Intelligent diluent concentration adjusting method and system

Abstract

The invention relates to the field of intelligent manufacturing, in particular to an intelligent diluent concentration adjusting method and system. The intelligent diluent concentration blending system comprises a parameter setting and self-learning module, a controller initialization module, an intelligent control module and a closed-loop control module. According to the invention, by introducing deep fusion of a material characteristic self-learning mechanism and fuzzy adaptive PID control, full-automatic intelligent setting and optimization of control parameters are realized; the system can automatically trigger a self-learning process when mother liquor varieties are replaced, key dynamic characteristic parameters such as lag time, response time constants and steady-state gain coefficients of the system are identified in real time by analyzing concentration response data, and a fuzzy control rule base is automatically corrected based on the parameters; and the controller has the capability of adapting to different material characteristics.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent manufacturing, in particular to a diluent concentration intelligent matching method and system. BACKGROUND

[0002] In many industrial fields such as chemical industry, paint, pesticide, electronic chemicals, etc., accurate diluent concentration matching is a key process link to ensure product quality; traditional concentration matching method mainly relies on manual experience or simple automatic control, which has the following obvious defects: Firstly, the traditional method generally uses fixed parameter PID control strategy, which is rigid in the face of different material characteristics. Since different batches of mother liquor have differences in viscosity, density, mixing characteristics, etc., fixed control parameters are difficult to adapt to all working conditions, resulting in problems such as overshoot, oscillation or long adjustment time in the matching process, which seriously affects production efficiency and product consistency.

[0003] Secondly, the existing technology lacks adaptive learning ability to material dynamic characteristics. When the material variety is changed or the environmental conditions are changed, the traditional system needs to be manually reset by the operator. This not only depends on the experience level of the operator, but also consumes time and effort in the debugging process, which cannot meet the requirements of modern production for flexibility and intelligence.

[0004] Furthermore, the traditional concentration control strategy is usually based on static model or simple feedback control, without fully considering the dynamic characteristics of the mixing process, such as system lag, inertia and other key parameters; which leads to the difficulty of control precision and stability to meet the process requirements when the control system faces the mixing process with large lag and nonlinear characteristics; at the same time, the correlation between historical matching data, material characteristic parameters and control effect cannot be effectively mined and utilized, which cannot form an optimized closed loop of getting smarter, restricting the continuous optimization of the production process.

[0005] Therefore, the present application proposes a diluent concentration intelligent matching method and system to meet the needs of modern industrial production for high quality, high efficiency and high flexibility. SUMMARY

[0006] The present application realizes the automatic intelligent setting and optimization of control parameters by introducing the deep integration of material characteristic self-learning mechanism and fuzzy adaptive PID control; the system can automatically trigger the self-learning process when the mother liquor variety is changed, identify the key dynamic characteristic parameters such as lag time, response time constant and steady-state gain coefficient of the system in real time through analyzing the concentration response data, and automatically correct the fuzzy control rule base based on these parameters, so that the controller has the ability to adapt to different material characteristics.

[0007] A diluent concentration intelligent matching method, comprising: Step S1: receiving a target concentration value through a human-computer interaction interface, and calling corresponding basic process parameters from a formula database; When the mother liquor type is detected to be changed or a forced learning instruction is received, a material property self-learning process is performed, the diluent flow rate is changed in a preset mode, and the concentration response data of the mixed liquid is collected through an online concentration sensor; The collected concentration response data and the corresponding diluent flow rate change data are identified and analyzed, and a parameter set representing the dynamic characteristics of the current material is calculated, including the system lag time, the response time constant, and the steady-state gain coefficient; Step S2: taking the dynamic characteristic parameter set as the initial configuration parameters of the fuzzy adaptive PID controller; Step S3: starting the mother liquor delivery and diluent adjustment, starting the mixing process, and continuously obtaining real-time concentration detection values; The current concentration deviation is calculated according to the difference between the target concentration value and the real-time detection value, and the deviation change rate relative to the last sampling time is calculated; The concentration deviation and the deviation change rate are input into the fuzzy inference engine, and the fuzzy rule base based on the dynamic characteristic parameters is used for inference calculation to output the real-time adjustment amount of the PID controller proportionality coefficient, integral coefficient, and differential coefficient; The running parameters of the PID controller are updated according to the adjustment amount output by the fuzzy inference, and the adjustment value of the diluent flow rate is calculated through the updated PID controller; The flow rate adjustment value is converted into a control signal and sent to the diluent adjustment device; Step S4: repeating step S3 to stabilize the concentration value within the preset error range of the target concentration through continuous concentration detection and flow rate adjustment; when the concentration value reaches a stable state, the blending operation is stopped, and the dynamic characteristic parameters, real-time concentration curve, and control instruction sequence are stored in a historical database.

[0008] Preferably, in step S1, the diluent flow rate is changed in a preset mode, specifically including any one of the following modes: In the step signal mode, the diluent flow rate is suddenly increased from the initial value to the set value and kept; In the ramp signal mode, the diluent flow rate is increased from the initial value to the set value at a constant rate; In the pseudo-random binary sequence signal mode, the diluent flow rate is randomly switched between multiple preset level values.

[0009] Preferably, in step S1, the collected concentration response data and the corresponding diluent flow rate change data are identified and analyzed, specifically including: A dynamic model of the relationship between the diluent flow rate and the concentration variation is established based on system identification theory. The model uses a standard difference equation to describe the mathematical relationship between the concentration values at the current and historical time points and the diluent flow rate values. The model includes model order parameters and lag time parameters that represent the dynamic characteristics of the system. A recursive least squares estimation algorithm known in the control field is used to identify the model parameters. The algorithm initializes a parameter vector and a covariance matrix and recursively updates the parameter estimates based on new measurement data at each sampling period. The gain matrix is used to weigh the weight relationship between the historical estimates and the new measurement data, and the covariance matrix reflects the uncertainty of the parameter estimates. Based on the identified model parameters, the system lag time, response time constant, and steady-state gain coefficient are calculated.

[0010] Preferably, based on the identified model parameters, the system lag time, response time constant, and steady-state gain coefficient are calculated, and the specific operations are as follows: According to the standard method in control theory, the system characteristic roots are obtained by solving the characteristic equation corresponding to the difference equation model. The modal time constant corresponding to the dominant characteristic root with the largest modulus is taken as the response time constant, which represents the inertia characteristics of the system response. Based on the final value theorem, the steady-state gain coefficient is obtained by calculating the input-output relationship of the difference equation model under steady-state conditions. This coefficient represents the steady-state sensitivity of the system. By analyzing the response characteristics of the system under a step input signal, the system lag time is determined according to the starting time of the significant change in the concentration response. This parameter represents the transmission delay characteristics of the system.

[0011] Preferably, the specific process of modifying the fuzzy rule base based on the dynamic characteristic parameters in step S2 includes: According to the length of the system lag time, the strength of the integral action in the fuzzy rule is adjusted. The longer the lag time, the weaker the integral action weight. According to the size of the response time constant, the strength of the proportional action in the fuzzy rule is adjusted. The larger the time constant, the stronger the proportional action weight. According to the size of the steady-state gain coefficient, the scaling factor of all output variables in the fuzzy rule is adjusted. The larger the gain coefficient, the smaller the domain range of the output variables.

[0012] Preferably, the construction process of the fuzzy inference engine includes: The basic domain of the concentration deviation is divided into seven fuzzy levels: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large. The membership degree distribution of each fuzzy level is determined by combining triangular membership functions and trapezoidal membership functions. The basic domain of the deviation change rate is also divided into seven fuzzy levels, and the division method is consistent with the fuzzy level division of the concentration deviation. The domain range of each fuzzy level is determined through actual debugging; The output domain of the proportional coefficient adjustment amount, the integral coefficient adjustment amount, and the differential coefficient adjustment amount is divided into seven fuzzy levels. The same fuzzy division strategy as the input variable is adopted; A complete fuzzy control rule base is established, which consists of 49 conditional statements. The antecedent of each rule is the fuzzy level combination of the concentration deviation and the deviation change rate, and the consequent is the fuzzy level combination of the three parameter adjustment amounts; Mamdani-type fuzzy reasoning method is used for reasoning operation, and the fuzzy output obtained by reasoning is de-fuzzified by the center of gravity method to obtain accurate proportional coefficient adjustment amount, integral coefficient adjustment amount, and differential coefficient adjustment amount.

[0013] Preferably, in step S3, the acquisition of real-time concentration detection value, the calculation of concentration deviation and deviation change rate, fuzzy reasoning, PID parameter update, diluent flow adjustment value calculation, and control signal issuance are completed within one control period.

[0014] A diluent concentration intelligent deployment system, comprising: A parameter setting and self-learning module, comprising: A parameter setting unit for receiving a target concentration value through a human-computer interaction interface and calling corresponding basic process parameters from a formula database; A self-learning execution unit for performing a material characteristic self-learning process when detecting a mother liquor type change or receiving a forced learning instruction, controlling the diluent flow to change in a preset mode, and collecting concentration response data of the mixed liquid through an online concentration sensor; A characteristic identification unit for identifying and analyzing the collected concentration response data and corresponding diluent flow change data to calculate a parameter set representing the dynamic characteristics of the current material, including system lag time, response time constant, and steady-state gain coefficient; A controller initialization module for using the dynamic characteristic parameter set as the initialization configuration parameters of a fuzzy adaptive PID controller; An intelligent control module, comprising: A data acquisition unit for starting mother liquor delivery and diluent adjustment, starting the mixing process, and continuously acquiring real-time concentration detection values; A deviation calculation unit for calculating the current concentration deviation based on the difference between the target concentration value and the real-time detection value, and calculating the deviation change rate relative to the previous sampling time; A fuzzy inference unit is configured to input the concentration deviation and the deviation change rate into a fuzzy inference device, perform inference calculation based on a fuzzy rule base modified according to the dynamic characteristic parameters, and output real-time adjustment amounts of the proportional coefficient, the integral coefficient and the differential coefficient of the PID controller; A parameter updating unit is configured to update the operating parameters of the PID controller according to the adjustment amounts output by the fuzzy inference, and calculate the adjustment value of the diluent flow through the updated PID controller; A control execution unit is configured to convert the flow adjustment value into a control signal and send the control signal to the diluent adjustment device. A closed-loop control module is configured to repeatedly call the real-time intelligent control module, and stabilize the concentration value within a preset error range of the target concentration through continuous concentration detection and flow adjustment. When the concentration value reaches a stable state, the dispensing operation is stopped, and the dynamic characteristic parameters, the real-time concentration curve and the control instruction sequence are stored in a historical database.

[0015] The present application has the following advantages: 1. The present application realizes full-automatic intelligent setting and optimization of control parameters by introducing deep integration of material characteristic self-learning mechanism and fuzzy adaptive PID control. The system can automatically trigger the self-learning process when the mother liquor variety is replaced, identify the key dynamic characteristic parameters such as the lag time, response time constant and steady-state gain coefficient of the system in real time through analysis of concentration response data, and automatically modify the fuzzy control rule base based on these parameters, so that the controller has the ability to adapt to different material characteristics.

[0016] 2. The present application realizes precision and optimization of the dispensing process by constructing a complete intelligent control closed loop. The system integrates real-time concentration detection, fuzzy inference decision, PID parameter online adjustment and actuator control in one fast response control cycle, forming a complete closed loop of perception-decision-execution. At the same time, the system also establishes an associated database of formula parameters and characteristic parameters, and has the ability of continuous optimization through accumulation and reuse of historical data. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 FIG. 1 is a structural schematic diagram of a diluent concentration intelligent dispensing system according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to enable personnel in the technical field to better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0019] Embodiment 1: A diluent concentration intelligent dispensing method, comprising: Step S1: Receive the target concentration value through the human-machine interface and call the corresponding basic process parameters from the formula database; the human-machine interface is implemented using an industrial touch screen, and the formula database adopts a relational database management system to store data including mother liquor type, target concentration range, and safe operating parameters; When a change in mother liquor type is detected or a forced learning command is received, a material characteristic self-learning process is executed. The detection of the change in mother liquor type is achieved through RFID tag identification technology. Each mother liquor container is equipped with a unique RFID tag, and the self-learning process is automatically triggered when new tag information is read. During the self-learning process, the diluent flow rate is controlled to change in a preset mode, while the concentration response data of the mixed liquid is collected by an online concentration sensor. The online concentration sensor is a refractive index concentration meter, whose measurement principle is based on the difference in the refractive index of light for liquids of different concentrations, and the measurement accuracy reaches ±0.1%. The collected concentration response data and corresponding diluent flow rate change data are identified and analyzed to calculate a set of parameters characterizing the current dynamic characteristics of the material, including system lag time, response time constant, and steady-state gain coefficient. The identification and analysis process is based on dynamic system modeling theory. By analyzing the correspondence between system input (diluent flow rate) and output (concentration value), a dynamic mathematical model of the system is established. Step S2: Use the set of dynamic characteristic parameters as the initialization configuration parameters of the fuzzy adaptive PID controller; the fuzzy adaptive PID controller is an intelligent controller that introduces a fuzzy logic reasoning mechanism on the basis of the traditional PID controller and can automatically adjust the control parameters according to the system operating status. Step S3: Start the mother liquor delivery and diluent adjustment, begin the mixing process, and continuously acquire real-time concentration detection values; The current concentration deviation is calculated based on the difference between the target concentration value and the real-time detection value, and the rate of change of the deviation relative to the previous sampling time is also calculated. The concentration deviation and the rate of change of deviation are input into the fuzzy inference engine, and the inference calculation is performed based on the fuzzy rule base corrected by the dynamic characteristic parameters. The output is the real-time adjustment of the proportional coefficient, integral coefficient and derivative coefficient of the PID controller. The operating parameters of the PID controller are updated based on the adjustment amount output by fuzzy inference, and the adjustment value of the diluent flow rate is calculated through the updated PID controller. The flow rate regulation value is converted into a control signal and sent to the diluent regulating device; Step S4: Repeat step S3, continuously monitor the concentration and adjust the flow rate to stabilize the concentration value within the preset error range of the target concentration. The condition for determining that the concentration value has reached a stable state is that the absolute deviation between the real-time concentration detection value and the target concentration value is less than half of the preset error range within 5 consecutive control cycles. When the concentration value reaches a stable state, stop the mixing operation and store the dynamic characteristic parameters, real-time concentration curve, and control command sequence in the historical database. The dynamic characteristic parameters stored in the historical database are associated with the basic process parameters. When the same basic process parameters are called again, the system will preferentially recommend using the stored set of dynamic characteristic parameters.

[0020] In step S1, controlling the diluent flow rate varies in a preset mode, specifically including any of the following modes: A step signal mode is employed, where the diluent flow rate is abruptly increased from its initial value to a set value and held. This mode observes the system's step response characteristics by instantaneously adjusting the diluent flow rate from its initial value to the set value and maintaining it. In practice, the initial value is typically set to 20%-30% of the normal operating flow rate, and the set value is set to 70%-80% of the normal operating flow rate. For example, when preparing a certain type of coating, the initial flow rate is set to 2 L / min, and at the adjustment moment, it is instantaneously switched to 8 L / min and held, while the response data of the concentration sensor is recorded at a sampling frequency of not less than 10 Hz. The advantage of this mode is that it can quickly excite the main dynamic modes of the system, making it easy to directly observe the system's lag time, steady-state gain, and main time constant. Its mathematical principle is that the initial stage of the step response reflects the system's lag characteristics, the rising stage reflects the system's inertial characteristics, and the steady stage directly reflects the system's steady-state gain. The ramp signal mode increases the diluent flow rate from its initial value to the set value at a constant rate. This mode gradually increases the diluent flow rate from its initial value to the set value at a constant rate of change. In practice, the choice of the rate of change is crucial and is usually adjusted based on the estimated system inertia. For materials with high viscosity or slow-changing mixing processes, a lower rate of change of 0.5-1 L / min / s is used; for low-viscosity, easily mixed materials, a higher rate of change of 2-5 L / min / s can be used. For example, in the dilution of electronic chemicals, a ramp rate of 1.5 L / min / s is used to linearly increase the flow rate from 3 L / min to 33 L / min within 20 seconds. The advantage of this mode is that the excitation signal is smooth, avoiding severe impact on sensitive materials, while providing dynamic characteristic information of the system at different operating points, making it particularly suitable for mixing processes with strong nonlinearity. A pseudo-random binary sequence signal mode is employed to randomly switch the diluent flow rate between multiple preset levels. This mode switches the diluent flow rate between multiple preset levels according to a pseudo-random sequence. In specific implementation, the upper and lower limits of the flow rate are first determined, typically set between 20% and 80% of the normal operating range, and then this range is divided into several levels. For example, four flow rates are set: 4 L / min, 8 L / min, 12 L / min, and 16 L / min, generating a switching sequence. The holding time of each level is set to 1.5-2 times the estimated system lag time. The advantage of this mode is that it can excite the system within a wide frequency band and obtain rich dynamic information, making it particularly suitable for occasions requiring precise establishment of a system mathematical model. Its theoretical basis lies in the continuous excitation condition in system identification theory, ensuring that all major dynamic characteristics of the system can be identified. In practical applications, the appropriate mode can be selected according to the material characteristics and identification requirements. For conventional liquid mixing, it is recommended to use the step signal mode for rapid identification first. For materials that are prone to precipitation or are sensitive to shear, it is recommended to use the ramp signal mode. When it is necessary to establish an accurate mathematical model for advanced control strategies, the pseudo-random binary sequence signal mode is preferable.

[0021] Step S1 involves identifying and analyzing the collected concentration response data and the corresponding diluent flow rate change data, specifically including: Establish a system difference equation model for the relationship between diluent flow rate and concentration changes:

[0022] in for Concentration value at time, for The diluent flow rate at any given time. For system lag time, , The model order; In practical applications, model order and The choice of model needs to balance model accuracy and computational complexity; for most liquid mixing processes, it is recommended to use... =2, A second-order model with a time lag of 2 can accurately describe the inertial and oscillatory characteristics of the system; the system lag time... The initial value can be estimated based on the pipeline length and flow velocity, and then accurately corrected in the subsequent identification process; this difference equation model describes the dynamic characteristics of the mixing process, where the left-hand side reflects the inertial characteristics of the system and the right-hand side reflects the excitation response characteristics of the system. Parameter identification is performed using the recursive least squares method, specifically including: Parameter initialization: Set the initial parameter vector Initial covariance matrix ,in For sufficiently large positive numbers, It is the identity matrix; During parameter initialization, The value is typically taken to be between 1000 and 10000 to ensure that the algorithm has a sufficient search range in the initial stage; identity matrix The dimension is (n+m+1)×(n+m+1), corresponding to the total number of parameters to be identified. Initial parameter vector. It is usually set to a zero vector, indicating that there are no prior assumptions about the system characteristics before the identification begins; Recursive calculation: For each sampling time... Calculate the gain matrix Update parameter estimates Update the covariance matrix ; in It is a data vector; The recursive calculation process is executed in real time during each sampling period, ensuring that the parameter estimation can track changes in system characteristics; gain matrix Essentially, these are weighting coefficients, which determine the degree to which new measurement data influences parameter correction; as identification progresses, the covariance matrix... As the data vector gradually decreases, the corresponding gain matrix also decreases, reflecting the adaptive process of the algorithm from "coarse tuning" to "fine tuning"; It contains historical state information of the system, providing sufficient excitation information for parameter estimation; Based on the identified parameter vector Calculate the system lag time, response time constant, and steady-state gain coefficient.

[0023] The calculation of system dynamic characteristic parameters based on the identified parameter vectors specifically includes: According to the characteristic equation corresponding to the difference equation model Solve for the eigenvalues, and use the modal time constant corresponding to the dominant eigenvalue (i.e. the eigenvalue with the largest modulus) as the response time constant; The characteristic equation is the core mathematical tool for analyzing the dynamic characteristics of a system, where λ represents the characteristic roots of the system, and each characteristic root corresponds to a specific dynamic mode in the system response. Taking a real mixing process as an example, when the diluent flow rate changes, the concentration does not immediately reach a new steady state, but rather undergoes a dynamic change process; this change process can be decomposed into the superposition of multiple dynamic modes, each with its own rate of change (determined by characteristic roots); the characteristic equation... The solution is to find the patterns of change in all these dynamic modes; The dominant eigenvalue is the eigenvalue with the largest modulus. It corresponds to the slowest decay mode in the system response and is the key factor determining the overall system response speed; the modal time constant is equal to... ,in The sampling period is The dominant characteristic root is the time constant, which quantifies the inertial characteristics of the mixing process. The larger the time constant, the slower the system responds to changes in flow rate, requiring the controller to adopt a more gradual adjustment strategy. Conversely, a smaller time constant means that the system responds quickly and can use more aggressive control parameters. The steady-state gain coefficient is calculated using the following formula: ; steady-state gain coefficient It has a clear physical meaning, representing the steady-state concentration change caused by a unit change in flow rate; the derivation of this formula is based on the final value theorem, which states that when the system reaches steady state, all difference terms tend to be constant, thus yielding this concise expression; in practical applications, the gain coefficient... The higher the gain coefficient, the more sensitive the system is to changes in flow rate, and the lower the output gain of the controller needs to be to avoid overshoot. The system lag time It can be directly determined by analyzing the initial lag phase of the system's step response.

[0024] The specific process of correcting the fuzzy rule base based on dynamic characteristic parameters in step S2 includes: The strength of the integral action in the fuzzy rules is adjusted according to the length of the system lag time; the longer the lag time, the weaker the weight of the integral action. The integral action is used to eliminate steady-state error in control, but in the presence of significant lag, an excessively strong integral action can cause severe integral saturation in the early stages of the response, resulting in a large overshoot. When a long lag time is identified, the system automatically reduces the weight of all outputs related to the integral action in the fuzzy rule base. For example, when preparing high-viscosity materials transported through long pipelines, the lag time may reach several seconds. In this case, the system will significantly weaken the strength of the integral action and rely more on proportional and derivative actions to maintain stability. Once the concentration begins to change significantly, the integral action will be gradually restored, thereby effectively suppressing the overshoot phenomenon. The strength of the proportional action in the fuzzy rules is adjusted according to the magnitude of the response time constant. The larger the time constant, the stronger the weight of the proportional action. The response time constant reflects the inertia of the mixing process. The larger the time constant, the slower the concentration response, and the stronger the initial driving force required for the system to begin effective regulation. Therefore, when a large response time constant is identified, the system will correspondingly increase the weight of the proportional action in the fuzzy rules. Taking the preparation of viscous emulsions as an example, the mixing process has large inertia and slow response. By increasing the proportional action, the controller can issue a stronger regulation command at the initial stage of concentration deviation, effectively overcoming system inertia, speeding up the response speed, and avoiding excessively long regulation time due to insufficient regulation force. The scaling factor of all output variables in the fuzzy rule is adjusted according to the magnitude of the steady-state gain coefficient. The larger the gain coefficient, the smaller the universe of discourse of the output variables. The steady-state gain coefficient characterizes the sensitivity of the process. The larger the gain coefficient, the more significant the concentration fluctuation caused by a small change in the diluent flow rate, indicating that the process is very sensitive. For such sensitive processes, the system will proportionally reduce the universe of discourse of all output variables (including the adjustment of proportional, integral, and derivative coefficients). For example, when diluting certain highly sensitive chemical reagents, the gain coefficient may be very high. In this case, the system will automatically limit the maximum adjustment range of the controller output, transforming the originally large adjustment into a series of fine adjustments, thereby avoiding oscillation and instability caused by excessive control action and ensuring smooth convergence of the control process.

[0025] The construction process of the fuzzy inference engine includes: The basic domain of concentration bias is divided into seven fuzzy levels: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large. The membership distribution of each fuzzy level is determined by combining triangular membership functions and trapezoidal membership functions. The basic domain of the deviation change rate is also divided into seven fuzzy levels, in the same way as the fuzzy level division of the concentration deviation, and the domain range of each fuzzy level is determined through actual debugging. The output domains of the proportional coefficient adjustment, integral coefficient adjustment, and differential coefficient adjustment are divided into seven fuzzy levels, using the same fuzzy partitioning strategy as the input variables. A complete fuzzy control rule base is established, which consists of 49 conditional statements. The antecedent of each rule is a fuzzy level combination of concentration deviation and deviation change rate, and the consequent is a fuzzy level combination of three parameter adjustment amounts. The Mamdani-type fuzzy inference method is used for inference calculation, and the fuzzy output obtained by inference is defuzzified by the centroid method, so as to obtain the precise adjustment of the proportional coefficient, integral coefficient and differential coefficient. During the construction process, the basic domain of concentration deviation is set according to the target concentration value. For example, when the target concentration is 15%, the domain range can be set to [-6%, +6%]; the domain of deviation change rate is determined by debugging the actual response data of the system, with a typical range of [-1.5% / s, +1.5% / s]; the domain of output variable is set according to the initial parameters of the PID controller to ensure that the adjustment can effectively change the control characteristics without causing system oscillation. The triangular membership function is used for the intermediate fuzzy levels to provide precise sensitivity; the trapezoidal membership function is used for the negative and positive levels at both ends to better express extreme cases; the 49 fuzzy rules completely cover all possible input combinations, and each rule embodies a specific control strategy. For example, when the concentration deviation is positive and the deviation change rate is negative, the output proportional coefficient adjustment is negative, the integral coefficient adjustment is negative, and the derivative coefficient adjustment is positive, so as to achieve fast and stable adjustment when the deviation is high but the change is slow. Mamdani-type inference methods map inputs to fuzzy outputs through fuzzy implication and aggregation operations; the centroid method for defuzzification calculates the centroid of the output fuzzy set to obtain precise parameter adjustment values, ensuring smooth changes in control parameters.

[0026] In step S3, the acquisition of real-time concentration detection values, the calculation of concentration deviation and deviation change rate, fuzzy inference, PID parameter update, calculation of diluent flow rate adjustment value, and the issuance of control signals are all completed within one control cycle; the control cycle ranges from 50 milliseconds to 500 milliseconds.

[0027] Example 2: A smart diluent concentration dispensing system, such as... Figure 1 As shown, it includes: The parameter settings and self-learning module includes: The parameter setting unit is used to receive the target concentration value through the human-machine interface and call the corresponding basic process parameters from the formula database. The self-learning execution unit is used to execute the material characteristic self-learning process when a change in the type of mother liquor is detected or a forced learning command is received. It controls the diluent flow rate to change in a preset mode, and at the same time collects the concentration response data of the mixed liquid through an online concentration sensor. The characteristic identification unit is used to identify and analyze the collected concentration response data and the corresponding diluent flow rate change data, and calculate the set of parameters characterizing the current dynamic characteristics of the material, including system lag time, response time constant and steady-state gain coefficient. The controller initialization module is used to use the set of dynamic characteristic parameters as the initialization configuration parameters of the fuzzy adaptive PID controller. The intelligent control module includes: The data acquisition unit is used to initiate the mother liquor delivery and diluent adjustment, start the mixing process, and continuously acquire real-time concentration detection values. The deviation calculation unit is used to calculate the current concentration deviation based on the difference between the target concentration value and the real-time detection value, and to calculate the rate of change of the deviation relative to the previous sampling time. The fuzzy inference unit is used to input the concentration deviation and the rate of change of deviation into the fuzzy inference engine, perform inference calculations based on the fuzzy rule base modified by dynamic characteristic parameters, and output the real-time adjustment of the proportional coefficient, integral coefficient and derivative coefficient of the PID controller. The parameter update unit is used to update the operating parameters of the PID controller based on the adjustment amount output by fuzzy inference, and to calculate the adjustment value of the diluent flow rate through the updated PID controller. The control execution unit is used to convert the flow rate regulation value into a control signal and send it to the diluent regulating device; The closed-loop control module is used to repeatedly call the real-time intelligent control module to stabilize the concentration value within the preset error range of the target concentration through continuous concentration detection and flow regulation. When the concentration value reaches a stable state, the mixing operation stops and the dynamic characteristic parameters, real-time concentration curve, and control command sequence are stored in the historical database.

[0028] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method of intelligent dosing of diluent concentration, characterized by, The method comprises the following steps: Step S1: receiving a target concentration value through a human-computer interaction interface, and calling corresponding basic process parameters from a formula database; When the mother liquor type is detected to be changed or a forced learning instruction is received, a material characteristic self-learning process is performed, the diluent flow is controlled to change in a preset mode, and the concentration response data of the mixed liquid is collected through an online concentration sensor; The collected concentration response data and corresponding diluent flow change data are identified and analyzed, and a parameter set representing the dynamic characteristics of the current material is calculated, including a system lag time, a response time constant, and a steady-state gain coefficient; Step S2: taking the dynamic characteristic parameter set as the initial configuration parameters of a fuzzy adaptive PID controller; Step S3: starting the mother liquor delivery and diluent adjustment, starting the mixing process, and continuously acquiring real-time concentration detection values; A current concentration deviation is calculated according to the difference between the target concentration value and the real-time detection value, and a deviation change rate relative to the last sampling time is calculated; The concentration deviation and the deviation change rate are input into a fuzzy inference engine, and a fuzzy rule base based on the dynamic characteristic parameters is used for inference calculation to output real-time adjustment amounts of the proportional coefficient, the integral coefficient, and the differential coefficient of the PID controller; The running parameters of the PID controller are updated according to the adjustment amounts output by the fuzzy inference, and the adjustment value of the diluent flow is calculated through the updated PID controller; The flow adjustment value is converted into a control signal and sent to the diluent adjustment device; Step S4: repeating step S3 to stabilize the concentration value within a preset error range of the target concentration through continuous concentration detection and flow adjustment; when the concentration value reaches a stable state, the blending operation is stopped, and the dynamic characteristic parameters, the real-time concentration curve, and the control instruction sequence are stored in a historical database.

2. The method of claim 1, wherein, In step S1, the diluent flow is controlled to change in a preset mode, which includes any one of the following modes: In a step signal mode, the diluent flow is suddenly increased from an initial value to a set value and kept; In a ramp signal mode, the diluent flow is increased from an initial value to a set value at a constant rate; In a pseudo-random binary sequence signal mode, the diluent flow is randomly switched between multiple preset level values.

3. The intelligent diluent concentration adjustment method according to claim 2, characterized in that, In step S1, the collected concentration response data and corresponding diluent flow change data are identified and analyzed, which specifically includes: A dynamic relationship model of the diluent flow and the concentration change is established based on system identification theory, which describes the mathematical relationship between the current and historical concentration values and the diluent flow values in the form of a standard difference equation, and includes model order parameters and lag time parameters representing the dynamic characteristics of the system; A recursive least squares estimation algorithm commonly known in the control field is used for model parameter identification, which initializes the parameter vector and the covariance matrix, and recursively updates the parameter estimation value based on new measurement data at each sampling period, wherein the gain matrix is used to weigh the weight relationship between the historical estimation and the new measurement data, and the covariance matrix reflects the uncertainty of the parameter estimation; Based on the identified model parameters, the system lag time, the response time constant, and the steady-state gain coefficient are calculated.

4. The method of claim 3, wherein the concentration of the diluent is adjusted based on the concentration of the diluent in the diluent tank. Based on the model parameters obtained by identification, the system lag time, response time constant and steady-state gain coefficient are calculated, and the specific operations are as follows: According to the standard method in control theory, the system characteristic root is obtained by solving the characteristic equation corresponding to the difference equation model, and the modal time constant corresponding to the maximum modulus of the dominant characteristic root is taken as the response time constant, which represents the inertia characteristic of system response. Based on the final value theorem, the steady-state gain coefficient is obtained by calculating the input-output relationship of the difference equation model under steady-state conditions, which represents the steady-state sensitivity of the system. By analyzing the response characteristics of the system under step input signal, the system lag time is determined according to the starting time of the obvious change of concentration response, which represents the transmission delay characteristic of the system.

5. The method of claim 4, wherein the step of determining the concentration of the diluent comprises: The specific process of correcting the fuzzy rule base based on dynamic characteristic parameters in step S2 includes: According to the length of the system lag time, the strength of the integral action in the fuzzy rule is adjusted, and the longer the lag time, the weaker the integral action weight; According to the size of the response time constant, the strength of the proportional action in the fuzzy rule is adjusted, and the larger the time constant, the stronger the proportional action weight; According to the size of the steady-state gain coefficient, the scaling factor of all output variables in the fuzzy rule is adjusted, and the larger the gain coefficient, the smaller the domain range of the output variable.

6. The method of claim 5, wherein the step of determining the concentration of the diluent comprises: The construction process of the fuzzy inference engine includes: The basic domain of concentration deviation is divided into seven fuzzy levels, namely negative large, negative medium, negative small, zero, positive small, positive medium and positive large, and the membership degree distribution of each fuzzy level is determined by combining triangular membership function and trapezoidal membership function; The basic domain of deviation change rate is also divided into seven fuzzy levels, and the division method is consistent with the fuzzy level division of concentration deviation, and the domain range of each fuzzy level is determined through actual debugging; The output domain of proportional coefficient adjustment amount, integral coefficient adjustment amount and differential coefficient adjustment amount is divided into seven fuzzy levels, and the same fuzzy division strategy as the input variable is adopted; A complete fuzzy control rule base is established, which is composed of 49 conditional statements, and the antecedent of each rule is the fuzzy level combination of concentration deviation and deviation change rate, and the consequent is the fuzzy level combination of three parameter adjustment amounts; Mamdani type fuzzy reasoning method is used for reasoning operation, and defuzzification calculation is carried out on the fuzzy output obtained by reasoning, and finally the accurate proportional coefficient adjustment amount, integral coefficient adjustment amount and differential coefficient adjustment amount are obtained.

7. The method of claim 6, wherein the step of determining the concentration of the diluent comprises: In step S3, the real-time concentration detection value is obtained, the concentration deviation and the deviation change rate are calculated, the fuzzy reasoning is carried out, the PID parameters are updated, the diluent flow adjustment value is calculated, and the control signal is issued in a control period.

8. An intelligent diluent concentration system, comprising: a diluent tank; a diluent pump; a diluent line; a diluent valve; a diluent sensor; a controller; and a user interface. The system is applied to the intelligent diluent concentration deployment method of any one of the above claims 1-7, comprising: Parameter setting and self-learning module, including: Parameter setting unit, for receiving target concentration value through human-computer interaction interface, and calling corresponding basic process parameters from formula database; The self-learning execution unit is configured to execute a material characteristic self-learning process when a mother liquor variety change is detected or a forced learning instruction is received, control the diluent flow to change in a preset mode, and collect concentration response data of the mixed liquid through an online concentration sensor; The characteristic identification unit is configured to perform identification analysis on the collected concentration response data and corresponding diluent flow change data, and calculate a parameter set representing the dynamic characteristics of the current material, including a system lag time, a response time constant, and a steady-state gain coefficient; The controller initialization module is configured to use the dynamic characteristic parameter set as an initialization configuration parameter of a fuzzy adaptive PID controller; The intelligent control module includes: The data acquisition unit is configured to start mother liquor delivery and diluent adjustment, start a mixing process, and continuously acquire real-time concentration detection values; The deviation calculation unit is configured to calculate a current concentration deviation based on a difference between a target concentration value and a real-time detection value, and calculate a deviation change rate relative to a previous sampling time; The fuzzy reasoning unit is configured to input the concentration deviation and the deviation change rate into a fuzzy reasoning device, perform inference calculation based on a fuzzy rule base corrected based on the dynamic characteristic parameters, and output real-time adjustment amounts of a proportional coefficient, an integral coefficient, and a differential coefficient of the PID controller; The parameter updating unit is configured to update running parameters of the PID controller based on the adjustment amounts output by the fuzzy reasoning, and calculate an adjustment value of the diluent flow through the updated PID controller; The control execution unit is configured to convert the flow adjustment value into a control signal and send the control signal to a diluent adjustment device; The closed-loop control module is configured to repeatedly call the real-time intelligent control module, stabilize the concentration value within a preset error range of a target concentration through continuous concentration detection and flow adjustment, stop the adjustment operation when the concentration value reaches a stable state, and store the dynamic characteristic parameters, a real-time concentration curve, and a control instruction sequence into a historical database.

Images