A method for constructing a shoe press control system

Abstract

The present invention relates to the technical field of shoe presses, and specifically to a method for constructing a shoe press control system, which is characterized by comprising the following steps: (1) adopting an online system identification strategy based on a weak coupling model structure to solve the modeling difficulty problem caused by the complex nonlinearity, strong coupling and multi-input and multi-output characteristics of the dynamic behavior of the ultra-wide shoe press system; (2) finding solutions for multi-modal conversion, diverse control targets and complex constraints through a double-layer structure predictive control of centralized optimization and decentralized control. (3) using a partitioned controllable medium-high shoe press structure to adjust errors of different wavelengths, thereby eliminating most errors. The method for constructing a shoe press control system of the present invention constructs a partitioned controllable medium-high technology for ultra-wide shoe presses, and the partitions can be arbitrarily adjusted, thereby increasing the adjustment range and expanding the adjustment function.

Description

Technical Field

[0001] The invention relates to the technical field of shoe presses, and in particular to a method for constructing a shoe press control system. Background Art

[0002] The application of shoe press technology is a milestone revolution in the pressing and dehydration technology of the modern paper machine press section. In the process of dehydrating and pressing wet paper sheets on the paper machine, this technology is characterized by ultra-high line pressure (up to 1000kN / m), ultra-large dehydration area (pressing area width 200-300mm), long dehydration time, low pressing pressure ratio, and low dehydration resistance compared to the traditional high-efficiency vacuum pressing technology. Therefore, it has very outstanding advantages and strengths in the papermaking process. Paper machines using this technology have high speed, low energy consumption, and good quality of finished paper (especially good bulk). Therefore, this pressing technology and equipment have become the first choice or necessary equipment technology in the speed-up and energy-saving transformation of old paper machines and the design of new paper machines.

[0003] Among them, the core technology is the ultra-wide controllable medium and high shoe press system (such as Figure 1 As shown), compared with the traditional pressing technology (fixed medium and high pressure pressing, and vacuum pressing), this technology has controllable medium and high pressure, larger pressure area, wider width, more hydraulic cylinders required, and complex parameters. To improve the pressing quality and reduce losses, it is necessary to ensure the balance of line pressure, which requires highly coordinated, precise and adaptive control of each hydraulic subsystem.

[0004] Controllable medium-high boot press is a new papermaking pressing and dewatering technology. Its application has greatly improved the dewatering performance of paper machines under high-speed operation conditions. Compared with traditional pressing, due to the increase in width, controllable medium-high and relatively uniform linear pressure is required, the linear pressure is high, and the linear pressure is adjustable. The pressure zone of the shoe press is 6 to 10 times that of the traditional pressing zone. Due to this feature, the residence time of the paper sheet in the shoe press is 6 to 10 times that of the traditional pressing zone under the same vehicle speed conditions. At the same time, the linear pressure that can be obtained by the shoe press is much higher than the traditional roller linear pressure. Controllable medium-high boot press is more helpful to improve the dewatering efficiency of the paper sheet by 15-25%, is more conducive to the curing and shaping of the paper web, saves energy in the drying area of ​​the paper web, reduces the steam consumption by 15-25%, increases the vehicle speed, and increases the production capacity. Thereby obtaining good economic benefits, so that the pressing section can obtain better operating performance. At present, it is the most respected high-end equipment for papermaking equipment, and it has a huge market prospect for upgrading.

[0005] However, there are three challenges in the coordinated, precise and adaptive control of the hydraulic system of the high-end components of the shoe press controllable system: 1) the motion mechanism is complex and modeling is difficult; 2) the control objectives are diverse and the constraints are complex; 3) the reliance on traditional PID has low precision, poor adaptability and high energy loss. The above problems will become more difficult when the nip area and width increase. Therefore, in view of the above status quo, it is urgent to develop a construction method for the shoe press control system to overcome the shortcomings in current practical applications. Summary of the invention

[0006] The object of the present invention is to provide a method for constructing a shoe press control system to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for constructing a shoe press control system comprises the following steps:

[0009] (1) An online system identification strategy based on a weakly coupled model structure is used to solve the modeling difficulties caused by the complex nonlinearity, strong coupling, and multi-input and multi-output characteristics of the dynamic behavior of the ultra-wide shoe press system;

[0010] (2) Find solutions for multi-modal conversion, diverse control objectives and complex constraints through a two-layer predictive control structure of centralized optimization and decentralized control;

[0011] (3) Using the zoned controllable mid-height shoe press structure to adjust the errors at different wavelengths, most errors are eliminated;

[0012] In step (1), the online identification strategy of the weakly coupled model structure includes the following steps:

[0013] Step 1, establishing a seed model through the dynamic mechanism of the hydraulic system;

[0014] Step 2: judge the model quality of the uncertain sub-models and eliminate the sub-models with poor quality;

[0015] Step 3: Progressively correct the model accuracy through online parameter identification.

[0016] As a further solution of the present invention: in step 1, by establishing a seed model structure for the mechanism of the process, taking a single nip hydraulic cylinder as an example, its mechanism model can be simply expressed as:

[0017]

[0018] Where: Q L is the load flow; p L is the load pressure; C teis the equivalent leakage coefficient; C k is the additional leakage coefficient; y is the displacement of the hydraulic cylinder; V t is the equivalent total volume; B is the equivalent damping coefficient of the hydraulic cylinder; F 1 is the equivalent external load force;

[0019] Performing LapLace transformation on the above formula and simplifying it yields:

[0020]

[0021] Where: K ce is the total flow pressure coefficient, K ce =K c +C te .

[0022] As a further solution of the present invention: In step 3, taking the input-output model as an example:

[0023]

[0024] Where:

[0025]

[0026] Where ξ(z) is the measurement noise; denote the pure lag time constants of input and disturbance respectively, and is the parameter to be identified.

[0027] As a further solution of the present invention: in step (2), the two-layer structure predictive control structure includes a model, optimization, control and an external target;

[0028] Among them, the model function is to make output predictions based on the identified system model, including both steady-state predictions and dynamic predictions;

[0029] Optimization refers to steady-state target calculation, and its function is to dynamically calculate the optimal operating point and pass it to the control module;

[0030] Control refers to model predictive control, whose main function is to calculate the control quantity of the underlying control to track the set value, and its output is generally PID set value or valve opening, etc.;

[0031] The external target comes from the calculation result of real-time optimization or the experience of the operator. At this time, the purpose of steady-state target calculation is to track the real-time optimization calculation target.

[0032] As a further solution of the present invention: if no external target is given, the steady-state target calculation will optimize the economic performance indicators within its own constraint space according to the optimized input parameters.

[0033] As a further solution of the present invention: in step (2), the centralized optimization refers to centralized optimization of the optimal working point of the system based on the global process model at the steady-state target calculation layer, and then passing the optimization result as a set value to the dynamic control layer;

[0034] The decentralized control means that at the dynamic control layer, each shoe press pressure zone is divided into several subsystems, and each subsystem is equipped with a predictive controller based on its own sub-process model, and the feedforward controller between the subsystems is used to compensate for disturbances to achieve accurate tracking of the centralized optimization set value.

[0035] As a further solution of the present invention: in step (3), the partitioned controllable middle and high shoe press structure is as follows:

[0036] The two ends of the zoned controllable medium-height shoe press are supported by rolling bearings. A plurality of piston holes are distributed on the generatrix of the mandrel between the two bearings facing the direction of the pressing zone, and a static pressure support head is arranged in the piston hole.

[0037] The hydraulic system is divided into several groups to supply oil to the piston hole. The high-pressure oil entering the piston hole presses the hydrostatic support head to the inner surface of the roller shell, and enters the top of the hydrostatic head through the throttle hole, forming a hydrostatic support oil film between the top and the inner surface of the roller shell, which plays a role in supporting, lubricating and cooling, offsetting the deflection deformation of the working roller surface, and adjusting the size and distribution of the linear pressure and spacing between the two rollers.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] It replaces the existing traditional shoe pressure structure and control system, changes the fixed middle and high into zoned controllable middle and high, and upgrades from single-loop PID control to model predictive control MPC.

[0040] The high-tech zone-controlled ultra-wide shoe press is constructed, and the zone can be adjusted arbitrarily, which increases the adjustment range and expands the adjustment function. It can eliminate the differences caused by various factors such as roll surface deformation, manufacturing errors, and wear and tear in the pressing area, creating conditions for computer transverse control, meeting the requirements of longitudinal and transverse quality self-control of the paper web pressing process in papermaking production, and ensuring the efficient and stable operation of wide-width, high-speed, high-quality papermaking equipment;

[0041] The system's model predictive control (MPC) theoretical system and verification system ensure that the pressing process runs smoothly and that key variables are kept within a reasonable range. Specifically, in terms of control quality, the system's dynamic response process is accelerated by 5%, and the control accuracy is improved by 6%; the strength and tightness of the paper web are enhanced, and the broken paper web is reduced by 7%; in terms of economic benefits, the paper web dryness is increased by 15-25%, the steam consumption in the drying section is reduced by about 15-25%, and the comprehensive economic benefits are increased by 8.6%. Develop an MPC ultra-wide shoe press prototype, finalize the product, and put it on the market. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a system structure block diagram of the double-layer structure model predictive control in an embodiment of the present invention.

[0043] Figure 2 Schematic diagram of a double-layer structure predictive control scheme for centralized optimization of decentralized control in an embodiment of the present invention.

[0044] Figure 3 It is a schematic diagram of the system control research and development plan and route in the embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0047] See also Figure 1-Figure 3 , a method for constructing a shoe press control system, the system control structure adopts a hierarchical mode. Among them, the first layer is the basic control level, which uses hardware such as PLC or DCS to realize direct control of the underlying device, generally composed of basic algorithms such as PID, cascade, and feedforward; the second layer is the advanced control level, which requires a dedicated host computer to optimize the optimal set value of the basic control level. This scheme adopts a multivariable double-layer structure predictive control. The third layer is the device optimization level, which uses real-time optimization (RTO) to ensure that the production device operates in the optimal condition and achieves global or local economic optimization.

[0048] The so-called two-layer structure predictive control refers to the two-layer predictive control strategy adopted by this scheme, which includes a steady-state target calculation module (SSTC) and a dynamic optimization module. Among them, the steady-state target calculation module receives the optimal operating condition target from the RTO, updates the steady-state target in real time by solving a linear programming or quadratic programming problem, and transmits it to the dynamic optimization layer in real time to realize the self-optimization of the basic control level set point and control parameters. The main function of the model identification and prediction module is to continuously update the identification model parameters according to the input and output data, so that the system model is as close to the actual process as possible, and predict the future output in the future limited time domain based on the identified model and future input.

[0049] Partitioned controllable mid-height shoe press structure:

[0050] The two ends of the zoned controllable medium-height shoe press are supported by rolling bearings, and many piston holes are distributed on the generatrix of the mandrel between the two bearings facing the direction of the pressing zone. A static pressure support head is arranged in the piston hole. The hydraulic system is divided into several groups to supply oil to the piston hole. The high-pressure oil entering the piston hole presses the static pressure support head to the inner surface of the roller shell, and enters the top of the static pressure head through the throttle hole, forming a static pressure support oil film on the top and the inner surface of the roller shell, which plays the role of support, lubrication and cooling, offsets the deflection deformation of the working roller surface, and adjusts the size and distribution of the linear pressure and spacing between the two rollers.

[0051] Online identification strategy based on weakly coupled model structure:

[0052] By taking advantage of the respective advantages of centralized control and distributed control, an online identification strategy for weakly coupled model structures is proposed. First, a seed model is established through the dynamic mechanism of the hydraulic system; second, the model quality is judged for the uncertain sub-models, and the sub-models with poor quality are eliminated; finally, the model accuracy is progressively corrected through online parameter identification. The specific steps are as follows:

[0053] The first step is to establish a seed model structure by analyzing the mechanism of the process. Taking a single pressure zone hydraulic cylinder as an example, its mechanism model can be simply expressed as:

[0054]

[0055] In the formula: is the load flow; is the load pressure; is the equivalent leakage coefficient; is the additional leakage coefficient; is the hydraulic cylinder displacement; is the equivalent total volume; is the equivalent damping coefficient of the hydraulic cylinder; is the equivalent external load force.

[0056] Performing LapLace transformation on the above formula and simplifying it yields:

[0057]

[0058] Where: K ce is the total flow pressure coefficient, K ce =Kc +C te .

[0059] The second step is to determine the quality of the uncertain sub-model. In the actual shoe press application, the signals collected on site are highly correlated, and the process model identification is difficult. In addition, the hydraulic cylinder sealing, back roller pressure, valve switching and other operations in the pressing process bring large unmeasurable disturbances to the identification. Therefore, the identified model has a large uncertainty, and uncertainty analysis of the uncertain sub-model is required. The hypothetical seed model is shown in the following table.

[0060]

[0061] Among them, √ indicates a sub-model that is certain to exist, X indicates a sub-model that is certain not to exist, and ? indicates an uncertain sub-model. Sub-models MV2-CV5, MV2-CV6, MV3-CV2, and MV3-CV3 are uncertain sub-models. After uncertainty analysis, the model quality of MV2-CV5 and MV3-CV3 is good, and the weak coupling model shown below can be obtained.

[0062]

[0063] The uncertainty of the sub-model is evaluated by the confidence interval of the estimated parameter vector and different confidence levels are set to determine the quality level of the model. Under a certain significance level, different sub-model quality levels are set according to different confidence interval scales, such as excellent, good, and poor. Then it is determined whether the sub-model is part of the multivariate control model.

[0064] The third step is to gradually correct the model accuracy through online parameter identification. Take the input-output model as an example:

[0065]

[0066] In the formula

[0067]

[0068] Where ξ(z) is the measurement noise; represent the pure lag time constants of input and disturbance respectively. and is the parameter to be identified.

[0069] Conventional parameter estimation methods, iteration methods, least squares methods, subspace identification and other methods can be used for parameter identification.

[0070] The model identification in this study is divided into two parts: the initial model identification in the implementation phase of model predictive control and the re-identification of the model during the online operation of the system. The initial model identification adopts a semi-empirical identification method that combines mechanism and experience, and is obtained by simple identification based on the seed model of the shoe press movement mechanism and historical data or pulse test. The re-identification of the model mainly uses the least squares method and other methods based on the seed model and field data to correct the model accuracy until the performance requirements are met.

[0071] Two-layer structure predictive control based on centralized optimization of decentralized control strategy:

[0072] The two-layer predictive control is a new control structure that has recently emerged in the field of model predictive control. It adds a steady-state target calculation layer to the conventional model predictive control, thereby solving the self-optimization problem of the set value design of the multivariable control system. Figure 1 As shown, the "model" function is to predict the output based on the identified system model, including both steady-state prediction and dynamic prediction. "Optimization" refers to steady-state target calculation, whose function is to dynamically calculate the optimal working point and pass it to the "control" module; "control" refers to model predictive control, whose function is mainly to calculate the control quantity of the underlying control to track the set value, and its output is generally PID set value or valve opening, etc. "External target" can come from the calculation result of RTO or the operator's experience. At this time, the purpose of steady-state target calculation is to track the RTO calculation target. If there is no external target given, the steady-state target calculation can be optimized for economic performance indicators within its own constraint space according to the optimized input parameters.

[0073] In view of the multivariable, constrained and strongly coupled characteristics of the ultra-wide shoe press hydraulic control system, this study adopts a double-layer structure predictive control scheme based on centralized optimization and decentralized control strategy, such as Figure 2 As shown. The so-called centralized optimization refers to the centralized optimization of the optimal working point of the system based on the global process model in the steady-state target calculation layer, and then the optimization result is passed to the dynamic control layer as the set value; the so-called decentralized control refers to the dynamic control layer, dividing each pressure zone of the shoe press into several subsystems, and equipping each subsystem with a predictive controller based on its own subprocess model, and using the feedforward controller between the subsystems to compensate for the disturbance, so as to achieve accurate tracking of the centralized optimization set value. This centralized optimization decentralized control strategy considers the soft and hard constraints and economic benefits of the system from a global level, which can ensure the global optimization performance of the shoe press system; in addition, the predictive controllers equipped in each pressure zone subsystem respectively use their own subprocess models and the synergy between subsystems to improve the overall dynamic control performance of the system.

[0074] Centralized optimization strategy:

[0075] The purpose of steady-state target calculation is to perform self-optimization with economy as the goal according to the actual situation of the process itself under the existing configuration mode of the model predictive controller, which can avoid the optimal set point deviation caused by operating variable disturbances such as inappropriate intermittent paper and oil feeding of the shoe press, structural disturbances such as hydraulic cylinder sealing, and unavoidable model errors. This study adopts the steady-state target calculation method of centralized optimization strategy.

[0076] The steady-state target calculation module is connected to the lower-level subsystem control station through a communication system. The steady-state target calculation module integrates the constraints of the global process, solves the optimization problem based on the full process model, calculates the set values ​​of each subsystem of the lower-level control system structure, and transmits them to the prediction controller of each subsystem through the communication cable as their control targets. Compared with the method of solving the optimal set point independently by the subsystem, this centralized optimization steady-state target calculation method based on the full process model can not only ensure the global optimization performance of the shoe press system, but also improve the overall dynamic control performance of the system.

[0077] 1) Problem description of steady-state target calculation

[0078] The operation variables are divided into cost operation variables and minimum movement operation variables. The steady-state target calculation optimization problem can be described as:

[0079]

[0080] For solving the optimization problem of steady-state target calculation, it is first necessary to determine whether the optimization problem under the constraint conditions has a feasible domain. When the feasible domain exists, solve the optimization problem in the feasible domain and calculate the optimal steady-state target value. When the feasible domain is empty, it means that the optimization is not feasible, and it is necessary to relax the constraints and then solve the optimization problem in the feasible domain after the constraints are relaxed to obtain the optimal steady-state target value.

[0081] 2) Feasibility judgment of steady-state target calculation

[0082] The feasibility judgment of steady-state target calculation is the premise to ensure that the optimization problem has a solution. The feasibility judgment of the optimization problem is converted into a constraint target adjustment problem as follows:

[0083]

[0084] When J=0, it means that the constraint set is non-empty and the optimization problem is feasible; when J≠0, it means that the constraint set is empty, the optimization problem is not feasible, and the constraints of the controlled variables need to be adjusted.

[0085] 3) Solving optimization problems based on the priority of operating variables

[0086] When the feasible region does not exist, that is, the optimization problem is not feasible, the constraints on the controlled variables need to be relaxed. The constraint target adjustment problem is transformed into a standard linear programming problem. The corresponding slack variable solutions are calculated in order of the priority of the controlled variables, and then the slack variable solutions of the controlled variable constraints are obtained. Then the optimization problem is solved.

[0087] When the feasible region exists, the optimization problem is feasible. Let the slack variable be 0, and the following optimization problem is obtained:

[0088]

[0089] First, determine the cost priority of all cost operation variables, and together with the minimum mobile variable priority, form the operation variable priority; second, sort the operation variable priorities in ascending order within the feasible domain. Calculate the corresponding operation variable steady-state target increments in the order of operation variable priority until the lowest priority operation variable steady-state target is obtained; finally, the steady-state target increments of each priority constitute the optimal operation variable steady-state target increment to obtain the optimal steady-state target.

[0090] Decentralized control strategy:

[0091] In the dynamic control layer, each shoe press area is divided into several subsystems, and each subsystem is equipped with a model prediction controller based on the model of each subsystem. The feedforward controller between the subsystems is used to compensate for the disturbance and achieve accurate tracking of the centralized optimization setting value. The dynamic control layer is composed of multiple subcontrollers connected in series. Each controller tracks the optimal set point transmitted by the steady-state target calculation module and controls the subsystem corresponding to its own controller. At the same time, the subcontrollers exchange information through communication equipment, and the coupling effect between the subsystems can be compensated by adding a feedforward control device.

[0092] 1) Decentralized dynamic control algorithm

[0093] The decentralized control consists of two parts: a feedforward controller and a sub-controller, which are described as follows.

[0094] Feedforward controller: The shoe press system pressing area is divided into several subsystems, and each subsystem is equipped with an independent subcontroller (model predictive controller). The subsystem exchanges information with the adjacent subsystem through communication equipment. There is usually a certain coupling relationship between adjacent subsystems. For example, the system output of a subsystem will affect the subsystem, that is, the system input of a subsystem can be regarded as a disturbance of the subsystem. Therefore, the feedforward model can be used to design a feedforward control device to compensate for the measurable disturbance of the associated subsystem.

[0095] Assume that the response matrix of the control input of the subsystem to the subsystem disturbance is identified as, when it is zero, it means that there is no correlation between the two, and when it is not zero, the subsystem is affected by the subsystem. Therefore, the prediction model of the subsystem can be described as:

[0096]

[0097] in, S is the output value of subsystem i at the next P moments and the initial value of the current moment. sub,i is the step response matrix of the input to the output of subsystem i.

[0098] The feedforward control device utilizes the delayed characteristics of the manipulated variable disturbances between coupled subsystems to compensate for the coupling effects between the subsystems and improve the robustness of the system.

[0099] Sub-controller: The sub-controller adopts the conventional predictive control algorithm to achieve accurate tracking of the optimal set point sent by the steady-state target calculation module.

[0100] Assuming that subsystem i has s∈p control outputs and h∈m control inputs, the subcontroller solves the following optimization problem:

[0101]

[0102] Among them, R i =block-diag{R i,1 …,R i,h}, Q i =block-diag{Q i,1 ,…,Q i,s} is the control input weight matrix and the error weight matrix.

[0103] Substituting the subsystem output prediction model (5) into the performance index function (6) can calculate the optimal control increment of the subsystem:

[0104]

[0105] The immediate control increment can be given by:

[0106]

[0107] in,

[0108]

[0109] At the next sampling moment, the sub-controller detects the actual output value of the subsystem and corrects the model-based prediction based on the error vector formed by subtracting the predicted output value from the actual output value.

[0110] 2) Specific implementation steps of the two-layer structure predictive control based on centralized optimization of decentralized control strategy

[0111] Step 1: Initialization.

[0112] After the two-layer structure predictive control is started, the initialization operation must be performed first. The control input and controlled output at the current moment are read from the distributed control system (DCS) through communication technologies such as OPC, and the values ​​of the control input and controlled output are saved in the memory as steady-state initial values. Then, various parameters of the two-layer structure predictive control are given, including the cost factor of the control input variable, the constraints of the input and output variables, etc.

[0113] Step 2 periodically performs steady-state target calculation and dynamic optimization functions.

[0114] 1) Obtain the real-time values ​​of control input and controlled output through OPC communication;

[0115] 2) The steady-state target calculation module uses the process output signal to calculate the prediction error, combines the steady-state values ​​of the input and output variables stored in the memory to calculate the steady-state target, uses the optimization control result at the current moment to update the steady-state values ​​of the input and output in the memory, and transmits the steady-state value of the control input and the output steady-state value of the open-loop stable process as the set point to the dynamic optimization module;

[0116] 3) The dynamic optimization module makes dynamic predictions based on the instantaneous values ​​of the control input and controlled output, performs control calculations, and transmits the control input results to the basic control layer DCS for execution. Generally, the two-layer structure predictive control should be started when the system is stable.

[0117] It should be noted that in the present invention, it should be understood that although this specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that those skilled in the art can understand.

Claims

1. A method for constructing a shoe press control system, characterized in that: The following steps are involved: (1) An online system identification strategy based on a weakly coupled model structure is used to solve the modeling difficulties caused by the complex nonlinearity, strong coupling, and multi-input and multi-output characteristics of the dynamic behavior of the ultra-wide shoe press system; The online identification strategy of the weak coupling model structure includes the following steps: Step 1, establishing a seed model through the dynamic mechanism of the hydraulic system; Step 2: judge the model quality of the uncertain sub-models and eliminate the sub-models with poor quality; Step 3, progressively correcting the model accuracy through online parameter identification; (2) Through the two-layer predictive control structure of centralized optimization and decentralized control, we can find solutions for multi-modal conversion, diverse control objectives, and complex constraints; (3) The errors of different wavelengths are adjusted by using a zoned controllable mid-height shoe-type pressing structure, thereby eliminating most of the errors.

2. The method for constructing a shoe press control system according to claim 1, characterized in that: In step 3, the input-output model is adopted: Where: Where ξ(z) is the measurement noise; denote the pure lag time constants of input and disturbance respectively, and are the parameters to be identified; the parameter estimation method, iteration method, least square method and subspace identification method are used to identify the parameters to be identified of the system.

3. The method for constructing a shoe press control system according to claim 1, characterized in that: In step (2), the two-layer predictive control structure includes modeling, optimizing, and controlling external targets; Among them, the model function is to make output predictions based on the identified system model, including both steady-state predictions and dynamic predictions; Optimization refers to steady-state target calculation, and its function is to dynamically calculate the optimal operating point and pass it to the control module; Control refers to model predictive control, whose main function is to calculate the control quantity of the underlying control to track the set value, and its output is the PID set value or valve opening; The external target comes from the calculation result of real-time optimization or the experience of the operator. At this time, the purpose of steady-state target calculation is to track the real-time optimization calculation target.

4. The method for constructing a shoe press control system according to claim 3, characterized in that: If no external target is given, the steady-state target calculation will optimize the economic performance indicators within its own constraint space based on the optimization input parameters.

5. The method for constructing a shoe press control system according to claim 1, characterized in that: In step (2), the centralized optimization refers to the centralized optimization of the optimal working point of the system based on the global process model at the steady-state target calculation layer, and then the optimization result is passed to the dynamic control layer as a set value; The decentralized control means that at the dynamic control layer, each shoe press pressure zone is divided into several subsystems, and each subsystem is equipped with a predictive controller based on its own sub-process model, and the feedforward controller between the subsystems is used to compensate for disturbances to achieve accurate tracking of the centralized optimization set value.

6. The method for constructing a shoe press control system according to claim 1, characterized in that: In step (3), the structure of the zoned controllable middle and high shoe press is as follows: The two ends of the zoned controllable medium-height shoe press are supported by rolling bearings. A plurality of piston holes are distributed on the generatrix of the mandrel between the two bearings facing the direction of the pressing zone, and a static pressure support head is arranged in the piston hole. The hydraulic system is divided into several groups to supply oil to the piston hole. The high-pressure oil entering the piston hole presses the hydrostatic support head to the inner surface of the roller shell, and enters the top of the hydrostatic head through the throttle hole, forming a hydrostatic support oil film between the top and the inner surface of the roller shell, which plays a role in supporting, lubricating and cooling, offsetting the deflection deformation of the working roller surface, and adjusting the size and distribution of the linear pressure and spacing between the two rollers.

Images