proposed solution is only -0.6%, and the recovery time is shortened to 90 seconds, both significantly better than the expected linear superposition of individual technologies (theoretical maximum deviation -1.0%, theoretical recovery time 120 seconds). This synergistic effect stems from the LSTM predicting the load change trend in advance, the fuzzy PID pre-adjusting the control parameters based on the prediction results, and the salt layer thickness compensation synchronously correcting the salt dissolution rate prediction. The three form a complete closed loop of "prediction-decision-compensation".

[0207] Example 4: Salt System Adaptation Scheme

[0208] This embodiment, based on Embodiment 1, extends the design to accommodate the differences in physicochemical properties of different salt systems. Compared to Embodiment 1, this embodiment differs in the control method as follows:

[0209] In the temperature compensation model of step 2, due to the significant differences in the conductivity-temperature characteristics of different salts in aqueous solutions, the system has multiple sets of compensation parameters built-in. After the operator selects the type of salt currently being produced through the human-machine interface, the controller automatically calls up the corresponding compensation coefficient.

[0210] Linear compensation model (applicable to systems containing sodium chloride, potassium chloride, etc.):

[0211]

[0212] in, To convert to the standard conductivity value at 25℃, This is the measured conductivity value. The actual temperature (°C) is the measured temperature. The temperature compensation coefficient was determined based on a large sample of salt dissolution experimental data. Within the temperature range of 10℃ to 60℃, standard saline solutions with concentration gradients of 5%, 10%, 15%, 20%, 25%, and 30% were prepared at 5℃ intervals. The conductivity values ​​at each temperature were measured using a high-precision laboratory conductivity meter, yielding a total of 180 data points. The least squares method was used for fitting, resulting in an α=0.021 goodness-of-fit coefficient (R²=0.998) for the sodium chloride system and an α=0.020 goodness-of-fit coefficient (R²=0.998) for the potassium chloride system. =0.997.

[0213] In this embodiment, the compensation coefficients for each salt system are set as follows:

[0214] Sodium chloride system: α = 0.021 (2.1% / ℃)

[0215] Potassium chloride system: α = 0.020 (2.0% / ℃)

[0216] Nonlinear compensation model (applicable to systems with phase transition points, such as sodium sulfate):

[0217]

[0218] Among them, the compensation coefficient The values ​​are segmented according to the temperature range: when t ≥ 32.4℃, α(t) = 0.018; when t < 32.4℃, α(t) = 0.016. This segmented design is based on experimental determination of the phase transition characteristics of sodium sulfate at 32.4℃. Experiments show that using a single compensation coefficient, the concentration conversion error near the phase transition point can reach ±0.8%, while segmented compensation controls the error within ±0.3%.

[0219] The systematized preset of control parameters is implemented in step 3, where different salts exhibit varying sensitivities to concentration stability during production. The system pre-sets differentiated initial PID parameters and target value fluctuation tolerances for different salt systems.

[0220] Operators can also customize target values ​​and control parameters for specific production processes. The system supports the storage and one-click recall of multiple sets of formula parameters.

[0221] The execution driver in step 4 is the same as in Example 1, and will not be repeated here.

[0222] The adaptive transition during salt switching, in the closed-loop feedback and verification in step 5, involves the system automatically executing the following transition process when the production line switches between different salts:

[0223] Parameter loading: After the operator selects the target salt type through the human-machine interface, the system automatically loads the corresponding temperature compensation coefficient, PID initial parameters, and target value tolerance.

[0224] Residual rinsing: If there is a risk of cross-contamination between the salt systems before and after the switch (e.g., switching from sodium chloride to sodium sulfate), the system will automatically trigger a rinsing procedure. The rinsing process lasts for 5 minutes, with the influent volume set to 150% of the normal fill volume, to clean the salt tank and pipelines. PID control is paused during rinsing, and the solution in the salt tank is drained to the preset level after rinsing is complete.

[0225] After sensor zero-point calibration and rinsing, the system refills the brine tank with process water to the preset baseline level and initiates the salt dissolution cycle. During salt dissolution, the system monitors the concentration change rate in real time. Once the concentration change rate is less than 0.1% / min for three consecutive sampling cycles, the system determines that the solution has reached a stable state. At this point, the system records the current reading of the conductivity sensor and compares it with the theoretical concentration value calculated based on the influent flow rate and salt dissolution volume, correcting the zero-point offset of the conductivity-concentration conversion table to ensure the accuracy of subsequent measurements. After calibration, normal control circulation begins.

[0226] Example 5: Complete Security Protection Mechanism Solution

[0227] This embodiment, based on Embodiment 1, constructs a multi-layered, hardware-software collaborative security protection system. Compared to Embodiment 1, this embodiment differs in the control method as follows:

[0228] Sensor signal integrity monitoring

[0229] During the data acquisition process in step 1, the system monitors the signal integrity of the sensor array in real time:

[0230] For normal range determination, the theoretical effective range of all 4–20mA standard current signals is 3.8–20.5mA (considering the transmitter range margin). The controller samples and verifies each analog channel at a period of 100ms.

[0231] For open circuit fault detection, if the current signal of any sensor remains below 3.5mA for 500ms, the system determines that the sensor is experiencing an open circuit fault.

[0232] For short-circuit fault detection, if the current signal of any sensor is higher than 20.8mA for 500ms, the system determines that the sensor is short-circuited.

[0233] Fault Response: Once the above fault is detected, the system will immediately perform the following operations:

[0234] Automatically switch to manual safety mode to disable automatic water filling command;

[0235] An alarm window pops up on the human-machine interface, displaying the name and location of the faulty sensor;

[0236] The audible and visual alarm emits a continuous buzzing sound and a red flashing signal;

[0237] If the faulty sensor is a conductivity sensor or a level gauge, the system will suspend the automatic control process and wait for manual intervention.

[0238] Steps 2-4 are the same as in Example 1, and will not be repeated here.

[0239] In the closed-loop feedback and verification in step 5, an independent mechanical high-level safety device is added. The salt tank is equipped with a mechanical high-level float switch independent of the control system to achieve hardware-level safety redundancy protection.

[0240] Float switch selection: Select a corrosion-resistant side-mounted float switch with a float material of polypropylene (PP), a housing material of 316L stainless steel, a protection rating of IP68, and a contact capacity of 220VAC / 5A.

[0241] Installation location: The float switch is installed on the top side wall of the brine tank, with the trigger level set to 95% of the safe height of the brine tank (i.e., 50mm below the overflow port). During installation, ensure that the float moves freely without obstruction and maintain a safe distance from the stirring device.

[0242] Electrical connection: The normally closed contact of the float switch is directly connected in series with the power supply circuit of the water inlet pump and the 24V control circuit of the electric regulating valve, without going through the programmable logic controller (PLC) or any intermediate relays, ensuring independence.

[0243] Triggering logic: Under normal operating conditions, the normally closed contacts of the float switch remain closed. When the liquid level rises to the trigger height, the float rises with the liquid level, the mechanical mechanism drives the contacts to open, and simultaneously cuts off the power supply to the inlet pump and the control power supply to the electric regulating valve, forcibly stopping the water intake.

[0244] Manual reset: After troubleshooting, the operator must wait for the liquid level to drop below the safe height and then manually restore the power supply circuit using the on-site reset button before the system can be restarted.

[0245] The dual security protection system of "software monitoring + hardware redundancy" constructed in this embodiment produced unexpected synergistic effects in actual operation:

[0246] When using software monitoring alone (sensor signal integrity monitoring + pipe burst fault diagnosis), the fault identification accuracy is 95%, with an average response time of 3 seconds. When using hardware redundancy alone (mechanical float switch + independent power supply circuit), the fault identification accuracy is 99%, but it cannot provide fault type diagnosis. Combining the two methods achieves a fault identification accuracy of 99.5%, while also providing fault type diagnosis (distinguishing between sensor disconnection, pipe burst, and excessive liquid level), with the average response time reduced to 0.5 seconds. This synergistic effect stems from the complementarity of software monitoring and hardware redundancy: software monitoring provides early warning and fault diagnosis, while hardware redundancy provides ultimate safety assurance. The combination of the two achieves full-chain safety protection from "post-event disconnection" to "pre-event warning + in-event disconnection + post-event diagnosis," significantly improving the inherent safety level of the system.

[0247] Example 6: Periodic Water Addition Control Method in Quenching Scenarios

[0248] This embodiment provides an automatic water addition control method applicable to the quenching process in which a salt bath is used. The difference from the previous embodiment is that the brine in the salt bath is used to cool high-temperature metal workpieces. During the quenching process, the water evaporates violently, causing the salt concentration to rise and the liquid level to drop. The control objective is to maintain a stable water content.

[0249] The control method in this embodiment uses the water content of the liquid salt as the control target and employs periodic, scenario-specific conditional judgment rules to calculate the water addition amount within each time period. The specific steps are as follows:

[0250] Step 1: Parameter Initialization

[0251] Set the following control parameters:

[0252] Target water content: The percentage of liquid salt water content to be maintained;

[0253] Cyclic water replenishment constant: The basic water replenishment volume within each time period under normal operating conditions;

[0254] Maximum water addition per cycle: The maximum amount of water allowed to be added within a single time cycle;

[0255] Water addition coefficient per unit weight: The amount of water evaporation compensation corresponding to a unit mass of workpiece during a single quenching process;

[0256] Total weight of workpiece in a single quenching cycle: The total mass of workpiece immersed in the salt bath during a single quenching cycle.

[0257] Step 2: Status Determination

[0258] Determine whether the salt bath is currently in a quenching state and whether it is in a moisture content correction state. Specifically:

[0259] The quenching state refers to the process in which a high-temperature metal workpiece is being immersed in a salt bath for cooling.

[0260] Moisture content correction status refers to a state where the deviation between the current moisture content and the target moisture content exceeds a preset threshold, requiring active water replenishment to eliminate the cumulative deviation.

[0261] Step 3: Calculate the water addition volume for different scenarios.

[0262] Based on different combinations of quenching and correction states, the amount of water added during the current time period is calculated according to the following four scenarios:

[0263] Scenario 1: Neither quenched nor corrected

[0264] When the salt bath is not undergoing quenching and moisture content correction:

[0265] If the periodic water addition constant is greater than or equal to the periodic maximum water addition, then the water addition for this period shall be the periodic maximum water addition.

[0266] If the periodic water addition constant is between 0 and the maximum periodic water addition, then the water addition for this period is taken as the periodic water addition constant.

[0267] If the periodic water addition constant is less than or equal to 0, then the water addition for this period is taken as 0.

[0268] Scenario 2: Quenching without correction

[0269] When the salt bath is in the quenching process but not in the moisture content correction process:

[0270] The amount of water added during this period is equal to the sum of the periodic water addition constant and the single quenching evaporation compensation amount;

[0271] Among them, the evaporation compensation amount for a single quenching is equal to the water addition coefficient per unit weight multiplied by the total weight of the workpiece in a single quenching.

[0272] Scenario 3: Non-quenched and corrected

[0273] When the salt bath is not undergoing quenching but is in the process of moisture content correction:

[0274] First, calculate the total amount of water that needs to be added. The value is equal to the difference between the target moisture content and the current moisture content, divided by 100, and then multiplied by the current liquid salt volume in the salt tank.

[0275] Then, the sum of the total amount of water to be added and the constant amount of water added in the cycle is allocated to each subsequent time cycle, with the maximum amount of water added in the cycle as the upper limit for a single allocation, until the total amount of water to be added is allocated.

[0276] After the water distribution is completed, the system automatically exits the calibration state and resumes normal water filling mode.

[0277] Scenario 4: Quenching and Correction

[0278] When the salt bath is being quenched and undergoing moisture content correction:

[0279] First, calculate the total amount of water that needs to be added according to the method described in Scenario 3;

[0280] Then calculate the allocation base, which is equal to the sum of the total amount of water to be added, the periodic water addition constant, and the single quenching evaporation compensation amount;

[0281] Using the maximum water addition in a cycle as the upper limit for a single allocation, the allocation base is sequentially distributed to subsequent time cycles until the total amount of water to be added is allocated.

[0282] Step 4: Add water

[0283] Based on the water addition amount calculated in step 3, the actuator is driven to complete the water addition within each time cycle. The actuator includes an electric regulating valve or a variable frequency water pump.

[0284] Step 5: Status Exit

[0285] When the total amount of water to be added has been allocated, and the deviation between the current water content and the target water content is less than the preset dead zone, the system exits the calibration state and resumes the normal water addition mode of Scenario 1 or Scenario 2.

[0286] In this embodiment, parameters such as the periodic water addition constant, the maximum periodic water addition, and the water addition coefficient per unit weight are set and modified through a human-machine interface. The quenching state can be obtained by detecting the workpiece immersion signal or by manual triggering by the operator.

[0287] The control method provided in this embodiment is particularly suitable for processes where the salt bath also serves as the quenching medium, effectively compensating for moisture evaporation loss during quenching and maintaining stable moisture content. Compared with the previous embodiments, this embodiment uses discrete periodic condition judgment rules, which are simple to implement, require less computation, and are suitable for quenching scenarios where high control precision is not required.

[0288] Example 7: Automatic water replenishment control of isothermal quenching nitrate bath

[0289] This embodiment provides an automatic water replenishment control method suitable for isothermal quenching nitrate baths. The difference from the previous embodiment is that the salt bath contains molten nitrate containing water, the working temperature is 150℃~400℃, and the control target is to accurately maintain the trace water content in the molten liquid, usually 0.5%~4.0%, in order to stabilize the quenching cooling rate.

[0290] This embodiment uses the same system architecture as Embodiment 1: a perception module, a data preprocessing module, a decision control module, an execution drive module, a closed-loop feedback module, and a bidirectional coupling architecture. Only some parameters and control logic are adaptively adjusted, as follows:

[0291] Step 1: Real-time data acquisition. Raw physical signals are synchronously acquired through a sensor array deployed inside the nitrate bath and on the inlet pipe. The sensor array includes:

[0292] Moisture content detection: A high-frequency impedance spectroscopy sensor is used to indirectly estimate the moisture content by measuring the electrical properties of the molten nitrate. Because the dielectric constant of water (approximately 80) differs significantly from that of molten nitrate (approximately 10~20), changes in moisture content will cause a shift in the characteristic frequency of the impedance spectrum. The sensor electrodes are made of high-temperature resistant and corrosion-resistant chromium-nickel-iron alloy material, with an operating temperature range covering 150℃~400℃.

[0293] Liquid level monitoring: High-temperature guided wave radar level gauge is used, with the waveguide rod made of 316L stainless steel and Teflon coating that is heat resistant to 260℃ (or ceramic coating that is heat resistant to 400℃). The measurement accuracy is ±2mm.

[0294] Temperature monitoring: A type K thermocouple (nickel-chromium-nickel-silicon) is used, with a temperature range of 0~600℃ and an accuracy of ±1.5℃. It is encapsulated in a high-temperature resistant stainless steel protective tube and installed at different depths on the side wall of the salt tank.

[0295] Water supply flow monitoring: A high-temperature electromagnetic flow meter or vortex flow meter is used and installed in the water supply pipeline. The measurement range is 0~100L / h and the accuracy is ±0.5%.

[0296] The 4~20mA standard current signal output by the above sensor is converted into a digital signal by the analog signal acquisition module and then transmitted to the controller.

[0297] Step 2: Data preprocessing and fusion. The controller employs a composite filtering algorithm to remove random noise and outliers from the original physical signal. For the high-temperature, highly corrosive environment of the nitrate bath, the composite filtering algorithm combines median filtering and exponentially weighted moving average filtering: the median filtering window length N=10 effectively suppresses instantaneous level jumps caused by molten salt flow or agitation; the exponentially weighted moving average filtering has a smoothing factor λ=0.2 and weighting coefficients...

[0298]

[0299] Where λ is the smoothing factor, and in this embodiment, λ=0.2 is taken; i is the time distance between sampling points (i=0 represents the current time, i=1 represents the previous time, and so on) to ensure signal smoothness while retaining dynamic response characteristics.

[0300] A temperature compensation model is constructed based on the nonlinear functional relationship between water content, conductivity, and temperature of nitrate melt. Since the conductivity of nitrate melt increases with temperature, the system uses the target process temperature (e.g., 200℃) as a reference, and converts the conductivity measurements using real-time collected melt temperatures. The compensation coefficient is determined through large-sample experiments: nitrate samples with different water contents are prepared at 10℃ intervals within the temperature range of 150℃ to 400℃. The conductivity values ​​at each temperature are measured using a high-frequency impedance spectrometer. The compensation coefficient α = 0.015 (i.e., 1.5% / ℃) is determined by least-squares fitting, ensuring that the water content conversion error is ≤ ±0.1% within the operating temperature range.

[0301] The compensated conductivity data is converted into real-time water content percentage data through a water content conversion table (pre-stored in the controller), and output to the decision control module along with the standardized liquid level data.

[0302] Step 3: The decision control module compares the real-time water content data and standardized liquid level data with the preset target water content and target liquid level values, respectively, and calculates the water content deviation and liquid level deviation. This embodiment adopts a full closed-loop control strategy, without setting an open-loop high-flow water replenishment mode. All water replenishment commands are calculated through PID + fuzzy logic to ensure that the water replenishment volume is always within a controllable range and to avoid molten salt splashing caused by instantaneous large-volume water replenishment.

[0303] The control strategy employs an improved PID algorithm with parameter self-tuning capabilities combined with fuzzy logic inference, dividing the control interval based on the magnitude of the deviation:

[0304] Fuzzy PID control range: When the moisture content deviation or liquid level deviation exceeds a preset threshold, such as ±10% of the target moisture content or ±10mm of the liquid level deviation, the system activates the fuzzy logic controller. The input variables are the moisture content deviation E (quantized to 7 fuzzy levels, with a domain of discourse of [-2%, +2%]) and the rate of change of deviation. (Quantized to 5 fuzzy levels, with a universe of discourse range of [-0.5% / s, +0.5% / s]). The system stores 35 fuzzy rules, and the PID coefficient correction is obtained by defuzzifying using the centroid method.

[0305]

[0306]

[0307]

[0308] in, , , The initial PID coefficients are set to 1.0, 0.06, and 0.5 respectively, taking into account the response characteristics of the nitrate bath. These coefficients are then substituted into the positional PID control algorithm to calculate the control output u(t), which is then converted into a 4~20mA standard signal and output to the actuator module after range conversion.

[0309] Fine-tuning mode: When the deviation is less than the preset dead zone, such as ±0.2% of the target moisture content and ±3mm of the liquid level deviation, the system switches to fine-tuning mode. This mode employs strategies such as integral separation: integration is activated when the absolute value of the deviation is less than 0.1%, and frozen when it is greater than 0.1%; integral limiting: the integral term output is limited to ±15% of the control range; derivative priority: only the measured moisture content value is differentiated to avoid shocks caused by sudden changes in the setpoint; and control dead zone: when the absolute value of the deviation is less than 0.05% for 30 seconds, the system enters a dormant monitoring state to eliminate static errors and prevent overshoot.

[0310] Water replenishment regulation principle: When the detected water content is lower than the target value, the controller linearly adjusts the opening of the electric regulating valve (20%~70%) and the frequency of the water replenishment pump according to the magnitude of the deviation. The larger the deviation, the stronger the water replenishment command, but the maximum opening does not exceed 70%, and the maximum frequency does not exceed 45Hz, ensuring that the water replenishment volume is always within a safe range. The system monitors the rate of change of water content after water replenishment in real time. If the rate of change is abnormal (e.g., the water replenishment volume is normal but the water content does not increase significantly), the control parameters are automatically adjusted and the sensor verification process is triggered.

[0311] Mapping relationship between moisture content and cooling rate: The system has a built-in mapping model between the moisture content of the nitrate melt and the quenching cooling rate. Through pre-experimental determination of cooling curves at different moisture contents, a database of the relationship between moisture content, maximum cooling rate, and cooling time at 300℃ is established. When the current moisture content deviates from the target value, the controller automatically adjusts the water replenishment command to restore the moisture content to the target range, thereby ensuring the stability of the quenching cooling rate. For example, for the isothermal quenching process of GCr15 bearing steel, the target moisture content is set to 1.2%~1.8%, corresponding to a cooling time of 15~20 seconds at 300℃.

[0312] Step 4: Drive the actuator to replenish water. The controller generates a drive signal of corresponding strength according to the drive command, and adjusts the opening of the high-temperature electric regulating valve on the water replenishment pipeline or the operating frequency of the water replenishment pump through the analog output module.

[0313] High-temperature electric control valve: Utilizing a high-temperature and corrosion-resistant Hastelloy valve body, the valve core features equal percentage flow characteristics. Equipped with an intelligent actuator, it supports 4~20mA signal adjustment and has position feedback functionality. The controller achieves closed-loop control of the valve position by comparing the commanded opening degree with the feedback opening degree, achieving a positioning accuracy of ±0.5%.

[0314] Water replenishment safety protection: Due to the high operating temperature of the nitrate bath, an immersion-type water supply pipe must be used during water replenishment. The water pipe outlet should be extended 100-200mm below the molten salt surface to utilize the pressure of the molten salt and prevent instantaneous vaporization and splashing. A check valve is installed at the end of the water replenishment pipe to prevent molten salt backflow. Before issuing the water replenishment command, the controller first confirms that the molten salt temperature is below the safety threshold (e.g., ≤450℃) and that the agitator is running to ensure safe water replenishment.

[0315] Step 5: Closed-loop feedback and verification, real-time monitoring of the dynamic changes in parameters inside the nitrate bath during water replenishment. The controller compares the correlation between the instantaneous water replenishment volume, the rate of liquid level rise, and the rate of change in water content fed back by the electromagnetic flowmeter. If the ratio of these three factors is abnormal (e.g., the water replenishment volume is normal but the water content does not increase significantly), the system determines that there is excessively rapid water evaporation or sensor drift, triggers compensation and correction logic, and automatically adjusts the water replenishment strategy.

[0316] When the water content and liquid level both reach the target value within the balance dead zone range of ±0.1% and ±2mm respectively, the system outputs a hold command to maintain the current actuator state or enter the sleep monitoring state, completing a complete automatic water replenishment control cycle.

[0317] Based on Example 5, this embodiment adds the following safety protection measures to address the high-temperature characteristics of the nitrate bath:

[0318] Over-temperature protection: When the temperature of the nitrate bath exceeds the preset safety threshold (e.g., 450℃), the system automatically blocks the water replenishment command and issues an alarm to prevent molten salt from splashing due to high-temperature water replenishment.

[0319] Stirring Confirmation: Before executing the water replenishment command, the controller must confirm that the stirrer is running and the speed is normal (judged by the frequency converter feedback signal) to ensure that the added water can be quickly and evenly dispersed.

[0320] Manual reset: After any safety protection is triggered, the system must be manually reset by the operator before the automatic water replenishment control can be restarted.

[0321] This embodiment was applied to the isothermal quenching production line of a bearing manufacturing company. The original manual water replenishment method had a water content fluctuation of ±0.5% and a control cycle of 30 minutes. After upgrading to this automatic control system, the water content fluctuation was reduced to ±0.1%, the control cycle was shortened to 5 seconds, and the uniformity of hardness of the quenched workpiece (the difference within the same batch) was reduced from HRC 3.5 to HRC 1.2, which significantly improved the consistency of product quality.

[0322] In summary, this invention achieves a high degree of automation and intelligence in the brine tank water addition process by constructing a complete system encompassing multi-dimensional perception, intelligent decision-making, precise execution, and closed-loop verification. By introducing advanced technologies such as temperature compensation, salt layer thickness correction, and deep learning prediction, this invention overcomes the shortcomings of traditional control technologies in terms of dynamic response, parameter coupling, and long-term stability. Experimental results show that the control accuracy of the brine tank concentration is improved from ±3% to within ±0.5%, the liquid level fluctuation range is reduced by 60%, and the energy consumption of the inlet pump is reduced by approximately 15%.

[0323] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automatic control system for adding water to a salt tank, characterized in that, include: The sensing module is used to synchronously acquire raw physical signals through a sensor array deployed inside the salt tank and on the inlet pipe. The sensor array includes an ultrasonic level gauge, a conductivity sensor, a temperature sensor, a flow meter, and a pressure transmitter installed at the bottom of the salt tank. The pressure transmitter is used to acquire the total pressure at the bottom of the tank, which is combined with the liquid level height acquired by the ultrasonic level gauge to calculate the real-time density of the solution, and then to estimate the thickness of the undissolved solid salt layer in the salt tank. The data preprocessing module is used to remove noise and outliers from the original physical signal using a composite filtering algorithm, and to recalculate and correct the conductivity measurement based on a temperature compensation model to obtain real-time salt concentration data and standardized liquid level data. The decision control module is used to calculate the deviation based on real-time salt concentration data and standardized liquid level data, and to calculate the control output using a PID control algorithm with parameter self-tuning function combined with fuzzy logic reasoning. The decision control module also dynamically adjusts the target value of water addition and the salt dissolution rate prediction based on the calculated salt layer thickness. When the salt layer thickness is lower than the preset threshold, the rising slope of the target value of water addition is automatically reduced. The execution drive module is used to adjust the opening degree of the electric regulating valve on the water inlet pipeline or the operating frequency of the variable frequency water pump according to the drive command. The closed-loop feedback module is used to monitor the dynamic change trend of parameters inside the salt tank in real time during the water addition process, and adjust the control weights according to the parameter regression rate until the liquid level and concentration reach the preset equilibrium dead zone range. The data preprocessing module and the decision control module are connected by a bidirectional coupling architecture: the signal characteristic parameters output by the data preprocessing module are fed back to the decision control module in real time for dynamic adjustment of control parameters, and the control output quantity output by the decision control module is fed back to the data preprocessing module synchronously for dynamic adjustment of filtering parameters.

2. The automatic water addition control system for the salt tank according to claim 1, characterized in that, In the workflow of the data preprocessing module, the composite filtering algorithm and the temperature compensation model are run through a controller. The controller is an industrial-grade programmable logic controller, equipped with a high-speed processor, program storage space, and data storage space. The controller cabinet is made of stainless steel and has a double-layer sealed structure, and is equipped with a temperature-controlled fan, a dehumidifier heater, and a surge protector. The software architecture inside the controller adopts a layered design, with the data acquisition module running in a timed interrupt, the control algorithm module running in a real-time task, and the communication module running in a non-real-time task. The automatic water addition control system for the salt tank also includes a human-machine interface and a data communication module. The human-machine interface adopts an industrial-grade touch screen to display the liquid level curve, concentration trend, valve status, and historical alarm records of the salt tank in real time. The data communication module supports industrial network protocols to upload process parameters to the production execution system. The automatic water supply control system for the salt tank also includes an edge computing gateway for running a long short-term memory neural network prediction model.

3. The automatic water addition control system for the salt tank according to claim 1, characterized in that, The automatic water supply control system for the salt tank also includes an anomaly monitoring and safety protection module, which monitors the signal integrity of the sensor array in real time. When the detected signal value exceeds the preset normal current range threshold, it is determined to be a sensor malfunction and automatically switches to manual safety mode. When the ultrasonic level gauge signal is abnormal, it automatically switches to the level calculation mode based on the pressure transmitter. The salt tank is also equipped with a mechanical high-level float switch independent of the control system. When the level exceeds the safety limit height, it forcibly cuts off the water supply.

4. An automatic control method for the amount of water added to a salt tank, applied in the automatic control system for the amount of water added to a salt tank as described in any one of claims 1-3, characterized in that, Includes the following steps: Step 1: Acquire raw physical signals synchronously through a sensor array deployed inside the salt tank and on the inlet pipe; the sensor array includes an ultrasonic level gauge, a conductivity sensor, a temperature sensor, a flow meter, and a pressure transmitter installed at the bottom of the salt tank; the total pressure at the bottom of the tank is obtained through the pressure transmitter, and combined with the liquid level height obtained by the ultrasonic level gauge to calculate the real-time density of the solution, thereby estimating the thickness of the undissolved solid salt layer in the salt tank; Step 2: Use a composite filtering algorithm to remove noise and outliers from the original physical signal, and adjust the conductivity measurement based on the temperature compensation model to obtain real-time salt concentration data and standardized liquid level data; Step 3: Compare the real-time salt concentration data and standardized liquid level data with the target value to calculate the deviation. Use a PID control algorithm with parameter self-tuning function combined with fuzzy logic reasoning to calculate the control output. Based on the salt layer thickness calculated in Step 1, dynamically adjust the target value of water addition and the salt dissolution rate prediction. When the salt layer thickness is lower than the preset threshold, automatically reduce the rising slope of the target value of water addition. Step 4: Drive the actuator to perform precise water addition, and adjust the opening of the electric regulating valve on the water inlet pipe or the operating frequency of the variable frequency water pump according to the drive command; Step 5: Monitor the dynamic change trend of parameters inside the salt tank in real time during the water addition process, and adjust the control weights according to the parameter regression rate until the liquid level and concentration reach the preset equilibrium dead zone range. In step 2, the signal characteristic parameters output by the data preprocessing module are fed back to step 3 in real time for dynamic adjustment of control parameters, and the control output quantity output by step 3 is fed back to step 2 for dynamic adjustment of filtering parameters, forming a two-way coupled architecture.

5. The automatic control method for water addition to the salt tank according to claim 4, characterized in that, In step 1, the ultrasonic level gauge is installed on a measuring bracket at the top of the salt tank. The measuring range of the ultrasonic level gauge is adapted to the depth of the salt tank, and the output signal is a standard current signal. The conductivity sensor is a four-electrode conductivity sensor with graphite electrodes that have anti-fouling properties. The measurement range covers a preset conductivity range and is suitable for long-term continuous monitoring in high-salinity environments. The temperature sensor is a platinum resistance temperature sensor, which uses a three-wire connection method to eliminate measurement errors caused by wire resistance. The flow meter is an electromagnetic flow meter, installed on the main inlet pipeline. It measures the flow rate by sensing the induced electromotive force generated when the conductive medium flows through a magnetic field. The electromagnetic flow meter does not have any flow-obstructing components inside.

6. The automatic control method for water addition to the salt tank according to claim 4, characterized in that, In step 2, the composite filtering algorithm combines median filtering and weighted moving average filtering. The composite filtering algorithm and the temperature compensation model are run by a controller. The controller sorts a preset number of continuously collected sampling points and extracts the median value, eliminating instantaneous liquid level jumps caused by stirring or material input inside the salt tank, forming a preliminary processing sequence. Subsequently, the preliminary processing sequence is subjected to sliding window weighted averaging, where the weight coefficient decays exponentially with the time distance of the sampling points, preserving dynamic response characteristics while ensuring signal smoothness. The temperature compensation model uses a preset reference temperature as a standard and converts the conductivity measurement value using the real-time collected solution temperature. The compensation coefficient is preset according to the characteristics of different salt systems to ensure that the conversion error of salt concentration is within a preset allowable range within the preset operating temperature range.

7. The automatic control method for water addition to the salt tank according to claim 4, characterized in that, In step 3, the deviation includes a concentration deviation term and a liquid level deviation term. The interval adjustment logic of the PID control algorithm is as follows: when the concentration deviation term or the liquid level deviation term is greater than a preset first threshold, the system adopts an open-loop high-flow water addition mode, driving the actuator to fully open to reduce the deviation; when the concentration deviation term or the liquid level deviation term is between the first threshold and the second threshold, the system activates the fuzzy logic controller, and retrieves PID coefficients from the preset fuzzy rule base according to the concentration deviation term or the liquid level deviation term and the deviation change rate; when the concentration deviation term or the liquid level deviation term is less than the second threshold, the system switches to fine adjustment mode, eliminating static error through the cumulative effect of the integral term.

8. The automatic control method for water addition to the salt tank according to claim 4, characterized in that, The automatic control method for adding water to the salt tank also includes a feedforward control step: the system automatically records the concentration recovery time after each water addition cycle and uses a long short-term memory neural network to model the dissolution load of the salt tank; based on the historical preset cycle operation data, the system predicts the concentration change trend in the future predetermined time period, and intervenes in adding water in advance when the concentration decline trend is initially apparent, and converts the prediction result into a feedforward compensation term and introduces it into the control loop.

9. The automatic control method for water addition to the salt tank according to claim 4, characterized in that, The actuator's drive logic includes an anti-oscillation algorithm. The controller monitors the direction of change of the control quantity output to the actuator. When the number of reversals of the control quantity's value within a unit time exceeds a preset threshold, the controller automatically increases the control dead zone and smooths the control quantity to prevent frequent valve jitter. The electric regulating valve has a percentage flow characteristic. The electric regulating valve actuator includes a position feedback potentiometer, which feeds back the actual valve opening signal to the controller in real time. The controller achieves closed-loop control of the valve position by comparing the commanded opening with the feedback opening. The variable frequency pump receives a preset voltage signal through a frequency converter to achieve stepless speed adjustment within a preset speed range.

Images