Reaction kettle control method and system for perfume production

Abstract

The present application belongs to the field of automatic control technology, and particularly relates to a reaction kettle control method and system for perfume production, which method comprises: selecting a production batch with the optimal aroma quality as a golden batch, collecting its spectral data, reference temperature curve and real concentration of key components, establishing a soft sensor model by using a chemometrics algorithm, and determining the component concentration evolution path of the golden batch as a component target trajectory. In the production process, the soft sensor model is used to calculate a real-time component vector according to real-time spectral data, and a quality deviation degree is obtained by comparing the real-time component vector with the target component vector; an outer loop calculates a temperature correction amount according to the quality deviation degree, and generates an adaptive temperature set value in combination with the reference temperature curve, which is used as the temperature set value of an inner loop, so as to realize dynamic adjustment of the temperature of the reaction kettle. The present application realizes disturbance compensation and improves the quality consistency among batches of perfume products through inner-outer loop cascade control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic control. More particularly, the present application relates to a reaction kettle control method and system for perfume production. BACKGROUND

[0002] In the field of fine chemical industry such as perfume production, reaction kettle is the core equipment to realize key chemical synthesis such as esterification and condensation. The final quality of perfume, especially its unique aroma characteristics, is not only determined by the purity of the main product, but also by the extremely complex proportion between the main product and various trace by-products. Therefore, ensuring the high consistency of product quality between different production batches is the core problem and goal of production control in this field.

[0003] Currently, the reaction kettle in perfume production is generally controlled automatically by programmable logic controller (PLC) or distributed control system (DCS). These systems usually rely on classic PID (proportional-integral-derivative) control algorithm, whose control target is to accurately maintain the stability of physical parameters in the reaction process, such as temperature, pressure, stirring rate and feed rate, etc. The operation logic of this control strategy is to first determine one or more production batches with optimal product quality, i.e. golden batch, through a large number of experiments, and then record the process data of this batch as the standard process curve for all subsequent production.

[0004] However, the existing technology has defects in practical application: the control target is the reproduction of process variables, not the reproduction of product quality, and the existing control system is blind to the real-time concentration of key chemical components that determine aroma. The control loop is essentially in an open-loop state in the dimension of product quality. This control strategy relies on a core assumption that as long as the process curves such as temperature and pressure of the golden batch are strictly reproduced, the product quality of the golden batch will also be reproduced.

[0005] Under ideal conditions, this assumption can be true, but in large-scale industrial production, the system will inevitably be disturbed by various unmeasured disturbances, such as purity fluctuations between raw material batches, slight differences in catalyst activity, or the introduction of trace impurities. When these disturbances occur, even if the PID controller perfectly controls the actual temperature of the reaction kettle on the preset reference curve with its strong feedback ability, these initial disturbances have already changed the chemical reaction kinetics balance in the kettle, causing the actual progress of the chemical reaction to deviate from the trajectory of the golden batch, resulting in changes in the final proportion of the main product and by-products, directly causing significant differences in the aroma characteristics of the product. Therefore, the existing technology cannot effectively compensate for disturbances caused by fluctuations in raw materials, etc., and it is difficult to ensure the high consistency of the final product quality between different batches. SUMMARY

[0006] To solve the technical problem of poor consistency between batches of product quality caused by the inability to compensate for disturbances in the prior art, the present application provides solutions in the following aspects.

[0007] In a first aspect, the present application provides a reaction kettle control method for perfume production, comprising: selecting a production batch with optimal aroma quality as a golden batch; collecting spectral data, a reference temperature curve, and real concentrations of key components of the golden batch during production; establishing a soft sensor model according to the spectral data and the real concentrations; determining a path of the real concentrations of the golden batch over time as a component target trajectory; deploying a double-loop cascade control architecture comprising an inner loop for performing temperature control and an outer loop for quality decision; during production, using the soft sensor model to calculate a real-time component vector according to real-time collected spectral data, and retrieving a target component vector corresponding to the current time from the component target trajectory; obtaining a quality deviation degree according to the real-time component vector and the target component vector; the outer loop calculates a temperature correction amount according to the quality deviation degree, and combines the temperature correction amount with the reference temperature curve to obtain an adaptive temperature set value as the temperature set value of the inner loop, so as to realize dynamic adjustment of the temperature of the reaction kettle.

[0008] The present application establishes a soft sensor model and a key component target trajectory, and constructs a double-loop cascade control architecture with an inner loop for performing temperature and an outer loop for decision-making quality. The architecture uses a real-time calculated quality deviation index to dynamically correct a reference temperature curve, thereby changing the control target from a traditional process variable to product quality, so that the system can actively compensate for disturbances such as raw material fluctuations, and help to improve the consistency of product quality between different production batches.

[0009] Preferably, the double-loop cascade control architecture comprising an inner loop for performing temperature control and an outer loop for quality decision-making comprises: configuring the inner loop as a PID temperature controller, the process variable of the PID temperature controller being the real-time temperature inside the reaction kettle, the set value being the adaptive temperature set value, and the output being a control signal sent to the heating or cooling valve; and configuring the outer loop as a PID quality controller, the process variable of the PID quality controller being the quality deviation degree, the set value being zero, and the output being the temperature correction amount.

[0010] The present application realizes decoupling of quality decision-making and temperature execution by configuring the inner loop as a PID temperature controller and the outer loop as a PID quality controller. This double-loop cascade control architecture allows the slow-period quality control loop and the fast-period temperature control loop to operate stably at their respective optimal time scales, ensuring the dynamic response performance and stability of the entire closed-loop control system.

[0011] Preferably, the quality deviation degree satisfies the expression: ; wherein, a quality deviation degree at time the calculated quality deviation degree as a process variable of the outer loop PID; a reaction time; a quality deviation degree at time a target concentration of the main product; a real-time concentration of the main product at time an index number of a key byproduct; a real-time concentration of the byproduct at time a target concentration of the byproduct at time a target concentration of the byproduct at time a target concentration of the byproduct at time a target concentration of the byproduct at time a target concentration of the byproduct at time a target concentration of the byproduct at time

[0012] The present application combines the concentration deviation of the main product with the concentration deviations of multiple key byproducts through a unified quality deviation index expression, wherein a negative contribution is exerted on the byproduct exceeding the standard, so as to collapse the complex multi-component quality state into a scalar control signal with clear physical meaning, so that the outer loop controller can correctly execute the regulation action of accelerating the main reaction by increasing the temperature or inhibiting the by-reaction by decreasing the temperature according to the positive or negative of the signal.

[0013] Preferably, the outer loop calculates a temperature correction amount according to the quality deviation degree, and combines the temperature correction amount with a reference temperature curve to obtain an adaptive temperature set value, including: loading the reference temperature curve as a feedforward term; the outer loop taking the quality deviation degree as an input and zero as a set value, and obtaining the temperature correction amount as a feedback term through PID operation; and adding the feedforward term and the feedback term to obtain the adaptive temperature set value.

[0014] Preferably, the adaptive temperature set value satisfies the expression: ; ; wherein, an adaptive temperature set value at time a reaction time; a value of the reference temperature curve at time a temperature correction amount calculated by the outer loop at time a proportional gain of the outer loop; an integral gain of the outer loop; a differential gain of the outer loop; a quality deviation degree at time an integral of the quality deviation degree with respect to time ​​​​​denotes the first derivative of the quality deviation degree with respect to time.

[0015] The present application combines the reference temperature curve from the golden batch as a feedforward item with the temperature correction calculated by the outer loop according to the real-time quality deviation as a feedback item to form an adaptive temperature set value, and the feedforward-feedback combination strategy makes the control system quickly enter the stable working condition by using historical experience and real-time compensate the quality deviation of the current batch to improve the response speed and robustness of the control.

[0016] Preferably, the soft sensor model is established by using a chemometrics algorithm to train the soft sensor model according to the spectral data and the true concentration.

[0017] Preferably, the true concentration of the key component is obtained by physically sampling at a key time point of the golden batch operation and using offline high-precision analysis to obtain the true concentration.

[0018] Preferably, the chemometrics algorithm is a partial least squares method.

[0019] Preferably, the real-time component vector satisfies the expression: ; wherein, denotes the real-time component vector at time . denotes the soft sensor model; denotes the spectral data collected in real time at time . denotes the reaction time.

[0020] The soft sensor model of the present application can effectively process the characteristics of high dimension and multiple collinearity of spectral data, thereby establishing a reliable quantitative relationship between the spectral data and the true concentration of the key component, and providing a feasible measurement means for the subsequent outer loop to realize real-time and quality index-based closed-loop control.

[0021] In a second aspect, the present application provides a reaction kettle control system for perfume production, comprising a processor and a memory, wherein the memory stores computer program instructions which, when executed by the processor, implement the reaction kettle control method for perfume production described above.

[0022] By using the above technical solution, the reaction kettle control method for perfume production described above is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal device is manufactured according to the memory and the processor, and convenient use is achieved.

[0023] The present application has the following advantages:

[0024] The present application can improve the consistency of the quality of the perfume product in different production batches by establishing a soft sensor model and a key component target trajectory, and using a double-loop cascade control architecture of an outer loop calculating a quality deviation index and an inner loop executing temperature regulation, so that the control target is changed from an indirect process variable to a direct product quality, and the system can actively compensate for the quality deviation caused by disturbances such as fluctuations in raw materials.

[0025] The present application can improve the consistency of the quality of the perfume product in different production batches by establishing a soft sensor model and a key component target trajectory, and using a double-loop cascade control architecture of an outer loop calculating a quality deviation index and an inner loop executing temperature regulation, so that the control target is changed from an indirect process variable to a direct product quality, and the system can actively compensate for the quality deviation caused by disturbances such as fluctuations in raw materials.

[0026] The present application can improve the consistency of the quality of the perfume product in different production batches by establishing a soft sensor model and a key component target trajectory, and using a double-loop cascade control architecture of an outer loop calculating a quality deviation index and an inner loop executing temperature regulation, so that the control target is changed from an indirect process variable to a direct product quality, and the system can actively compensate for the quality deviation caused by disturbances such as fluctuations in raw materials. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a flow chart schematically showing a reaction kettle control method for perfume production in the present application;

[0028] Figure 2 is a schematic diagram schematically showing a comparison between a real-time component vector and a target component vector;

[0029] Figure 3 is a schematic diagram schematically showing the input and output of an outer loop PID controller;

[0030] Figure 4 is a schematic diagram schematically showing the update of an adaptive temperature set value. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0032] The specific embodiments of the present application will be described in detail below with reference to the drawings.

[0033] The reaction kettle control method for perfume production disclosed in the embodiments of the present application is described below with reference to Figure 1 , which comprises steps S1-S4:

[0034] S1, selecting a production batch with the best aroma quality as a golden batch; collecting spectrum data, a benchmark temperature curve, and real concentrations of key components of the golden batch in a production process; and establishing a soft sensor model according to the spectrum data and the real concentrations.

[0035] It should be noted that, in order to realize closed-loop control of product quality, two prerequisite problems must be solved, one of which is that the best quality must be mathematically defined, and the other of which is that a means for measuring the current quality in real time in the production process must be provided; the present application constructs a component target trajectory and a soft sensor model by analyzing data of the most successful golden batch in history, the component target trajectory is a mathematical definition of the best quality, which describes the process of the concentration of each key chemical component evolving over time in the best production process; the soft sensor model is a means for measuring in real time, since it is impossible to directly install a concentration meter in the reactor, the present application uses spectrum data collected by an online spectrometer, combined with a chemometrics algorithm, to establish a mathematical model that can deduce real-time concentrations from spectrum data, namely a soft sensor model.

[0036] Specifically, one or more production batches that have been analyzed and confirmed to have the best aroma quality are selected as the golden batch; when the golden batch is running, spectrum data of the golden batch are collected at a high frequency by an online spectrometer installed on the reactor , and a benchmark temperature curve of the reactor and other process data are recorded; Meanwhile, physical sampling is performed at key time points of the reaction, and real concentrations of key components are obtained by using offline high-precision analysis .

[0037] Further, all data pairs are collected, a chemometrics algorithm such as a partial least squares method is used for training, and a mathematical model that can predict component concentrations from spectrum data, namely a soft sensor model, is established ; the soft sensor model is used to predict a real-time component vector according to real-time collected spectrum data in subsequent actual production, which satisfies the expression:

[0038]

[0039] wherein, represents a real-time component vector at time point ; represents the soft sensor model; represents real-time collected spectrum data at time point ; represents a reaction time.

[0040] Further, a real concentration evolution path of the golden batch is obtained The real concentration evolution path of the gold batch over time is determined as the component target trajectory , as the control reference for all subsequent production batches.

[0041] S2, determining the real concentration evolution path of the gold batch over time as the component target trajectory; deploying a dual-loop cascade control architecture comprising an inner loop for performing temperature control and an outer loop for quality decision.

[0042] It should be noted that the present application adopts a cascade control architecture, which contains two coordinated control loops, the outer loop and the inner loop, because the control system cannot directly control the valve opening at the bottom layer with the high-level index of quality deviation. The present application takes advantage of the physical control chain that the valve opening controls the temperature and the temperature affects the product quality, and separates the two tasks of quality control and temperature control through dual-loop design. Specifically, the outer loop is responsible for quality decision, and calculates an adaptive temperature set value that should be reached by comparing the difference between real-time quality and target quality; the inner loop is responsible for temperature execution, and receives the adaptive temperature set value given by the outer loop, immediately adjusts the heating or cooling valve, so that the actual temperature of the reactor accurately tracks the set value.

[0043] Specifically, the control system is configured such that the inner loop, i.e. the secondary loop, is a standard PID temperature controller, whose process variable is the real-time temperature inside the reactor , whose set value is the adaptive temperature set value dynamically given by the outer loop , and whose output is the control signal sent to the heating or cooling valve ; the outer loop, i.e. the primary loop, is a PID quality controller, whose process variable is the quality deviation degree calculated in real time , whose set value is zero deviation, and whose output is , and serves as the temperature set value of the inner loop.

[0044] S3, in the production process, the real-time component vector is calculated from the real-time spectrum data collected by using the soft sensor model, and the target component vector corresponding to the current time is retrieved from the component target trajectory; the quality deviation degree is obtained according to the real-time component vector and the target component vector.

[0045] It should be noted that the outer loop PID controller needs a single scalar error signal as input, but the real-time quality and the target quality are both vectors containing multiple component concentrations; the core goal of the present application is to design a function to collapse this multi-dimensional quality vector deviation into a single scalar ; this scalar must be designed to have a clear physical meaning, so that the PID controller can respond correctly, i.e. when is positive, it means that the main reaction is too slow and needs to be accelerated, and when When the value is negative, it means that the side reaction is too fast or the byproduct exceeds the standard and needs to be inhibited; the PID controller will increase the output, i.e. increase the temperature, when receiving a positive error, and will decrease the output, i.e. decrease the temperature, when receiving a negative error, which is exactly matched with the required physical adjustment action.

[0046] Specifically, in a new production batch, the online spectrometer acquires real-time spectral data ; soft sensor model Solve it to get the real-time component vector ; at the same time, the system retrieves the target component vector corresponding to the current time from the component target trajectory .

[0047] Further, the quality deviation degree is calculated:

[0048]

[0049] Wherein, The scalar quality deviation degree calculated at time is taken as the process variable of the outer loop PID; Indicates the reaction time; Indicates the target concentration of the main product at time ; Indicates the real-time concentration of the main product at time ; Indicates the index number of the key byproduct; Indicates the real-time concentration of the th byproduct at time ; Indicates the target concentration of the th byproduct at time .

[0050] It should be noted that the first term of the expression is the concentration deviation of the main product, and if the real-time concentration is lower than the target concentration, i.e. the main reaction is too slow, the term is positive, contributing a positive , indicating that it needs to be accelerated; the second term of the expression is the concentration deviation of the byproduct, and if the real-time byproduct is higher than the target, i.e. the side reaction is too fast, the term is positive; by setting a negative sign in front of the second term, the byproduct exceeding the standard is converted into a negative contribution to . Therefore, when is positive, it clearly indicates that the main reaction is lagging, and when is negative, it clearly indicates that the byproduct exceeds the standard, and this unified scalar signal can be directly sent to the PID controller.

[0051] Exemplarily, Figure 2is a schematic diagram of real-time component vector compared with target component vector, where green dotted line represents target concentration of main product, and blue solid line represents real-time concentration of main product.

[0052] S4, the outer loop calculates a temperature correction amount according to the quality deviation, and combines the temperature correction amount with the reference temperature curve to obtain an adaptive temperature set value as the temperature set value of the inner loop, so as to realize dynamic adjustment of the temperature of the reaction kettle.

[0053] It should be noted that the quality deviation generated in the foregoing steps indicates the direction and degree of quality deviation, and the outer loop PID controller is a mechanism for correcting the deviation. The outer loop PID controller calculates a temperature correction amount according to the quality deviation and the reference temperature curve. To avoid sluggish response of the control system, the present application adopts a combination strategy of feedforward and feedback, that is, the reference temperature curve of the golden batch is used as a feedforward term, and is used as a feedback term; the adaptive temperature set value of the present application is just the implementation of this feedforward-feedback strategy, obtaining a final temperature set value which is both experienced and real-time adaptive, and is executed by the inner loop.

[0054] Specifically, the system loads the reference temperature curve of the golden batch ; the outer loop PID controller calculates a temperature correction amount with a set value of 0, wherein the temperature correction amount satisfies the expression:

[0055]

[0056] Further, the inner loop set value is updated as follows:

[0057]

[0058] wherein, represents the adaptive temperature set value at time , which is used as the set value of the inner loop PID; represents the reaction time; represents the value of the reference temperature curve at time as a feedforward term; represents the temperature correction amount calculated by the outer loop PID at time as a feedback term; represents the proportional gain of the outer loop PID; represents the integral gain of the outer loop PID; represents the differential gain of the outer loop PID, and the three gains are tuning parameters of the outer loop PID; quality deviation degree at time t; integral of quality deviation degree over time ; and first derivative of quality deviation degree over time.

[0059] It should be noted that, the update logic of the adaptive temperature set point realizes the combination of feedforward and feedback; when the main reaction is slow, is positive, resulting in is positive, the temperature set point of the inner loop is adjusted upward, the reactor is warmed up, the main reaction is accelerated, and returns to zero; when the byproduct is too much, is negative, resulting in is negative, the temperature set point of the inner loop is adjusted downward, the reactor is cooled down, the side reaction is inhibited, and returns to zero; the present application constitutes a closed-loop control system by cyclic execution, forces the actual chemical component trajectory to track the preset optimal quality trajectory, and realizes direct control of the final fragrance quality.

[0060] Exemplarily, Figure 3 is a schematic diagram of the input and output of the outer loop PID controller, wherein the purple solid line represents the quality deviation degree, and the blue dashed line represents the temperature correction amount.

[0061] Exemplarily, Figure 4 is a schematic diagram of adaptive temperature set point update, wherein the green dashed line represents the reference temperature curve, and the red solid line represents the adaptive temperature set point.

[0062] The embodiment of the present application also discloses a reactor control system for fragrance production, comprising a processor and a memory, and the memory stores computer program instructions which realize the reactor control method for fragrance production according to the present application when executed by the processor.

[0063] The above system also comprises a communication bus and a communication interface and other components well known to those skilled in the art, the settings and functions of which are known in the art, and thus will not be described here.

Claims

1. A reaction vessel control method for perfume production, characterized by, The method comprises the following steps: Selecting a production batch with the best aroma quality as a golden batch; Collecting spectral data, reference temperature curve and real concentrations of key components of the golden batch during the production process; Establishing a soft sensor model according to the spectral data and the real concentrations; Determining the path of the real concentrations of the golden batch over time as a component target trajectory; deploying a double-loop cascade control architecture comprising an inner loop for performing temperature control and an outer loop for quality decision-making; During the production process, the real-time component vector is calculated according to the real-time collected spectral data by using the soft sensor model, and the target component vector corresponding to the current time is retrieved from the component target trajectory; Obtaining a quality deviation degree according to the real-time component vector and the target component vector; The outer loop calculates a temperature correction amount according to the quality deviation degree, and combines the temperature correction amount with the reference temperature curve to obtain an adaptive temperature set value as the temperature set value of the inner loop, so as to realize dynamic adjustment of the temperature of the reaction kettle.

2. The reaction vessel control method for perfume production according to claim 1, characterized by, The double-loop cascade control architecture comprising an inner loop for performing temperature control and an outer loop for quality decision-making comprises: The inner loop is configured as a PID temperature controller, the process variable of the PID temperature controller is the real-time temperature inside the reaction kettle, the set value is the adaptive temperature set value, and the output is the control signal sent to the heating or cooling valve; the outer loop is configured as a PID quality controller, the process variable of the PID quality controller is the quality deviation degree, the set value is zero, and the output is the temperature correction amount.

3. The reaction vessel control method for perfume production according to claim 1, characterized by, The quality deviation degree satisfies the expression: ； wherein, denotes the time instant the calculated quality deviation as a process variable for the outer loop PID; denotes the reaction time; denotes the time instant the main product target concentration; denotes the time instant the main product real-time concentration; denotes the index number of the key by-product; denotes the time instant the real-time concentration of the th by-product; denotes the time instant the target concentration of the th by-product.

4. The reaction vessel control method for perfume production according to claim 1, characterized by, The outer loop calculates a temperature correction amount according to the quality deviation degree, and combines the temperature correction amount with the reference temperature curve to obtain an adaptive temperature set value, comprising: Loading the reference temperature curve as a feedforward term; the outer loop takes the quality deviation degree as an input and zero as a set value, and obtains a temperature correction amount as a feedback term through PID operation; the feedforward term and the feedback term are added to obtain the adaptive temperature set value.

5. The reaction vessel control method for perfume production according to claim 4, characterized by, The adaptive temperature set value satisfies the expression: ； ； wherein, represents an adaptive temperature setting value at time ; represents a reaction time; represents a value of a reference temperature curve at time ; represents a temperature correction amount calculated by the outer loop at time ; represents a proportional gain of the outer loop; represents an integral gain of the outer loop; represents a differential gain of the outer loop; represents a quality deviation degree at time ; represents an integral of the quality deviation degree with respect to time ; represents a first derivative of the quality deviation degree with respect to time.

6. The reaction vessel control method for perfume production according to claim 1, wherein, The establishment of the soft sensor model comprises: Using a chemometrics algorithm to train the soft sensor model according to the spectral data and the real concentrations.

7. The reaction vessel control method for perfume production according to claim 1, characterized by, The method for obtaining the real concentrations of the key components comprises: Physically sampling at key time points during the operation of the golden batch, and obtaining the real concentrations by using offline high-precision analysis.

8. The reaction vessel control method for perfume production according to claim 6, characterized by, The chemometrics algorithm is a partial least squares method.

9. The reaction vessel control method for perfume production according to claim 1, characterized by, The real-time component vector satisfies the expression: ； wherein, represents a real-time component vector at time ; represents a soft-sensor model; represents a real-time collected spectrum data at time ; represents a reaction time.

10. A reaction vessel control system for perfume production, characterized by, The method comprises the following steps: A processor and a memory, the memory stores computer program instructions, when the computer program instructions are executed by the processor, the method for controlling the reaction kettle for perfume production according to any one of claims 1-9 is realized.

Images