A method and system for adaptive control of a multivariable coupled puffing process

Abstract

The present application relates to the technical field of extrusion puffing, and particularly relates to a multi-variable coupled puffing process self-adaptive control method and system, which comprises real-time collection of multi-dimensional operation parameters of the puffing production process, the operation parameters comprising puffing machine multi-section barrel temperature, cavity internal working pressure, screw operation rotating speed, material feeding flow, material feeding moisture content, die outlet discharging speed and discharging puffing degree quality parameters; based on the collected operation parameters, grey correlation analysis and partial least square regression algorithm are adopted. The present application firstly constructs a multi-variable dynamic coupling kinetics model of the puffing process, quantifies the coupling correlation weight of each process parameter, compensates parameter linkage deviation in real time through a self-adaptive decoupling control algorithm, breaks the technical limitation of traditional single-loop independent regulation, completely solves the coupling interference problem of "temperature regulation disturbs pressure, speed regulation disturbs material, flow regulation disturbs temperature", greatly improves the control precision of each process parameter, and significantly optimizes the working condition operation stability.

Description

Technical Field

[0001] This invention relates to the field of extrusion puffing technology, and in particular to an adaptive control method and system for a multivariable coupled puffing process. Background Technology

[0002] Extrusion puffing technology is a core process in the food, feed, and grain and oil deep processing industries. Utilizing a closed working environment with high temperature, high pressure, and high shear, it achieves material maturation, puffing, and modification, and is widely used in the production of puffed snacks, pet food, and deep-processed grains. The puffing production process is a typical complex industrial process characterized by strong multivariate coupling, nonlinearity, large time lag, and time-varying parameters. Core control variables include multi-section barrel temperature, cavity pressure, screw speed, material feed moisture content, feed flow rate, and die outlet velocity. These variables are deeply dynamically coupled, and adjusting a single parameter can trigger multi-dimensional fluctuations in operating conditions, making it a significant challenge for industry control.

[0003] Current puffing production control technologies generally suffer from numerous shortcomings, making it difficult to meet the demands for high-precision, stable, and intelligent production. Firstly, existing control schemes mostly employ single-loop independent PID control modes, individually regulating parameters such as temperature, pressure, and speed, completely ignoring the coupling interference characteristics between variables. For example, increasing screw speed increases material shear force and intracavity pressure, leading to a passive increase in barrel temperature; fluctuations in material moisture content simultaneously affect puffing pressure, degree of ripening, and discharge bulkiness. Single-variable control methods are prone to cascading deviations of "temperature adjustment causing pressure disturbance, speed adjustment causing temperature disturbance," resulting in operational imbalances.

[0004] Secondly, traditional control models are fixed and have poor adaptability, making them unable to adapt to dynamic changes in production conditions. During the puffing process, batch differences in raw materials, fluctuations in ambient temperature and humidity, and wear and tear of equipment due to long-term operation can all cause real-time deviations in the optimal operating point of process parameters. Existing fixed-parameter control algorithms cannot identify model disturbances and parameter drifts online, and can only rely on manual experience to periodically correct parameters. This results in a strong lag in regulation and easily leads to quality problems such as uneven puffing degree, texture deviation, and loss of heat-sensitive nutrients, resulting in large fluctuations in product yield.

[0005] Third, it suffers from weak anti-interference capabilities and low steady-state accuracy, making it impossible to achieve multi-objective collaborative optimization. Existing technologies struggle to simultaneously address multiple production objectives, such as puffing uniformity, nutrient retention, production energy consumption, and equipment operational stability. Under extreme conditions of high temperature and pressure, conventional control strategies cannot counteract disturbances caused by thermal inertia, mechanical lag, and changes in material rheological properties, easily leading to problems such as large batch-to-batch product quality differences, high energy consumption, and equipment overload and wear. Furthermore, traditional control methods heavily rely on operator experience, resulting in frequent manual intervention, low automation levels, and difficulty in adapting to the demands of large-scale, standardized, and intelligent modern puffing production.

[0006] In summary, current puffing process control technology suffers from core technical defects such as lack of multivariable coupling and decoupling capabilities, failure of parameter adaptive adjustment, poor dynamic anti-interference ability, lack of multi-objective collaborative control, and low level of intelligence. There is an urgent need for a puffing process control scheme that can accurately identify multivariable coupling relationships, adaptively correct control parameters online, and achieve dynamic and stable regulation under all operating conditions. Summary of the Invention

[0007] This invention addresses the core deficiencies of existing technologies by proposing an adaptive control method and system for a multivariable coupled puffing process. The core innovative logic is as follows: first, modeling and decoupling; then, online identification; finally, adaptive regulation and multi-objective collaborative optimization. By constructing a multivariable coupled dynamic model of the puffing process, the coupling correlation weights of each process variable are accurately quantified. Combined with online parameter identification and adaptive sliding mode control algorithms, the interference of variable coupling, operating condition disturbances, and model parameter drift are dynamically offset, achieving high-precision, lag-free, adaptive and stable control of the entire puffing process.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an adaptive control method for a multivariable coupled puffing process, comprising the following steps: S1. Real-time acquisition of multi-dimensional operating parameters of the extrusion production process, including multi-section barrel temperature of the extruder, working pressure inside the cavity, screw speed, material feed flow rate, material feed moisture content, die outlet speed and extrusion degree quality parameters. S2. Based on the collected operating parameters, the dynamic coupling correlation weights between each parameter are quantified through grey relational analysis and partial least squares regression algorithm. A nonlinear multivariate coupled dynamic model of the puffing process is constructed to characterize the coupling disturbance law of each process parameter under different working conditions. S3. Based on the recursive least squares algorithm, online parameter identification is performed on the multivariable coupled dynamic model, and the model parameter drift and operating condition disturbance caused by raw material fluctuations, changes in environmental temperature and humidity, and equipment aging are captured in real time, and the real-time identified model parameters are output. S4. Based on the identified model parameters and coupling correlation weights, the control parameters of each control loop are dynamically corrected through an improved adaptive sliding mode decoupling control algorithm to offset multivariable coupling interference and operating condition lag deviation, thereby achieving multi-loop coordinated decoupling control of temperature, pressure, speed, and flow. S5. Construct a multi-objective optimization function with puffing uniformity, nutrient retention rate, production energy consumption, and equipment steady-state operation as the core. Combine the online feedback data of output quality to iteratively optimize the control parameters and complete the full closed-loop adaptive control of the puffing process.

[0009] Preferably, the method for constructing the multivariable coupled dynamic model in step S2 includes: The collected raw operating parameters are preprocessed by filtering and noise reduction and outlier removal. Based on the preprocessed parameter data, the grey correlation degree between any two process parameters is calculated to quantify positive coupling, negative coupling, weak coupling, and strong coupling relationships. The nonlinear coupling equation between parameters is fitted by partial least squares regression to establish a coupling dynamic model that changes dynamically with the operating conditions, and the coupling disturbance compensation coefficient of each parameter is output in real time.

[0010] Preferably, the online parameter identification process in step S3 includes: The model parameters are set with initial thresholds and operating condition disturbance thresholds. The coefficient parameters of the coupled model are updated in real time through recursive least squares algorithm. When fluctuations in raw material moisture content, ambient temperature, or equipment speed deviation are detected to exceed the preset thresholds, they are identified as operating condition disturbances. The disturbance identification signal and model correction parameters are output in real time to achieve dynamic adaptive updating of model parameters.

[0011] Preferably, the control logic of the improved adaptive sliding mode decoupling control algorithm in step S4 includes: A decoupling compensation matrix is ​​constructed based on the coupling correlation weights. The sliding surface parameters and the approach law parameters are dynamically adjusted according to the real-time operating condition deviation to suppress system chattering. For the large lag and nonlinear characteristics of the expansion process, a lag compensation factor is introduced to eliminate the lag error of temperature and pressure control and achieve multi-variable uncoupled precise control.

[0012] Preferably, the multi-objective optimization function is constructed in step S5 as follows: Based on product production requirements, dynamic weighting coefficients are set for puffing uniformity, nutrient retention rate, production energy consumption, and equipment steady-state operation, respectively, and a weighted multi-objective optimization function is constructed. The optimal control parameters are solved iteratively through gradient descent algorithm to achieve synergistic optimization with the lowest production energy consumption and the most stable equipment operation while ensuring product quality and nutrient retention.

[0013] The control method for executing any one of claims 1-5 includes a data acquisition module, a coupling modeling module, an online identification module, an adaptive decoupling control module, a multi-objective optimization module, and a closed-loop feedback control module. The data acquisition module is used to collect real-time operating data of the extruder, including multi-segment temperature, cavity pressure, screw speed, feed flow rate, material moisture content, and discharge extrusion degree, and to complete data preprocessing. The coupling modeling module is electrically connected to the data acquisition module and is used to quantify the multi-parameter coupling correlation weights based on preprocessed data to construct a nonlinear multivariable coupling dynamics model. The online identification module is electrically connected to the coupled modeling module and is used to identify model parameter drift and external operating condition disturbances in real time, and output the iterated model parameters. The adaptive decoupling control module is electrically connected to the online identification module and is used to dynamically correct the control parameters based on the identification parameters and coupling weights through an adaptive sliding mode decoupling algorithm to counteract coupling interference. The multi-objective optimization module is electrically connected to the adaptive decoupling control module and is used to construct a multi-objective optimization function and iteratively optimize the control strategy. The closed-loop feedback control module is electrically connected to the multi-objective optimization module and the data acquisition module, respectively. It is used to receive optimized control commands, regulate each actuator of the extruder, and provide real-time feedback on the output quality and operating data to form a closed-loop control.

[0014] Preferably, the data acquisition module includes a temperature sensor, a pressure sensor, a speed encoder, a flow sensor, an online moisture detector, and an online puffing degree detection unit. Each detection unit collects data synchronously in real time, with a sampling frequency of not less than 100Hz.

[0015] Preferably, the adaptive decoupling control module has a built-in decoupling compensation matrix and a lag compensation unit to eliminate multi-variable coupling interference and system regulation lag error, thereby achieving multi-loop coordinated and precise regulation.

[0016] Preferably, the system further includes an anomaly warning module, which is used to monitor deviations in operating parameters, model identification anomalies, and equipment operation anomalies in real time. When the parameters exceed the preset threshold, the module automatically issues an early warning signal and executes parameter protection and control strategies.

[0017] Preferably, the computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the multivariable coupled adaptive control method for the expansion process according to any one of claims 1-5. By employing the above technical solution, the present invention provides an adaptive control method and system for a multivariable coupled puffing process, which has at least the following beneficial effects: 1. This invention is the first to construct a multivariate dynamic coupling dynamic model of the puffing process, quantifies the coupling correlation weight of each process parameter, and compensates for parameter linkage deviation in real time through an adaptive decoupling control algorithm. It breaks through the technical limitations of traditional single-loop independent control, and completely solves the coupling interference problems of "temperature regulation disturbing pressure, speed regulation disturbing material, and flow regulation disturbing temperature". The control accuracy of each process parameter is greatly improved and the stability of the working condition is significantly optimized.

[0018] 2. By using online parameter identification technology, the system can capture model parameter drift caused by batch differences in raw materials, environmental fluctuations, and equipment aging in real time. Without the need for manual modification of process parameters, the system can automatically complete the iterative correction of control parameters, effectively solving the problems of poor adaptability and control failure after the operating condition deviates in traditional fixed control models. It is adaptable to puffing production scenarios with different raw materials, different production capacities, and different equipment states, and has extremely strong versatility and robustness.

[0019] 3. Through high-precision dynamic adaptive control, stable and accurate control of puffing temperature, pressure, and speed is achieved, avoiding problems such as uneven puffing, clumping, undercooking, and overcooking caused by drastic fluctuations in working conditions, and significantly improving the consistency of product batches; at the same time, the duration of high temperature and high pressure can be precisely controlled, effectively reducing the degradation and loss of heat-sensitive nutrients, greatly improving the nutritional value and yield of products, and reducing the cost of defective products.

[0020] 4. This invention achieves optimal multi-parameter coordinated control, avoiding the ineffective energy waste caused by traditional over-adjustment and lag control, accurately matching the energy consumption requirements of production conditions, and effectively reducing production energy consumption; at the same time, it avoids equipment shock and overload operation caused by drastic parameter fluctuations, reduces wear on core components such as screws and barrels, reduces the probability of equipment failure and maintenance costs, and extends the service life of extrusion equipment.

[0021] 5. This invention achieves automatic sensing, autonomous identification, adaptive control, and closed-loop optimization throughout the entire puffing production process, eliminating the need for real-time monitoring and parameter correction by professional operators. This significantly reduces the frequency of manual intervention and labor costs, completely freeing production from the high dependence on operator experience. It is suitable for large-scale, standardized, and intelligent industrial production scenarios, resulting in a significant improvement in production efficiency.

[0022] 6. This invention constructs a multi-objective collaborative optimization mechanism, which can dynamically balance multiple objectives such as product quality, nutrient retention, production efficiency, and energy consumption control according to production needs. This solves the shortcomings of traditional single-parameter optimization, which often results in neglecting one aspect for another, and maximizes the comprehensive benefits of puffing production. Attached Figure Description

[0023] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a block diagram of the multivariable coupled adaptive control system for the puffing process of the present invention. Figure 2 This is a flowchart of the adaptive control method for the multivariable coupling puffing process of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Example 1 Please refer to Figures 1-2 This embodiment uses a mass-production twin-screw extruder, suitable for extruding and processing powdered materials such as corn flour, rice flour, and mixed grain powder. The equipment is divided into a feeding mechanism, a three-section heating barrel, a screw extrusion mechanism, and a die outlet mechanism. It is equipped with online detection equipment and an industrial controller. The entire system supports automated data acquisition and real-time control. The feeding mechanism is equipped with a variable frequency speed control screw feeder, which can achieve uniform and continuous material conveying and eliminate problems such as intermittent feeding, material bridging, and feed deviation. The three-section heating barrel adopts an independent cast aluminum heating ring and air-cooled heat dissipation structure. Each section of the barrel is equipped with a high-precision temperature sensor and temperature control module to achieve precise temperature control in zones, with a temperature control accuracy of ±0.5℃. The screw extrusion mechanism adopts a wear-resistant alloy twin-screw structure with a screw length-to-diameter ratio of 18:1, which is suitable for extrusion and melting processing of grain powders. It is wear-resistant, durable, and has a long service life. The die outlet mechanism is equipped with a detachable and customized die head, which can be replaced with different diameter dies according to product specifications to meet the forming requirements of different extruded products. The supporting online testing equipment consists of industrial-grade high-precision testing devices, which are suitable for the complex production environment of high temperature, high dust and strong vibration in the puffing workshop. The industrial controller uses a PLC programmable controller, which supports multi-channel data synchronous acquisition, algorithm calculation and real-time command output, with fast response speed and strong anti-interference ability.

[0026] In this embodiment, the default initial reference production working conditions are set as follows: the extruder has three sections of independently temperature-controlled barrels, wherein the temperature of the barrel in the first-stage feeding zone is 85°C, the temperature of the barrel in the second-stage melting zone is 120°C, and the temperature of the barrel in the third-stage voltage-stabilizing extrusion zone is 155°C; the reference value of working pressure in the expansion cavity is 3.8MPa; the reference value of working rotating speed of the screw is 280r / min; the reference value of material feeding flow is 120kg / h; the reference value of material feeding moisture content is 18%; the reference value of die outlet discharging speed is 0.8m / s. The production target quality indicators require that: the discharging expansion uniformity is not less than 96%, and the material nutrient retention rate is not less than 92%. The above reference working condition is the optimal working condition for standardized production of corn and rice mixed powder (ratio 1:1), and can be adaptively fine-tuned according to single material and differentiated material ratio. The low-temperature working condition in the feeding zone can preheat and loosen the material, avoid powder material caking at high temperature, and ensure that the material smoothly enters the melting zone; the medium-temperature working condition in the melting zone provides suitable temperature conditions for grain starch gelatinization and protein modification, and ensures sufficient melting and plasticization of the material; the high-temperature and high-pressure working condition in the voltage-stabilizing extrusion zone can realize uniform expansion and shaping of the material, and guarantee the discharging regularity and expansion consistency. The corresponding quality indicators are the qualification standards for industrial mass production. The expansion uniformity is calculated through statistical calculation of the expansion ratio deviation of discharged particles, and the nutrient retention rate is calculated based on the retention ratio of vitamins, dietary fiber and crude protein of the material before and after processing.

[0027] All online detection sensors and detection units matched with the present invention are uniformly set with a sampling frequency of 100Hz, which can realize millisecond-level synchronous data acquisition, ensure real-time, continuous and synchronous updating of working condition data, and provide a reliable data basis for subsequent modeling, identification and regulation. The fixed 100Hz sampling frequency is determined through optimization of a large number of working condition tests, giving consideration to data acquisition accuracy and system operation load. It can not only capture millisecond-level instantaneous working condition fluctuations in the production process and avoid missing high-frequency disturbance data, but also avoid controller operation redundancy and system lag caused by excessively high sampling frequency. All detection units adopt a time-sequence synchronous triggering mechanism, and unify clock signals through an industrial bus, ensuring that the acquisition time stamps of all parameters such as temperature, pressure, rotating speed, flow, moisture content and expansion degree are completely aligned, and completely eliminating the problems of data错位 and time-sequence confusion caused by asynchronous acquisition of multiple parameters.

[0028] S1, Real-time acquisition and data preprocessing of multi-dimensional operating parameters in production process During the normal production operation of the extruder, this system continuously collects key operating and quality parameters throughout the entire production process in real time. These parameters include: real-time temperature values ​​of each of the three barrel sections, real-time working pressure within the extrusion chamber, real-time screw speed, real-time material feed flow rate, real-time material feed moisture content, real-time die discharge speed, and real-time extrusion degree. All these parameters are collected synchronously through corresponding dedicated detection units to avoid data misalignment caused by delays in single-parameter acquisition. The specific configuration of the dedicated detection units for each parameter is as follows: PT100 industrial platinum resistance temperature sensors are used for all three barrel sections, with a temperature range of 0-200℃ and a measurement accuracy of ±0.1℃. These sensors are installed flush against the inner wall of the barrel to accurately collect the actual working temperature of the chamber. A high-temperature piezoelectric pressure sensor is used for the extrusion chamber pressure, with a range of 0-10MPa and an accuracy class of 0.3. This sensor is directly embedded inside the extrusion chamber to collect the extrusion pressure in real time. An incremental photoelectric speed encoder is used for the screw speed. With a resolution of 1000 lines, it accurately captures the real-time rotational speed and fluctuations of the screw; the feed flow rate uses a spiral weighing flow sensor to measure the mass flow rate of the material in real time, suitable for powder material metering scenarios; the material moisture content uses a near-infrared online moisture detector to detect the dryness and wetness of the feed material in real time without contact, with a detection accuracy of ±0.2%; the die outlet discharge speed uses a laser speed sensor to detect the running speed of the discharged material without contact; the discharge expansion degree uses a machine vision online detection unit, which uses high-definition cameras combined with image recognition algorithms to calculate the expansion ratio and uniformity of the discharged particles in real time, realizing online quantitative acquisition of quality parameters.

[0029] The raw data collected on-site is affected by factors such as equipment mechanical vibration, workshop electromagnetic interference, and instantaneous feed fluctuations, resulting in noise and abnormal abrupt changes. Therefore, the system needs to preprocess the raw data, mainly including two operations: filtering and noise reduction, and outlier removal. This embodiment adopts the commonly used engineering moving average filtering method to smooth data fluctuations and retain the actual operating condition trend. At the same time, a reasonable fluctuation range of parameters is set to directly remove abnormal abrupt data that exceeds the normal fluctuation range, avoiding the introduction of erroneous data into subsequent modeling calculations and ensuring the accuracy and validity of the input data. The moving average filtering is set with a sliding window length of 10 sampling points, targeting a sampling frequency of 100Hz, corresponding to a data smoothing period of 0.1s. This can effectively filter high-frequency noise generated by high-frequency mechanical vibration of equipment and electromagnetic interference from workshop frequency converters, while completely retaining the effective data trends of manual parameter adjustment, raw material fluctuations, and gradual changes in equipment operating conditions during the production process, avoiding the loss of operating condition characteristics due to over-filtering. The reasonable fluctuation range of parameters is set based on the benchmark operating conditions and the normal operating error of the equipment. Specifically, it is as follows: barrel temperature fluctuation ±1℃, cavity pressure fluctuation ±0.2MPa, screw speed fluctuation ±3r / min, feed flow rate fluctuation ±2kg / h, and material moisture content fluctuation ±0.5%. Single-point abrupt changes outside this range are judged as invalid abnormal data. The system automatically marks and removes them. After removal, the data sequence is completed by interpolation of the valid data before and after to ensure data continuity and provide a complete and accurate dataset for subsequent modeling.

[0030] S2. Construct a nonlinear multivariable coupled dynamic model of the puffing process. In the extrusion production process, the various process parameters do not change independently; adjusting any one parameter will cause a coordinated change in other parameters, a phenomenon known as parameter coupling. To accurately grasp the linkage patterns between these parameters, this invention uses preprocessed effective operating data and employs grey relational analysis combined with partial least squares regression to complete the coupling modeling.

[0031] First, using a grey relational analysis algorithm, the grey relational degree between any two process parameters is calculated one by one. The coupling type is distinguished based on the magnitude of the relational degree, clarifying the linkage characteristics of each parameter. In the actual working conditions of this embodiment, an increase in barrel temperature directly increases the extrusion pressure within the cavity, and an increase in the material feed moisture content significantly improves the uniformity of the output expansion. Both sets of parameters show a clear positive trend, indicating a strong positive coupling relationship. An increase in screw speed shortens the material extrusion heating time, leading to an increase in the output moisture content. An increase in feed flow rate increases the instantaneous extrusion pressure within the cavity. These parameter changes mutually restrict each other, indicating a strong negative coupling relationship. Meanwhile, the ambient temperature and humidity of the workshop and minor mechanical vibrations of the equipment have minimal impact on the core production parameters, indicating a weak coupling relationship. This step allows for the complete quantification of the positive, negative, weak, and strong coupling correlation weights of all parameters. The specific operational parameters for the grey relational analysis algorithm in this embodiment are set as follows: 1000 sets of preprocessed effective steady-state operating condition data are selected as analysis samples, and the resolution coefficient is set to 0.5 (an industry-standard optimal value that balances correlation resolution and stability). Dimensionless normalization is performed on all core process parameters to eliminate calculation interference from parameters of different dimensions such as temperature, pressure, speed, and flow rate. A correlation threshold is set: correlation ≥ 0.7 is considered strong coupling, 0.3 ≤ correlation < 0.7 is considered weak coupling, and correlation < 0.3 is considered no significant coupling. Quantitative calculations show that the correlation between barrel temperature and internal pressure is 0.86, and the correlation between feed moisture content and expansion uniformity is 0.82, indicating typical strong positive coupling. The correlation between screw speed and discharge moisture content is -0.81, and the correlation between feed flow rate and dynamic fluctuations in internal pressure is -0.79, indicating typical strong negative coupling. The correlations between ambient temperature and humidity, equipment vibration, and core process parameters are all below 0.4, indicating weak coupling, which can mitigate the impact of interference during precise control. Finally, precise coupling weight coefficients for each parameter are output, providing a quantitative basis for subsequent decoupling control.

[0032] Subsequently, partial least squares regression was used to fit multiple sets of operating data. This algorithm can effectively avoid the problem of multivariate collinearity and adapt to the nonlinear variation characteristics of the puffing process, fitting nonlinear coupling equations between various process parameters. Combining the operating condition changes caused by different equipment loads, raw material moisture differences, and equipment operating time, a multivariate coupled dynamic model that can be dynamically updated with the operating conditions was established. The model can output the coupling disturbance compensation coefficients corresponding to each parameter in real time, accurately depicting the specific laws of mutual interference and fluctuation disturbances of various process parameters under different production conditions, providing model support for subsequent decoupling control. Compared with the traditional multiple linear regression algorithm, this algorithm can effectively solve the technical problems of high multivariate collinearity, strong nonlinearity, and time-varying operating conditions in the puffing process, without the need for pre-screening of variables, and can achieve fitting of full parameter coupling relationships. This embodiment selects 2000 sets of operating condition data under different loads (80%-100% rated capacity), different moisture contents (16%-20%), and different equipment operating times (0-2000h) as the training set, and 300 sets of data as the test set. The model fits a coefficient of determination R² ≥ 0.98, indicating extremely high fitting accuracy. After the model is built, the equation coefficients can be dynamically updated based on real-time operating condition data, and the cross-disturbance compensation coefficients between temperature, pressure, speed, and flow rate can be output in real time. This accurately quantifies the interference amplitude of a certain parameter fluctuation on other parameters, completely solving the defect of traditional fixed models that cannot adapt to time-varying operating conditions.

[0033] S3. Online parameter identification of models based on recursive least squares algorithm In actual production, issues such as batch differences in raw materials, variations in workshop temperature and humidity, long-term screw wear from equipment operation, and component aging can all lead to deviations in the parameters of the previously established fixed model, i.e., model parameter drift. Using the old model parameters will result in a decrease in control accuracy. Therefore, this invention employs a recursive least squares algorithm to identify and update the parameters of the coupled dynamics model online and in real time.

[0034] In this embodiment, fixed operating condition disturbance thresholds and initial thresholds for model parameters are pre-set. Specifically, the thresholds are: allowable fluctuation range of raw material moisture content ±1.2%, allowable fluctuation range of ambient temperature ±3℃, and allowable deviation range of equipment speed ±5 r / min. During equipment operation, the system relies on a recursive least squares algorithm to iteratively update the coefficient parameters of the coupled model every second, achieving adaptive model correction. The above thresholds are the optimal fault-tolerant thresholds for grain puffing production, which can avoid system oscillations caused by frequent model updates triggered by minor operating condition fluctuations, and can also promptly capture effective operating condition disturbances and complete model correction. The recursive least squares algorithm initialization parameters are set as follows: the forgetting factor is set to 0.98 to ensure a reasonable weight distribution between old and new operating condition data, preserving the characteristics of historical effective operating condition models while highlighting the corrective role of real-time new operating condition data, adapting to the operating conditions of slow equipment aging and gradual raw material changes; the initial covariance matrix is ​​set to an identity matrix to ensure stability in the early stages of algorithm iteration. The system collects real-time operating condition data every second with an iteration cycle of 1 second to complete the iterative update of model coefficients and parameter identification.

[0035] When the system detects fluctuations in key parameters such as raw material moisture content, ambient temperature, and screw speed exceeding the preset thresholds, or detects model parameter drift caused by long-term equipment operation, the system directly determines that a disturbance has occurred in the current production condition, immediately outputs a condition disturbance identification signal, and synchronously updates the model correction parameters. This completes the dynamic adaptive update of model parameters, ensuring that the model always adapts to the current real production condition and avoids control failure caused by model lag. A model parameter drift judgment mechanism is set: when the relative error between the model's predicted output value and the actual detected value exceeds 3% for 5 consecutive seconds, it is determined to be a model parameter drift, and a forced identification update is immediately triggered. The system distinguishes between normal small fluctuations and large condition disturbances. Fluctuations within a small threshold only undergo minor adaptive correction of the model, while large disturbances exceeding the threshold initiate a full parameter iterative update. At the same time, the disturbance type, disturbance amplitude, and correction parameters are recorded to form a condition disturbance database, continuously optimizing the model identification accuracy and fundamentally solving the problem of control accuracy decay after long-term operation of traditional fixed models.

[0036] S4, Improved Adaptive Sliding Mode Decoupling Control with Multi-Loop Coordinated Regulation After completing the real-time identification and update of model parameters, the system uses an improved adaptive sliding mode decoupling control algorithm based on the latest model parameters and the coupling weights of each parameter to dynamically adjust each control loop and counteract multivariate coupling interference.

[0037] First, a decoupling compensation matrix is ​​constructed based on the coupling weights of each parameter to specifically counteract coupling interference between different parameters. Traditional sliding mode control algorithms are prone to system chattering, causing equipment instability. This invention dynamically adjusts the sliding surface parameters and reaching law parameters according to the real-time operating condition deviation, effectively suppressing high-frequency chattering in the control system and significantly improving the stability of equipment regulation. Based on the coupling weights of each parameter obtained from grey relational analysis, a 4×4 decoupling compensation matrix for temperature, pressure, speed, and flow is constructed. The matrix coefficients correspond one-to-one with the coupling weights, accurately matching the coupling interference intensity of different parameters to achieve targeted decoupling compensation. Strongly coupled parameters correspond to large compensation coefficients, and weakly coupled parameters correspond to small compensation coefficients, avoiding over-regulation or under-compensation caused by uniform compensation. Addressing the drawback of traditional sliding mode control's fixed parameters easily causing system chattering, this invention optimizes the design of an adaptive reaching law. When the operating condition deviation is large, the reaching speed is increased to quickly eliminate the deviation; when the deviation approaches 0, the reaching speed is decreased to weaken the high-frequency chattering caused by sliding mode switching, controlling the steady-state chattering amplitude within the allowable fluctuation range of the parameters, thus significantly improving the stability of equipment operation.

[0038] Meanwhile, the temperature rise and fall, and pressure adjustment of the extruder exhibit significant lag, representing a typical large-lag control scenario. This algorithm specifically introduces a lag compensation factor to accurately compensate for the lag deviations in temperature and pressure regulation, completely eliminating mutual interference between the control loops for temperature, pressure, speed, and flow. It achieves independent, precise, and coordinated decoupled regulation of the four core control loops, solving the problem that traditional single-loop control cannot simultaneously address multi-parameter coupling interference and lag deviations. The temperature regulation lag time of this extrusion system is 3-5 seconds, and the pressure regulation lag time is 2-4 seconds, classifying it as a typical large-lag nonlinear control object. This embodiment calculates the lag time constant of each loop through step response tests and sets a targeted dynamic lag compensation factor. The compensation factor adaptively fine-tunes with changes in operating temperature and pressure, offsetting regulation lag errors in real time. Through dual optimization of the decoupling matrix and lag compensation, the coupling relationship of the four core control loops is completely eliminated. Each loop can independently receive regulation commands and independently complete parameter tuning without interference. The regulation response speed is improved by more than 60%, and the steady-state control accuracy is improved to within ±0.3℃ and ±0.1MPa.

[0039] S5. Multi-objective optimization iteration and full closed-loop adaptive control Extrusion production not only needs to ensure the quality of the output material, but also needs to consider the retention of nutrients in the material, production energy consumption, and equipment operational stability, which is a multi-objective collaborative optimization problem. This invention uses four indicators as the core—extrusion uniformity, nutrient retention rate, production energy consumption, and equipment steady-state operation—to construct a multi-objective optimization function.

[0040] To adapt to different production needs, this invention sets dynamically adjustable weight coefficients for the four indicators to suit different production scenarios. In high-quality production mode, priority is given to ensuring the quality of the finished product, with the weights for puffing uniformity set at 0.35, nutrient retention rate at 0.35, production energy consumption at 0.15, and equipment steady-state operation at 0.15. In batch production and energy-saving production modes, the weights are adjusted to focus on energy saving and stable production, with the weights for production energy consumption adjusted to 0.3, equipment steady-state operation at 0.25, and the total weight of quality indicators adjusted accordingly to 0.45. By setting differentiated weights to adapt to different production needs, in addition to the two core production modes mentioned above, this system supports custom weight ratios. Users can manually fine-tune the weight coefficients of the four indicators according to their personalized needs such as special product production, trial production and debugging, and high and stable production to adapt to the production needs of all scenarios. The quantitative calculation methods for each indicator are as follows: the uniformity of puffing is quantified by the standard deviation of the particle expansion ratio detected by machine vision. The smaller the standard deviation, the higher the uniformity. The nutrient retention rate is calculated by detecting the retention content of soluble protein, dietary fiber, and B vitamins before and after material processing. The production energy consumption is statistically analyzed in real time by the power consumption per unit output of the equipment through the power acquisition module. The steady-state operation of the equipment is quantified by the comprehensive variance of screw speed fluctuation, pressure fluctuation, and temperature fluctuation. The smaller the variance, the more stable the equipment operation.

[0041] The system employs a gradient descent algorithm to iteratively solve a multi-objective optimization function, continuously receiving online feedback data on the degree of puffing of the output material, nutrient loss, equipment operating load, and real-time production energy consumption. It continuously iterates and optimizes the control parameters of each control loop. Ultimately, it achieves the following: while ensuring stable finished product puffing quality and sufficient nutrient retention, it minimizes production energy consumption and improves equipment operational stability, forming a fully closed-loop adaptive control process from data acquisition, modeling and identification, decoupled control, quality feedback, and parameter optimization. The gradient descent algorithm has an initial learning rate of 0.01, with adaptively adjusted iteration step size and an iteration convergence threshold of 0.001. When the change in the output value of the multi-objective optimization function is less than the convergence threshold, the parameters are determined to be optimal, iteration stops, and the current control parameters are locked. If the operating conditions continue to fluctuate, the system continues to iterate and optimize, updating the optimal control parameters in real time. The system uses a 2-second optimization iteration cycle to continuously receive feedback data from the entire process, forming a complete closed-loop control link of "data acquisition - preprocessing - coupled modeling - parameter identification - decoupled control - quality feedback - multi-objective optimization - parameter update". This enables unmanned adaptive optimization of the production process, which can reduce overall energy consumption by 8%-12% and reduce equipment failure downtime by more than 15% in long-term production.

[0042] Example 2 Please refer to Figures 1-2This embodiment is basically the same as Embodiment 1. This embodiment is made on the basis of Embodiment 1 and has the same beneficial effects as Embodiment 1. The same parts can be referred to each other, and will not be described in detail here.

[0043] The control system of the present invention is used to fully execute the above-mentioned adaptive control method. It consists of a data acquisition module, a coupling modeling module, an online identification module, an adaptive decoupling control module, a multi-objective optimization module, a closed-loop feedback control module, and an anomaly early warning module. Each module has a clear division of labor, links up sequentially, and works in coordination.

[0044] The data acquisition module is equipped with high-precision temperature sensors, pressure sensors, speed encoders, flow sensors, online moisture detectors, and online puffing degree detection units. All detection units collect data synchronously at a stable sampling frequency of 100Hz. It is responsible for collecting operating conditions and quality data throughout the entire production process, and performs filtering and outlier removal preprocessing before transmitting qualified data to the coupled modeling module. The data acquisition module is the core of the system's data input. All detection devices are industrially calibrated and adapted to the high-temperature, high-dust, and strong electromagnetic interference conditions of the puffing workshop. They are dustproof, waterproof, and vibration-resistant, with a protection level of at least IP65. The module has a built-in data storage buffer that can cache 10 seconds of raw data to prevent data loss due to data transmission delays. It also has an independent data preprocessing subroutine that implements the moving average filtering algorithm and outlier removal rules defined in this invention. This allows for independent raw data purification, uploading only valid and accurate standardized data to the upper-level module, reducing the computational load on the main control system and ensuring the real-time performance and stability of data transmission and processing.

[0045] The coupled modeling module receives preprocessed operating condition data, quantifies the coupling weights of each parameter, calculates parameter coupling relationships, fits nonlinear coupling equations, constructs a dynamically updated multivariate coupled dynamic model, and outputs coupling disturbance compensation coefficients in real time, providing a model foundation for subsequent control. The module has built-in grey relational analysis and partial least squares regression algorithms, requiring no external computing power, and can independently complete multi-dimensional parameter coupling analysis and model fitting. The module features model storage and iteration capabilities, accumulating coupled model parameters under different raw materials and operating conditions to form an operating condition model database. Historical models can be directly called during subsequent production under similar operating conditions, shortening model adaptation and iteration time and improving system response speed. Simultaneously, the module outputs visualized coupling weight data and disturbance compensation coefficients in real time, supporting manual viewing and tracing of operating condition coupling changes in the background.

[0046] The online identification module relies on the recursive least squares algorithm to iteratively update model coefficient parameters in real time. It continuously monitors model parameter drift and operating condition disturbances caused by raw material fluctuations, environmental changes, and equipment aging, and promptly outputs the latest corrected model parameters to ensure that the model matches the actual operating conditions. As the core unit of model adaptive optimization, the online identification module runs an independent model identification subroutine, operating in parallel with the production control process without interference. The module incorporates dual judgment logic for operating condition disturbance determination and model drift identification, accurately distinguishing between normal operating condition fluctuations and abnormal operating condition disturbances, avoiding invalid identification iterations. The module has parameter traceability and recording functions, automatically storing the time, disturbance type, and correction parameter magnitude of each model correction. This facilitates equipment maintenance personnel in analyzing equipment aging trends and raw material fluctuation patterns, providing data support for equipment maintenance and raw material pretreatment optimization.

[0047] The adaptive decoupling control module incorporates a decoupling compensation matrix and a hysteresis compensation unit. Combining the latest identification model parameters and coupling weights, it dynamically corrects the control parameters of each loop using an improved adaptive sliding mode decoupling algorithm. This suppresses system chattering, eliminates control hysteresis errors, and achieves precise, uncoupled control of multiple loops. As the core control execution unit of the system, the adaptive decoupling control module interfaces with the control ports of all actuators in the equipment and can independently perform real-time control of four major loops: temperature, pressure, speed, and flow. The module includes a built-in dynamic decoupling compensation matrix generation subroutine, which automatically updates the matrix coefficients based on real-time coupling weights, eliminating the need for manual parameter presets. The hysteresis compensation unit adaptively adjusts the compensation intensity according to the hysteresis characteristics of different operating conditions, perfectly adapting to the hysteresis control requirements of different production scenarios such as low temperature, high temperature, high load, and low load. Simultaneously, the module incorporates chattering suppression logic, monitoring system output fluctuations in real time and dynamically optimizing sliding mode control parameters to ensure smooth and shock-free operation of the equipment throughout the entire process.

[0048] The multi-objective optimization module dynamically configures the weights of each evaluation indicator based on the production scenario, constructs a weighted multi-objective optimization function, and iteratively solves for the optimal control parameters through a gradient descent algorithm to achieve synergistic optimization of product quality, nutrient retention, production energy consumption, and equipment stability. The module includes two preset production modes and a custom mode switching program, supporting one-click switching between high-quality mode and energy-saving production mode. It also provides a manual weight configuration port to adapt to niche and customized production needs. The module can visualize the real-time values ​​and optimization trends of the four optimization indicators, allowing production personnel to monitor the production status in real time. The entire optimization iteration process is controllable, with iteration pause and parameter locking functions to meet the manual intervention needs during equipment debugging and new product trial production.

[0049] The closed-loop feedback control module bidirectionally connects to the multi-objective optimization module and the data acquisition module, receives optimized control commands, and precisely controls the extruder's heating mechanism, feeding mechanism, screw operation mechanism, and other actuators. At the same time, it continuously collects the latest operating conditions and output quality data and transmits them back to the system, forming a closed-loop control cycle to achieve continuous adaptive optimization.

[0050] The system is additionally equipped with an anomaly early warning module, which monitors deviations in operating parameters, model identification anomalies, and abnormal equipment loads in real time. When any parameter or operating status exceeds a preset safety threshold, the system immediately triggers an audible and visual warning, and automatically executes parameter protection and control strategies to fine-tune equipment operating parameters, avoid production failures, reduce defective products, and ensure continuous and stable operation of the production line. The anomaly early warning module presets multi-dimensional safety thresholds, covering more than ten types of fault scenarios, including over-temperature, over-pressure, abnormal speed, flow interruption, excessive moisture content, model identification failure, equipment overload, and abnormal energy consumption. All thresholds have been verified through equipment safety tests and comply with industrial equipment safety operation standards. The early warning method adopts a triple mode of high-decibel audible and visual alarms, industrial control screen pop-up prompts, and fault code recording, accurately indicating the fault type and location. At the same time, a built-in hierarchical protection strategy is implemented: minor anomalies are automatically fine-tuned and repaired, moderate anomalies are warned and adaptively controlled, and severe anomalies are immediately shut down for protection, completely eliminating problems such as equipment damage, batch defects, and safety accidents.

[0051] Example 3 Please refer to Figures 1-2 This embodiment is basically the same as Embodiment 1. This embodiment is made on the basis of Embodiment 1 and has the same beneficial effects as Embodiment 1. The same parts can be referred to each other, and will not be described in detail here.

[0052] This invention also includes a corresponding computer-readable storage medium, which can be a USB flash drive, hard disk drive, solid-state drive, or a conventional industrial-readable storage device such as an industrial controller's built-in storage chip. The storage medium contains a computer program that can be recognized and executed by the industrial controller's processor. When the program runs, it can completely execute all the adaptive control method steps described above, automatically completing the entire process of data acquisition, coupling modeling, online parameter identification, adaptive decoupling control, multi-objective optimization, and closed-loop feedback regulation. It can be directly integrated into existing automated puffing equipment, demonstrating strong versatility.

[0053] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Since the above embodiments are substantially similar to the method embodiments, their descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0054] It should be noted that the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0055] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive control method for a multivariable coupled puffing process, characterized in that, Includes the following steps: S1. Real-time acquisition of multi-dimensional operating parameters of the extrusion production process, including multi-section barrel temperature of the extruder, working pressure inside the cavity, screw speed, material feed flow rate, material feed moisture content, die outlet speed and extrusion degree quality parameters. S2. Based on the collected operating parameters, the dynamic coupling correlation weights between each parameter are quantified through grey relational analysis and partial least squares regression algorithm. A nonlinear multivariate coupled dynamic model of the puffing process is constructed to characterize the coupling disturbance law of each process parameter under different working conditions. S3. Based on the recursive least squares algorithm, online parameter identification is performed on the multivariable coupled dynamic model, and the model parameter drift and operating condition disturbance caused by raw material fluctuations, changes in environmental temperature and humidity, and equipment aging are captured in real time, and the real-time identified model parameters are output. S4. Based on the identified model parameters and coupling correlation weights, the control parameters of each control loop are dynamically corrected through an improved adaptive sliding mode decoupling control algorithm to offset multivariable coupling interference and operating condition lag deviation, thereby achieving multi-loop coordinated decoupling control of temperature, pressure, speed, and flow. S5. Construct a multi-objective optimization function with puffing uniformity, nutrient retention rate, production energy consumption, and equipment steady-state operation as the core. Combine the online feedback data of output quality to iteratively optimize the control parameters and complete the full closed-loop adaptive control of the puffing process.

2. The adaptive control method for multivariable coupled puffing process according to claim 1, characterized in that, The method for constructing the multivariable coupled dynamic model in step S2 includes: The collected raw operating parameters are preprocessed by filtering and noise reduction and outlier removal. Based on the preprocessed parameter data, the grey correlation degree between any two process parameters is calculated to quantify positive coupling, negative coupling, weak coupling, and strong coupling relationships. The nonlinear coupling equation between parameters is fitted by partial least squares regression to establish a coupling dynamic model that changes dynamically with the operating conditions, and the coupling disturbance compensation coefficient of each parameter is output in real time.

3. The adaptive control method for multivariable coupled puffing process according to claim 1, characterized in that, The online parameter identification process described in step S3 includes: The model parameters are set with initial thresholds and operating condition disturbance thresholds. The coefficient parameters of the coupled model are updated in real time through recursive least squares algorithm. When fluctuations in raw material moisture content, ambient temperature, or equipment speed deviation are detected to exceed the preset thresholds, they are identified as operating condition disturbances. The disturbance identification signal and model correction parameters are output in real time to achieve dynamic adaptive updating of model parameters.

4. The adaptive control method for multivariable coupled puffing process according to claim 1, characterized in that, The control logic of the improved adaptive sliding mode decoupling control algorithm described in step S4 includes: A decoupling compensation matrix is ​​constructed based on the coupling correlation weights. The sliding surface parameters and the approach law parameters are dynamically adjusted according to the real-time operating condition deviation to suppress system chattering. For the large lag and nonlinear characteristics of the expansion process, a lag compensation factor is introduced to eliminate the lag error of temperature and pressure control and achieve multi-variable uncoupled precise control.

5. The adaptive control method for multivariable coupled puffing process according to claim 1, characterized in that, The method for constructing the multi-objective optimization function in step S5 is as follows: Based on product production requirements, dynamic weighting coefficients are set for puffing uniformity, nutrient retention rate, production energy consumption, and equipment steady-state operation, respectively, and a weighted multi-objective optimization function is constructed. The optimal control parameters are solved iteratively through gradient descent algorithm to achieve synergistic optimization with the lowest production energy consumption and the most stable equipment operation while ensuring product quality and nutrient retention.

6. An adaptive control system for a multivariable coupled puffing process, characterized in that, The control method for executing any one of claims 1-5 includes a data acquisition module, a coupling modeling module, an online identification module, an adaptive decoupling control module, a multi-objective optimization module, and a closed-loop feedback control module. The data acquisition module is used to collect real-time operating data of the extruder, including multi-segment temperature, cavity pressure, screw speed, feed flow rate, material moisture content, and discharge extrusion degree, and to complete data preprocessing. The coupling modeling module is electrically connected to the data acquisition module and is used to quantify the multi-parameter coupling correlation weights based on preprocessed data to construct a nonlinear multivariable coupling dynamics model. The online identification module is electrically connected to the coupled modeling module and is used to identify model parameter drift and external operating condition disturbances in real time, and output the iterated model parameters. The adaptive decoupling control module is electrically connected to the online identification module and is used to dynamically correct the control parameters based on the identification parameters and coupling weights through an adaptive sliding mode decoupling algorithm to counteract coupling interference. The multi-objective optimization module is electrically connected to the adaptive decoupling control module and is used to construct a multi-objective optimization function and iteratively optimize the control strategy. The closed-loop feedback control module is electrically connected to the multi-objective optimization module and the data acquisition module, respectively. It is used to receive optimized control commands, regulate each actuator of the extruder, and provide real-time feedback on the output quality and operating data to form a closed-loop control.

7. The multivariable coupled adaptive control system for the puffing process according to claim 6, characterized in that, The data acquisition module includes a temperature sensor, a pressure sensor, a speed encoder, a flow sensor, an online moisture analyzer, and an online puffing degree detection unit. Each detection unit collects data synchronously in real time, with a sampling frequency of not less than 100Hz.

8. The multivariable coupled adaptive control system for the puffing process according to claim 6, characterized in that, The adaptive decoupling control module has a built-in decoupling compensation matrix and hysteresis compensation unit, which are used to eliminate multivariate coupling interference and system regulation hysteresis error, and realize multi-loop coordinated precise regulation.

9. The multivariable coupled adaptive control system for the expansion process according to claim 6, characterized in that, The system also includes an anomaly warning module, which is used to monitor deviations in operating parameters, model identification anomalies, and equipment operation anomalies in real time. When the parameters exceed the preset threshold, it automatically issues an early warning signal and executes parameter protection and control strategies.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the multivariable coupled adaptive control method for the expansion process as described in any one of claims 1-5.

Images