Intelligent regulation method for low-temperature evaporation process of industrial waste liquid

Abstract

The application relates to the technical field of industrial process automation control, and discloses an intelligent adjustment method for an industrial waste liquid low-temperature evaporation process, which comprises the following steps: superimposing a pressure step test excitation on a vacuum adjustment valve and synchronously collecting a temperature response data stream, calculating an absolute error integral of a response curve, SIAE and a recovery time characteristic, and substituting the absolute error integral, the SIAE and the recovery time characteristic into an inverse model to solve an equivalent process gain and a time constant of a controlled object; based on an internal model control principle, PID controller parameters are dynamically calculated and updated by using the solved gain and time constant, a control law is reconstructed to generate an execution instruction, through active energy characteristic identification and a pole configuration mechanism, the application solves a control divergence problem of a nonlinear time-varying object under model mismatch, realizes decoupling control and self-adaptive adjustment of fluid viscosity mutation and equipment fouling, and improves the stability and control precision of a system.

Description

TECHNICAL FIELD

[0001] The application relates to an intelligent adjustment method of an industrial waste liquid low-temperature evaporation process and belongs to the technical field of industrial process automation control. BACKGROUND

[0002] In the current industrial waste liquid low-temperature evaporation concentration treatment system, an automatic control scheme mainly adopts a fixed set value feedback control strategy. In this kind of scheme, the preset thermodynamic model parameters of the controlled object are constant or only slowly drift with time, the liquid phase temperature and vacuum degree deviation are collected to adjust the heating power or valve opening, and the system is maintained in a steady state when processing a conventional fluid with uniform physical properties. In the concentration process of high-salt and high-organic waste liquid, the physical properties of the fluid nonlinearly evolve with the increase of the concentration. Near the critical crystallization point or gelation point, the fluid is suddenly changed from a Newtonian fluid to a high-viscosity non-Newtonian fluid, which causes the process gain and dominant time constant of the controlled object to jump by orders of magnitude.

[0003] The fixed parameter controller lacks real-time perception of the change in the transfer function structure, and the cumulative effect of the integral element easily causes the control amount and the actual response characteristics to be mismatched, resulting in oscillation or overshoot of the heating unit, coking of the heat exchange surface, and the like. For example, a high-concentration online instrument waste liquid low-temperature negative pressure evaporation method is disclosed in the Chinese patent application CN111977732A. The scheme defines the vacuum degree as 0.095 MPa to 0.1 MPa and the heating temperature as 65 ~75 Process running interval, which is essentially a fixed value adjustment, focuses on static working condition parameter matching. In the deep concentration process of high-salt and high-organic waste liquid, a dynamic compensation mechanism is not introduced for the sudden change in the rheological properties of the fluid. Near the critical crystallization point or gelation point of the waste liquid, the fluid is suddenly changed from a Newtonian fluid to a high-viscosity non-Newtonian fluid, and the process gain and dominant time constant of the controlled object jump by orders of magnitude. The fixed heating and pressure parameters cannot immediately match the rapidly changing system impedance and lag time, which easily causes oscillation or overshoot of the heating unit, coking of the heat exchange surface, and a sharp drop in the evaporation efficiency. The single-variable feedback system is difficult to distinguish the physical causes of the decrease in the heat transfer efficiency. The increase in the fluid viscosity and the increase in the heat resistance caused by the fouling of the heat exchange wall both show temperature response lag. The existing system cannot decouple the two kinds of disturbance sources with different control requirements, and the misjudgment of the state causes the system to incorrectly increase the heating power under the fouling working condition where cleaning operation should be performed, thereby accelerating the equipment wear and tear. Although an online viscosity meter is introduced, the contact type instrument has high maintenance cost and insufficient stability under the high-temperature vacuum and fouling working conditions.

[0004] Therefore, the technical problem to be solved by the present application is to identify the kinetic model parameters of the controlled object online by using general process control variables without relying on special physical property detection instruments, to real-time reconstruct the control law according to the kinetic model parameters, and to realize the decoupling and adaptive adjustment of the nonlinear time-varying system. SUMMARY

[0005] To solve the problems presented in the background art, the technical solutions of the present application are as follows: An intelligent adjustment method for an industrial waste liquid low-temperature evaporation process, which is executed by a process control system and includes the following steps:

[0006] During the steady-state operation of the vacuum regulating valve maintaining a constant opening, a square wave pressure step signal with a duration of is superimposed and sent to the drive unit of the vacuum regulating valve as a test excitation, and a transient value sequence of the liquid-phase temperature sensor is synchronously collected through an analog input channel at a preset sampling frequency ;

[0007] An energy feature operation is performed on the transient value sequence to calculate an absolute error integral sum of a temperature response curve relative to a reference temperature before excitation, and the formula is , wherein is the transient value, is the reference temperature, and is the sampling period; a time feature required for the temperature response curve to return to the reference temperature by a preset proportion in the free recovery stage is extracted;

[0008] The absolute error integral sum and the time feature are substituted into a preset first-order plus pure lag inverse model to solve an equivalent process gain and an equivalent time constant of the controlled object at the current time, wherein the equivalent process gain characterizes the static sensitivity of the controlled object to the input energy, and the equivalent time constant characterizes the inertial lag degree of the controlled object to the input energy;

[0009] Based on the internal model control principle and the zero-pole cancellation logic, the equivalent process gain and the equivalent time constant are used to dynamically calculate the PID controller parameters of the next control period, wherein the proportional gain is set to be inversely proportional to the equivalent process gain , and the integral time is set to be proportional to the equivalent time constant ;

[0010] The calculated proportional gain and integral time are used to update the PID control law, and a heating power instruction or a discharge valve opening instruction is generated according to the reconstructed control law to configure the pole of the closed-loop control system in the stable region of the complex plane.

[0011] Preferably, the equivalent process gain and the equivalent time constant The step of dynamically calculating the PID controller parameters of the next control period specifically includes performing calculation according to the following formula: wherein, is a proportional gain, is an integral time, is a preset closed-loop response bandwidth adjustment factor, which is used to define the trade-off relationship between the response speed and stability of the closed-loop control system.

[0012] Preferably, the method further includes a multi-source disturbance decoupling step based on the orthogonal evolution characteristics of the parameters: constructing a parameter evolution trend monitoring window with time as the independent variable, and tracking the change rate of the equivalent time constant and the equivalent process gain in real time; when it is monitored that the equivalent time constant presents a monotonically increasing trend under constant liquid level conditions, and the change amplitude of the equivalent process gain does not exceed the preset tolerance threshold at the same period, it is determined that the system is in a heat transfer attenuation state dominated by the wall thermal resistance, and a thermal shock cleaning instruction for the heat exchange wall is generated and executed; when it is monitored that the equivalent process gain presents an exponential jump trend, it is determined that the system is in a phase change state dominated by the fluid flow characteristics, and a power adaptation adjustment instruction based on the reconstructed control law is generated and executed.

[0013] Preferably, the method further includes a phase state stability discrimination step based on signal cross-correlation, acquires a pressure instruction data stream corresponding to the test excitation, and performs cross-correlation coefficient calculation within a sliding window on the instantaneous value sequence and the pressure instruction data stream; compares the calculated cross-correlation coefficient with a preset coupling threshold, and when the cross-correlation coefficient is lower than the coupling threshold, generates a determination result that the system enters a thermodynamic non-equilibrium state dominated by the gas-liquid interface foam layer; in response to the determination result, interrupts the current PID regulation logic and switches to a high-frequency shear oscillation mode, and controls the actuator to perform a preset frequency of pneumatic defoaming action until the cross-correlation coefficient recovers to above the coupling threshold.

[0014] Preferably, the step of extracting the time characteristic required for the temperature response curve to recover to a preset proportion of the reference temperature in the free recovery stage specifically includes: continuously monitoring the difference between the instantaneous value sequence and the reference temperature during the free recovery stage after the test excitation ends; recording the time length experienced when the difference decays to 36.8% of the maximum transient deviation value, and directly defining the time length as the observed value of the equivalent time constant characterizing the thermal inertia of the controlled object.

[0015] Preferably, the method further includes a step of adjusting the closed-loop response bandwidth adjustment factor The adaptive correction step is to obtain the residual of the measured value and the target value of the physicochemical index of the residual liquid at the end of each control batch; the gradient descent algorithm is used to correct the closed-loop response bandwidth adjustment factor of the next batch according to the residual When the residual indicates that there is a history of overshoot oscillation, the value of the closed-loop response bandwidth adjustment factor is increased to reduce the system response bandwidth When the residual indicates that there is a history of response delay, the value of the closed-loop response bandwidth adjustment factor is reduced to increase the system response bandwidth.

[0016] The duration of the square wave pressure step signal is preferably set according to the following rule: the duration is set to be greater than the pure lag time of the controlled object and less than 10% of the dominant time constant of the controlled object, so as to ensure that the current evaporation steady state balance is maintained while the system dynamic characteristics are excited, and the specific value of the duration is dynamically updated according to the equivalent time constant identified in the last control period.

[0017] Preferably, before the step of performing energy feature operation on the instantaneous value sequence, a data preprocessing step is further included, which performs denoising and normalization processing on the collected instantaneous value sequence; a preset moving average filter is used to filter high-frequency random noise, and the filtered data sequence is mapped to a relative deviation domain with the reference temperature as the zero point, so as to eliminate the influence of sensor zero drift on the absolute error integral and The calculation of the absolute error integral and

[0018] Preferably, after the step of generating the heating power instruction or the discharge valve opening degree instruction according to the reconstructed control law, the following steps are further included: verifying whether all the roots of the closed-loop system characteristic equation under the generated instruction are located in the left half plane of the complex plane; if the verification result is no, then the proportional gain is forcibly adjusted until the root trajectory of the characteristic equation completely enters the left half plane, so as to ensure the absolute stability of the control loop when the model parameters of the controlled object nonlinearly mutate.

[0019] Preferably, the method is specifically applied to processing industrial waste liquid containing non-Newtonian fluid characteristics; the control system compensates for the model mismatch caused by the exponential change of fluid viscosity in real time by performing the step during the process that the industrial waste liquid changes from a low viscosity state to a high viscosity state, so as to maintain the control accuracy of the liquid phase temperature near the critical crystallization point.

[0020] Compared with the prior art, the method has the following beneficial effects:

[0021] ​​​1. In the process of low-temperature evaporation of industrial waste liquid, the vacuum regulating valve is used to actively superimpose micro-pulse pressure excitation during steady-state maintenance. The deviation integral area and decay characteristics of the synchronous analysis of the liquid phase temperature response curve are used to construct the online identification loop of the dynamics parameters of the controlled object. The mapping relationship between the time domain response energy spectrum and the system process gain and time constant is used to capture and quantify the model drift of the controlled object in real time when the physical properties change nonlinearly due to the change of the waste liquid from Newtonian fluid to non-Newtonian fluid. The controller executes the control law reconstruction based on the internal model principle, dynamically adjusts the proportional gain and integral time to match the current system inertia and lag, configures the closed-loop control system poles in the stable region, eliminates the risk of heating shock or response divergence caused by model mismatch at the end of the concentration period in the traditional fixed parameter PID control, and ensures the control stability under the condition of large fluctuation of fluid rheological characteristics.

[0022] 2. The orthogonal relationship between process gain and time constant in the time domain characteristics of step response is used to realize decoupling diagnosis of different nature disturbances of fluid body thickening and heat exchange wall fouling. The system monitors the monotonic evolution trend of the time constant under constant liquid level and the relative change of the process gain, distinguishes the increase of the heat transfer interface thermal resistance caused by thermal inertia drift and the change of the fluid viscosity caused by dynamics change, respectively triggers the heat shock cleaning program or the heating power adaptation action, and based on the soft sensing and fault separation mechanism of control theory, avoids the error increase of heating power caused by the failure to distinguish the cause of the decrease of heat transfer efficiency, and protects the evaporation efficiency and prolongs the continuous operation period of the equipment.

[0023] 3. A correlation coefficient stability discrimination mechanism based on the cross-correlation coefficient of pressure command data stream and temperature feedback data stream is established to solve the problem of distorted control signal caused by foam layer dominance or false boiling in the boiling process of high-viscosity waste liquid. When the cross-correlation coefficient is lower than the preset coupling threshold, the control system identifies the thermodynamic non-equilibrium state, automatically switches from the normal regulation mode to the high-frequency shear oscillation mode, uses the pneumatic action of the actuator to eliminate the accumulation of foam at the gas-liquid interface, ensures the observation and closed-loop regulation of complex multiphase flow process under the configuration of general sensors, prevents control blind area and misoperation caused by signal correlation decoupling, and ensures smooth transition at the critical phase change point. BRIEF DESCRIPTION OF DRAWINGS

[0024] Fig. 1 The intelligent regulation flowchart of the active energy feature identification and multi-mode decoupling of the present application;

[0025] Fig. 2 The adaptive iterative correlation graph of the closed-loop response bandwidth adjustment factor and batch residual of the present application;

[0026] Fig. 3 The principle diagram of the closed-loop control system integrated with inverse model solution and multi-dimensional actuator of the present application. DETAILED DESCRIPTION

[0027] The application will be further described in detail below with reference to the drawings, and it should be noted that the following examples are only used to illustrate the application, but not to limit the protection scope of the application.

[0028] The application provides an intelligent adjustment method for a low-temperature evaporation process of industrial waste liquid, which is executed by a process control system and is based on a closed-loop identification and self-adaptive adjustment circuit with active excitation. The control system does not rely on direct physical measurement of fluid viscosity or density, but injects a specific energy disturbance into an actuator, and observes the energy dissipation characteristics of the controlled object in the time domain, so as to invert the dynamic model parameters of the system, and reconstruct the PID control law in real time. After the system is started and enters the initial steady state, the control system executes a steady state judgment procedure, and monitors the liquid phase temperature and gas phase pressure in the evaporation kettle in real time. When the temperature fluctuation amplitude in the continuous N sampling periods is less than , and the pressure fluctuation amplitude is less than , it is determined that the system is in a quasi-steady state suitable for model identification. In view of the engineering problem that the conventional feedback control model is mismatched due to the nonlinear mutation of the rheological properties of the industrial waste liquid in the concentration process, an online identification mechanism based on a micro-pulse excitation is adopted. During the steady state operation of the vacuum regulating valve with constant opening degree, the controller superimposes a square wave pressure step signal with a duration of onto the driving unit of the vacuum regulating valve as a test excitation. Before the test excitation is applied, the control system executes a static friction force calibration of the vacuum regulating valve, outputs a ramp test signal with constant slope to the driving unit, synchronously monitors the response slope of the gas phase pressure sensor, identifies the minimum displacement driving instruction value of the actuator as the dead zone compensation threshold, and superimposes the dead zone compensation threshold to the basic instruction when generating the square wave pressure step signal, so as to eliminate the excitation energy loss and waveform distortion caused by the mechanical gap or static friction of the valve. The signal is used as a test probe to detect the dynamic characteristics of the system, and is not used to change the evaporation working condition.

[0029] During this period, the instantaneous value sequence of the liquid phase temperature sensor is synchronously collected through an analog input channel at a preset sampling frequency , which is usually set to be not less than . The operation unit in the controller performs energy characteristic operation on the instantaneous value sequence, and the core is to calculate the absolute error integral of the temperature response curve with respect to the reference temperature before excitation , and the calculation formula is , wherein, is the instantaneous value, is the reference temperature, is the sampling period. At the same time, the system extracts the time characteristics required for the temperature response curve to recover to the reference temperature by a preset proportion, for example, 63.2%, in the free recovery stage after the end of the test excitation. The numerical value directly maps to the static sensitivity of the controlled object to the input energy, i.e., the equivalent process gain. The recovery time characteristic directly characterizes the degree of inertial lag of the controlled object, i.e., the equivalent time constant. Equivalent process gain The numerical determination is based on the energy conservation logic of the impulse response of a linear system. The computing unit will calculate the integral of the absolute error of the obtained temperature response curve and sum the results. Divide by the amplitude and duration of the test excitation pressure Product, closed-loop response bandwidth adjustment factor to prevent parameter drift during iteration. The numerical constraints are based on the equivalent time constants obtained in the current identification. Within the closed interval of 0.5 to 1.5 times, the interval boundary is set according to the Nyquist stability criterion. The physical blocking model mismatch induces the distribution of closed-loop poles in the right half plane. Through this active detection mechanism, the control system can transform the unmeasurable fluid rheological state into quantifiable control theoretical parameters.

[0030] Based on the model parameters identified above, the control system performs parameter self-tuning based on the internal model control principle and zero-pole cancellation logic. Given that sudden changes in waste liquid viscosity can cause significant pole drift in the system, if the PID parameters remain unchanged, the closed-loop system will enter the oscillation or divergence region. Therefore, the controller utilizes the calculated equivalent process gain... With equivalent time constant The PID controller parameters for the next control cycle are dynamically calculated, and the specific parameter reconfiguration follows the deterministic procedure: proportional gain. Set to be equivalent to the process gain It exhibits an inverse relationship to compensate for changes in system sensitivity; integration time Set to the equivalent time constant They are directly proportional to each other to match the dominant time constant of the system. The specific calculation formula is as follows: and ,in This is a preset closed-loop response bandwidth adjustment factor. Instead of a fixed value, it is adaptively corrected using an online gradient descent algorithm: after each control batch, the residual values ​​of the measured physicochemical properties of the residual liquid are obtained compared with the target values. If the residual indicates a history of overshoot oscillation, the value is increased in the next batch. The value is used to reduce the system response bandwidth, and vice versa. To improve response speed, utilize the updated and The PID control law is reconstructed, and a heating power command or a discharge valve opening command is generated based on the reconstructed control law, thereby configuring the poles of the closed-loop control system in the stable region of the complex plane.

[0031] To solve the limitation of single variable feedback that cannot distinguish the change of fluid rheological property from the fouling of equipment, the present invention constructs a multi-source disturbance decoupling logic based on the orthogonal evolution characteristics of parameters. The system constructs a parameter evolution trend monitoring window in the time domain, and tracks the equivalent time constant and the rate of change of equivalent process gain in real time. When the equivalent time constant presents a monotonically increasing trend under constant liquid level conditions, and the change amplitude of the equivalent process gain does not exceed the preset tolerance threshold such as 5% at the same period, the system determines that the system is in a heat transfer attenuation state dominated by wall thermal resistance, i.e., the heat exchange tube wall is fouled, and then generates and executes a thermal shock cleaning instruction for the heat exchange wall surface, instead of incorrectly increasing the heating power. On the contrary, when the equivalent process gain presents an exponential jump trend, the system determines that it is in a phase change state dominated by fluid rheological property, i.e., the viscosity of the waste liquid increases dramatically, and at this time, a power adjustment instruction based on the reconstructed control law is generated and executed. This decoupling mechanism ensures that the control system can apply the correct control strategy for physically distinct disturbances. In addition, to address the false liquid level and foam interference that is easily generated in the boiling process of high-viscosity fluid, the system also integrates a phase state stability discrimination step based on signal cross-correlation. The system obtains the pressure command data stream corresponding to the test excitation in real time, and performs cross-correlation coefficient calculation in the sliding window for the instantaneous value sequence and the pressure command data stream. The calculated cross-correlation coefficient is compared with the preset coupling threshold, which is usually set to 0.8. When the cross-correlation coefficient is lower than the coupling threshold, it indicates that the linear transfer relationship between the pressure input and the temperature output is destroyed, and the system determines that it has entered a thermodynamic non-equilibrium state dominated by the gas-liquid interface foam layer. In response to this determination result, the controller immediately interrupts the current PID regulation logic and switches to a high-frequency shear oscillation mode. The control actuator such as the pneumatic valve executes the preset frequency of pneumatic defoaming action until the cross-correlation coefficient returns to above the coupling threshold.

[0032] After generating the control instruction each time, the system also performs a posteriori check of stability. The system uses the currently recognized object model parameters and the calculated PID parameters to construct the characteristic equation of the closed-loop system, and verifies whether all the roots of the characteristic equation are located in the left half of the complex plane. If the verification result is no, it indicates that the current parameter combination may cause system instability, and the controller will adjust the proportional gain , usually by gradually reducing the gain by introducing an attenuation coefficient, until the root trajectory of the characteristic equation completely enters the left half plane. This deterministic procedure serves as the last safety barrier of the system, ensuring that the control loop always remains absolutely stable when the object model parameters undergo nonlinear mutations. The duration of the square wave pressure step signal The settings also follow strict dynamic update rules: the specific values ​​are based on the equivalent time constant identified in the previous control cycle. The settings are limited to being greater than the pure time delay of the controlled object and less than 10% of the dominant time constant of the controlled object, to ensure that while effectively stimulating the dynamic characteristics of the system, the current steady-state equilibrium of evaporation is not excessively disrupted. For the collected raw data, after... Before the calculation, the system uses a preset moving average filter to filter out high-frequency random noise and maps the filtered data sequence to the relative deviation domain with the reference temperature as the zero point.

[0033] Example 1: In the industrial application scenario of low-temperature evaporation and concentration of high-salt organic waste liquid, the waste liquid being treated abruptly changes from a Newtonian fluid to a non-Newtonian fluid with strong thixotropy near the critical crystallization point. This nonlinear phase transition leads to process gain of the controlled object. It increases more than 10 times in a few seconds, while the time constant... This also results in a jump in magnitude. Traditional fixed-parameter PID controllers, unable to detect such drastic changes in the model structure, are prone to causing violent fluctuations in heating power at the end of concentration, leading to coking on the heat exchange wall and production interruption. When the system faces this condition, the intelligent adjustment system of this invention actively sends a duration of [unclear - possibly related to power generation] to its drive unit during the steady-state operation of the vacuum regulating valve maintaining a constant opening. For example, a 300ms square wave pressure step signal can be used as the test excitation, and the instantaneous liquid phase temperature sequence can be acquired at high frequency. Calculate the integral of the absolute error of the temperature response curve relative to the reference temperature before excitation. ,Right now Simultaneously, the system extracts the time characteristic required for the response curve to recover to 63.2% of the reference temperature during the free recovery phase. Using these two characteristic quantities, the system inverts the current equivalent process gain in real time. With equivalent time constant Based on the internal model control principle, the PID parameters for the next control cycle are dynamically calculated, including the proportional gain. Set as with Inversely proportional, that is Integral time Set as with Proportional This online identification mechanism based on active excitation enables the control system to instantly sense and compensate for the drift of the controlled object model when the fluid viscosity undergoes a nonlinear change. By reconstructing the control law, the closed-loop poles are forcibly locked in the stable region, thus smoothly navigating the fluid phase transition region without human intervention and avoiding heating oscillations and coking.

[0034] Embodiment 2: This embodiment aims to rigorously test the intelligent adjustment method of Embodiment 1 by constructing a low-temperature evaporation concentration test platform containing real industrial noise and process disturbances, and focuses on verifying the control stability and anti-interference ability of the gain inverse compensation mechanism based on the characteristics used in the present application when dealing with fluid nonlinear viscosity mutations; the test platform is built on a pilot-scale forced circulation evaporator with an effective volume of , equipped with an industrial-grade tube heat exchanger and a vacuum flash tank, to simulate the signal environment in a real industrial site. In the signal transmission loop of the liquid phase temperature sensor, a signal generator is used to actively superimpose Gaussian white noise with a signal-to-noise ratio of , and introduce a power frequency interference with a frequency of and an amplitude of to construct a challenging measurement environment. The processed material is selected as a simulated high-salt organic waste liquid with an initial viscosity of at room temperature, but when concentrated to a critical solid content of about 45%, the viscosity will show a nonlinear exponential jump, with a peak viscosity of up to ; the test design includes two parallel test groups: the control group and the sample group of the present application. The control group uses the fixed parameter PID control strategy commonly used in the industry, with a proportional gain and an integral time adjusted according to the empirical formula and kept constant; the sample group of the present application loads the intelligent adjustment algorithm, and the control parameters are dynamically reconstructed based on the real-time identified equivalent process gain and equivalent time constant ; the core goal of the test is to investigate the adjustment behavior of heating power and the stability of liquid phase temperature when the material concentration crosses the critical viscosity mutation point. During the test, the key sampling period is set to , which is not randomly selected, but is based on the trade-off between the viscosity mutation rate of the material and the thermal inertia of the system: given that the viscosity doubling time of the material in the phase transition region can be as short as 5 seconds, in order to ensure the information integrity under Shannon's sampling theorem and capture the small temperature response characteristics, the sampling frequency must be high enough; however, too high a frequency will introduce too much high-frequency noise and increase the computational load. After measurement , the sampling period can meet the needs of accurate integration of characteristics, and can effectively filter out the superimposed high-frequency noise with a sliding average filter.

[0035] After the experiment started, as the evaporation process progressed, the material concentration gradually approached the critical point. In the control group, when the material viscosity began to rise, the system exhibited a sluggish temperature response due to the sharp decrease in the fluid heat transfer coefficient. The fixed-parameter PID controller could not detect this change in the process model, and the integral term continuously accumulated errors, causing the heating power command to be pushed to the saturation limit in a short period of time, resulting in severe temperature overshoot and oscillation. Data showed that the liquid phase temperature in the control group fluctuated significantly near the target value. The experiment experienced continuous fluctuations, and at the 42-minute mark, a coking alarm was triggered due to excessively high heat exchanger tube wall temperature, forcing the experiment to stop. In contrast, the sample group of this invention exhibited drastically different control behavior. When the material entered the viscosity abrupt change region, the temperature response curve induced by the periodically superimposed pressure micro-pulse excitation of the system changed shape, and the decay time was significantly prolonged. As the integral value increases sharply, the controller of the prototype of this invention keenly detects this characteristic change and calculates the equivalent process gain in real time. It showed a trend of rapidly increasing from the initial value to 3.8 times, and the control law reconstruction logic immediately intervened, adjusting the proportional gain. The value was reduced from the initial setting of 2.5 to 0.65 according to the inverse relationship, and the integration time was extended accordingly. This parameter adjustment offsets the risk of excessive loop gain introduced by the increase in process gain. Table 1 below clearly shows the comparison of the core control indicators of the two systems during the critical viscosity change stage.

[0036] Table 1: Comparison of Control Performance During Viscosity Sudden Change Stage

[0037]

[0038] Data shows that the sample of this invention can still maintain precise control of liquid phase temperature when faced with severe nonlinear disturbances and background noise interference. The maximum temperature deviation is only 15.8% of that of the control group, and the heating power output is stable without any drastic fluctuations that cause coking.

[0039] Example 3: This example combines Figs. 1 to 3 The method for intelligent regulation of low-temperature evaporation process of industrial waste liquid is explained, such as... Fig. 1 As shown, the system operates in a quasi-steady-state mode, during which the vacuum valve maintains a constant opening and the amplitude of temperature and pressure fluctuations is monitored. When the fluctuations are less than 0.5... After the pressure reaches 0.2 kPa and meets the steady-state condition, the system enters the active excitation and parameter identification stage, which involves superimposing a pressure step signal. It also calculates SIAE and response time, and outputs the result. and After parameter identification is completed, the system immediately enters the multi-source disturbance decoupling determination stage for analysis. With the evolution trend and calculate the cross-correlation coefficient, according to the judgment result, the logical flow is divided into three branches: if the cross-correlation coefficient is less than the threshold value, it is judged as thermodynamic non-equilibrium (foam) and enters the high-frequency shear oscillation mode, executes the aerodynamic defoaming action until the signal recovers and returns to the monitoring state, if the cross-correlation coefficient is greater than the threshold value, it is judged as the wall heat resistance dominant and enters the thermal shock cleaning mode, executes the cleaning instruction to eliminate the heat transfer attenuation and returns to the monitoring state; if the cross-correlation coefficient is greater than the threshold value, it is judged as the fluid rheological property dominant and enters the control law reconstruction and adaptation stage, according to the evolution trend and calculate the cross-correlation coefficient, according to the judgment result, the logical flow is divided into three branches: if the cross-correlation coefficient is less than the threshold value, it is judged as thermodynamic non-equilibrium (foam) and enters the high-frequency shear oscillation mode, executes the aerodynamic defoaming action until the signal recovers and returns to the monitoring state, if the cross-correlation coefficient is greater than the threshold value, it is judged as the wall heat resistance dominant and enters the thermal shock cleaning mode, executes the cleaning instruction to eliminate the heat transfer attenuation and returns to the monitoring state; if the cross-correlation coefficient is greater than the threshold value, it is judged as the fluid rheological property dominant and enters the control law reconstruction and adaptation stage, according to the evolution trend and calculate the cross-correlation coefficient, according to the judgment result, the logical flow is divided into three branches: if the cross-correlation coefficient is less than the threshold value, it is judged as thermodynamic non-equilibrium (foam) and enters the high-frequency shear oscillation mode, executes the aerodynamic defoaming action until the signal recovers and returns to the monitoring state, if the cross-correlation coefficient is greater than the threshold value, it is judged as the wall heat resistance dominant and enters the thermal shock cleaning mode, executes the cleaning instruction to eliminate the heat transfer attenuation and returns to the monitoring state; if the cross-correlation coefficient is greater than the threshold value, it is judged as the fluid rheological property dominant and enters the control law reconstruction and adaptation stage, according to the evolution trend and calculate the cross-correlation coefficient, according to the judgment result, the logical flow is divided into three branches: if the cross-correlation coefficient is less than the threshold value, it is judged as thermodynamic non-equilibrium (foam) and enters the high-frequency shear oscillation mode, executes the aerodynamic defoaming action until the signal recovers and returns to the monitoring state, if the cross-correlation coefficient is greater than the threshold value, it is judged as the wall heat resistance dominant and enters the thermal shock cleaning mode, executes the cleaning instruction to eliminate the heat transfer attenuation and returns to the monitoring state; if the cross-correlation coefficient is greater than the threshold value, it is judged as the fluid rheological property dominant and enters the control law reconstruction and adaptation stage, according to the evolution trend and calculate the cross-correlation coefficient, according to the judgment result, the logical flow is divided into three branches: if the cross-correlation coefficient is less than the threshold value, it is judged as thermodynamic non-equilibrium (foam) and enters the high-frequency shear oscillation mode, executes the aerodynamic defoaming action until the signal recovers and returns to the monitoring state, if the cross-correlation coefficient is greater than the threshold value, it is judged as the wall heat resistance dominant and enters the thermal shock cleaning mode, executes the cleaning instruction to eliminate the heat transfer attenuation and returns to the monitoring state; if the cross-correlation coefficient is greater than the threshold value, it is judged as the fluid rheological property dominant and enters the control law reconstruction and adaptation stage, according to the evolution trend and calculate the cross-correlation coefficient, according to the judgment result, the logical flow is divided into three branches: if the cross-correlation coefficient is less than the threshold value, it is judged as thermodynamic non-equilibrium (foam) and enters the high-frequency shear oscillation mode, executes the aerodynamic defoaming action until the signal recovers and returns to the monitoring state, if the cross-correlation coefficient is greater than the threshold value, it is judged as the wall heat resistance dominant and enters the thermal shock cleaning mode, executes the cleaning instruction to eliminate the heat transfer attenuation and returns to the monitoring state; if the cross-correlation coefficient is greater than the threshold value, it is judged as the fluid rheological property dominant and enters the control law reconstruction and adaptation stage, according to

[0040] As shown in Fig. 2 , the horizontal axis represents the production batch sequence from batch 1 to batch 10, the left vertical axis corresponds to the closed-loop response bandwidth adjustment factor and is marked with a solid line, the right vertical axis corresponds to the batch residual absolute value and is marked with a dashed line, the data points in the figure clearly depict the dynamic correlation fluctuation trend of the closed-loop response bandwidth adjustment factor and the batch residual absolute value; as shown in Fig. 3 , the internal integration has an inverse model solving unit for calculating the process gain and time constant, a parameter self-tuning unit for dynamically updating the proportional and integral parameters, an active excitation generating unit for generating a pressure step signal, and a multi-source disturbance decoupling unit for distinguishing between scaling and viscosity mutation, the instructions issued by the host are divided into three paths: the first path is the pressure step excitation signal, which is transmitted to the valve driving unit to superimpose the step signal and control the vacuum regulating valve, the second path is the power regulation instruction, which is transmitted to the power regulation controller to execute the heating instruction and drive the heating unit, the third path is the high-frequency shear instruction, which is transmitted to the pneumatic executive electromagnetic valve to execute the high-frequency shear action and operate the defoaming pneumatic device, the above execution mechanisms all act on the low-temperature evaporation kettle, while the liquid phase temperature sensor monitors the instantaneous value sequence in the kettle, the temperature response data stream formed by converting the analog quantity to digital quantity through the high-frequency data acquisition module is fed back to the industrial control host, thereby forming a complete closed-loop control hardware architecture.

[0041] Embodiment 4: This embodiment aims to further disclose the adaptive correction logic and specific implementation details of the closed-loop response bandwidth adjustment factor in the intelligent adjustment method, to solve the engineering problem that a single fixed parameter cannot simultaneously meet the contradictory demands of fast response and steady-state disturbance rejection in a nonlinear time-varying process. This adjustment factor It's not just a static scaling factor, but a dynamic variable based on an Iterative Learning Control (ILC) mechanism. The correction process follows a pre-defined gradient descent optimization procedure. In a typical industrial application cycle, the system operates in batches. After each batch is completed, the background data analysis module of the control system will automatically extract two key performance indicators during the operation of that batch: maximum overshoot. With adjustment time The system defines the comprehensive objective function. This function quantifies the deviation of the current batch control effect from the ideal state, and its expression is: ,in, and These are the preset target overshoot (usually 0) and target settling time, respectively. and These are normalized weighting coefficients used to balance the focus on overshoot and velocity.

[0042] Based on the above objective function, the system uses the gradient descent algorithm to calculate the next batch ( (regulatory factors) The specific iterative update formula is as follows: ,in The learning rate is used to control the step size of parameter updates to prevent parameter oscillations caused by overly rapid corrections. In practical engineering implementations, the partial derivative is... It is usually estimated through difference approximation or sensitivity models based on historical data, when When the indicator system has overshoot, i.e., insufficient damping, gradient direction guidance is used. Increasing the damping capacity reduces the closed-loop bandwidth in the next batch, thus enhancing the system's damping characteristics; conversely, when the system exhibits hysteresis, i.e., overdamping, This reduces the size, thereby improving response speed; to ensure the safety of parameter iteration, the system also has a built-in... absolute security boundary Regardless of the calculation result, it is ultimately written to the controller. The values ​​are all forcibly limited to this range; for example, for high-viscosity materials that are prone to coking, It is set to be no less than 0.5 times the dominant time constant to physically prevent high-frequency oscillations in heating power caused by excessive bandwidth. Through this batch-to-batch closed-loop optimization, the control system can automatically converge to the optimal combination of control parameters under the current operating conditions as the production process is repeated.

[0043] Embodiment 5: This embodiment aims to provide a set of parameter boundary offline calibration procedures for the intelligent adjustment method before deployment under non-standard conditions, which constructs a simulation test platform with similar thermodynamic characteristics to the target industrial site in a laboratory environment. Select typical material samples with different solid content gradients and perform a series of step heating experiments under constant vacuum. For each material sample, record the steady-state boiling point temperature and response time constant under different heating power inputs to construct a multi-dimensional solid content-process gain-time constant reference database. Through least squares fitting, an empirical formula for the controlled object model parameters as a function of material concentration is obtained, which is used as a feedforward model for the initial operation of the control system to provide high-quality initial value estimates for the online identification algorithm.

[0044] In addition, for the problems of sensor zero drift and actuator nonlinear dead zone that may occur in actual engineering, this embodiment also defines a pre-deployment calibration process. Before the system is started for the first time, an automatic zero and dead zone scanning program is executed. The control system sets the heating power command to zero, continuously collects liquid temperature and pressure data within a preset time, calculates and stores the background noise baseline and zero deviation value of the sensor. The system increments the valve opening command with a small step size, while monitoring the response change of the pressure sensor, identifies the minimum command value at which the valve starts to produce effective action, and sets this value as the dead zone compensation threshold of the control output. Through the above offline calibration and on-site calibration steps, accurate physical inputs and execution basis are provided for the subsequent adaptive adjustment algorithm.

[0045] Embodiment 6: This embodiment aims to provide a systematic engineering debugging and parameter calibration procedure to solve the problem of initial control parameter mismatch and insufficient model identification accuracy caused by individual differences in physical parameters such as evaporator effective volume and heater thermal inertia of the controlled object and changes in the operating environment. The procedure defines a series of standardized test steps and data processing logic that must be performed before the system is formally put into closed-loop operation. The execution is a static calibration for the linearity and dead zone of the actuator. When the system is in a cold state, the controller outputs an opening command sequence from 0% to 100% of the full range of the vacuum regulating valve, with a step size of 1% and a holding time of 5 seconds. The valve position feedback signal and gas phase pressure response data are collected simultaneously. The least squares method is used to perform piecewise linear fitting on the command feedback data pair to identify the starting dead zone threshold and saturation zone boundary of the valve. These two parameters are written into the output compensation module of the controller to automatically eliminate the nonlinear errors of the actuator in subsequent operation.

[0046] The next step involved conducting an open-loop step response test on the thermodynamic characteristics of the controlled object to obtain the initial tuning parameters of the PID controller. Under conditions of maintaining a basic vacuum and constant liquid level, a heating power step signal with an amplitude of 10% was manually applied, and the response curve of the liquid phase temperature changing over time was continuously recorded until a new steady-state value was reached. The static gain of the system was then extracted from this response curve using the two-point method. Dominant time constant and pure lag time Based on the extracted feature parameters, the initial proportional gain of the PID controller is calculated using the ITAE (time multiplied by absolute error) criterion. With integration time This serves as the baseline parameter for starting the adaptive adjustment algorithm; finally, the sensitivity calibration of the online identification algorithm is performed. After the system enters initial steady-state operation, the micropulse excitation function is activated, and the initial pulse width is set. for 5% of the sensor noise level is used to monitor the resulting fluctuation in liquid phase temperature. If the fluctuation amplitude is less than three times the sensor noise baseline, the value is gradually increased. The pulse width is then determined as the optimal excitation parameter under the current operating condition until the signal-to-noise ratio reaches a preset effective recognition threshold, such as 10dB. This ensures that high signal-to-noise ratio identification of system model parameters is achieved while minimizing process disturbances.

[0047] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent regulation of an industrial waste liquid low-temperature evaporation process, characterized in that, The method is executed by a process control system, comprising the following steps: During steady-state operation with constant opening degree of the vacuum regulating valve, a square wave pressure step signal with a duration of is superimposed to the drive unit of the vacuum regulating valve as test excitation and a sequence of instantaneous values of the liquid phase temperature sensor is synchronously acquired via an analog input channel with a preset sampling frequency . performing energy feature operation on the instantaneous value sequence, calculating absolute error integral of the temperature response curve relative to the reference temperature before excitation , the formula is , wherein, is the instantaneous value, is the reference temperature, is the sampling period; extracting the time feature required for the temperature response curve to return to the preset proportion of the reference temperature in the free recovery stage; The absolute error integral is The equivalent process gain of the controlled object at the current time is calculated by substituting the time characteristics into the preset first-order plus pure lag inverse model The equivalent time constant The equivalent process gain characterizes the static sensitivity of the controlled object to input energy, and the equivalent time constant characterizes the inertial lag degree of the controlled object to input energy; Based on the internal model control principle and zero-pole cancellation logic, the equivalent process gain is calculated and used The PID controller parameters for the next control period are dynamically calculated, with the proportional gain being set in inverse proportion to the equivalent process gain , the integral time being set in direct proportion to the equivalent time constant ; Utilizing the calculated proportional gain with integration time updating the PID control law and generating the heating power command or the discharge valve opening degree command according to the reconstructed control law to configure the pole of the closed-loop control system in the stable region of the complex plane; and, using the solved equivalent process gain with the equivalent time constant The step of dynamically calculating the PID controller parameters for the next control period specifically comprises performing the calculation according to the following formula: wherein, Kp is the proportional gain, Ti is the integral time, Kb is a preset closed-loop response bandwidth adjustment factor for defining a trade-off relationship between the response speed and stability of the closed-loop control system; The method further comprises a multi-source disturbance decoupling step based on the parameter orthogonal evolution characteristics: constructing a parameter evolution trend monitoring window with time as the independent variable, tracking the equivalent time constant in real time and the change rate of the equivalent process gain ; when it is monitored that the equivalent time constant presents a monotonous increasing trend under the condition of constant liquid level, and the change amplitude of the equivalent process gain in the same period does not exceed the preset tolerance threshold, it is determined that the system is in a heat transfer attenuation state dominated by the wall thermal resistance, and a thermal shock cleaning instruction for the heat exchange wall is generated and executed; when it is monitored that the equivalent process gain presents an exponential jump trend, it is determined that the system is in a phase change state dominated by the fluid flow characteristics, and a power adaptation adjustment instruction based on the reconstructed control law is generated and executed.

2. The method as claimed in claim 1, wherein, The method further comprises a phase stability discrimination step based on signal cross-correlation, obtaining a pressure command data stream corresponding to the test excitation, and performing cross-correlation coefficient calculation within a sliding window on the instantaneous value sequence and the pressure command data stream; comparing the calculated cross-correlation coefficient with a preset coupling threshold value, and when the cross-correlation coefficient is lower than the coupling threshold value, generating a determination result that the system enters a thermodynamic non-equilibrium state dominated by the gas-liquid interface foam layer; In response to the determination result, the current PID regulation logic is interrupted and switched to a high-frequency shear oscillation mode, and the control execution mechanism performs a preset frequency of pneumatic defoaming action until the cross-correlation coefficient returns to above the coupling threshold value.

3. The method of claim 1, wherein the method further comprises: The step of extracting the time characteristic of the preset proportion of the temperature response curve in the free recovery phase to return to the reference temperature specifically comprises: continuously monitoring the difference between the instantaneous value sequence and the reference temperature in the free recovery phase after the test excitation ends; recording the time length experienced when the difference decays to 36.8% of the maximum transient deviation value, and directly defining the time length as an equivalent time constant characteristic of the thermal inertia of the controlled object the observation value.

4. The method as claimed in claim 1, wherein, The method also includes adjusting the closed-loop response bandwidth factor. The adaptive correction step involves obtaining the residual between the measured values ​​of the physicochemical properties of the residual liquid and the target values ​​after each control batch is completed. Using gradient descent algorithm, the closed-loop response bandwidth adjustment factor of next batch is corrected reversely according to the residual error ; when the residual error indicates that there is overshoot history, the value of the closed-loop response bandwidth adjustment factor is increased to reduce the system response bandwidth ; when the residual error indicates that there is response lag history, the value of the closed-loop response bandwidth adjustment factor is decreased to increase the system response bandwidth .

5. The method as claimed in claim 1, wherein the said method is characterized by, Duration of square wave pressure step signal The setting rule is: duration The time limit is set to be greater than the pure time delay of the controlled object and less than 10% of the dominant time constant of the controlled object, to ensure that the current steady-state equilibrium of evaporation is maintained while stimulating the dynamic characteristics of the system. The specific value is based on the equivalent time constant identified in the previous control cycle. It is dynamically updated.

6. The method of claim 1, wherein the method further comprises: Before the step of performing energy feature operation on the instantaneous value sequence, the method further comprises a data preprocessing step of performing denoising and normalization processing on the collected instantaneous value sequence; using a preset moving average filter to filter out high-frequency random noise, and mapping the filtered data sequence to a relative deviation domain with the reference temperature as the zero point.

7. The method as claimed in claim 1, wherein the said method is characterized by, After generating heating power commands or discharge valve opening commands based on the reconstructed control law, the process further includes: verifying whether all roots of the characteristic equation of the closed-loop system under the generated commands are located in the left half-plane of the complex plane; if the verification result is negative, then forcibly adjusting the proportional gain. , until the root locus of the characteristic equation completely enters the left half-plane.

8. The method of claim 1, wherein the method further comprises: The method is specifically applied to processing industrial waste liquid containing non-Newtonian fluid characteristics; the control system compensates for the model mismatch caused by the exponential change of fluid viscosity in real time by performing the steps during the process of the industrial waste liquid changing from a low viscosity state to a high viscosity state.

Images