Graphical system modeling method for nonlinear process mechanism of process industry

Abstract

The invention discloses a graphical system modeling method oriented to a nonlinear process mechanism of a process industry. The nonlinear process mechanism graphical system modeling method comprises an expansion mechanism element registration step method based on unit ontology constraint, a complex mechanism high-fidelity local agent description method based on third-order spline interpolation, and a system differential equation automatic complementation and positive qualitative inspection method based on Newman conditions. A complex mechanism high-fidelity local agent model is established through cubic spline fitting and solvability test, and on the basis of expanding mechanism element graphical representation and based on a bonding graph modeling method, a flow industry nonlinear process mechanism-oriented solvable graphical model can be constructed. According to the nonlinear process mechanism graphical system modeling method, the relation between a model mechanism and modeling personnel is separated to the maximum extent, the requirement of the modeling personnel for understanding a fine mechanism is lowered, and the modeling efficiency and the model precision are greatly improved.

Description

Technical Field

[0001] The present invention relates to a process industry process modeling method, in particular to a graphical system modeling method for multi-field nonlinear mechanism processes in the process industry. Background Art

[0002] As an important part of industrial control technology, industrial process modeling plays an important role in state monitoring, performance evaluation, process optimization, troubleshooting and decision-making. However, since complex industrial processes usually involve multiple disciplines such as chemistry, physics, biology, and multiple fields such as control, electronics, mechanics, power, and thermal power, and exhibit nonlinear, time-varying, and uncertain characteristics, modeling usually adopts a parallel development model with typical distribution, parallelism, collaboration, and integration characteristics. At present, with the continuous enhancement of modeling and simulation capabilities in a single field, the key issues restricting complex industrial process modeling are more focused on the integration and decoupling between multi-domain models. Graphical modeling, as an abstract modeling method that uses graphical modules to represent complex systems, can well express the connection relationship between systems in different fields or at different stages. Graphical modeling methods, represented by bond graph modeling, have been widely used in industrial process modeling.

[0003] However, facing the complex process mechanism of multi-domain coupling in the process industry, the limited basic elements of the traditional bond graph cannot cover the deep causal relationship of the real multi-parameter nonlinear coupling, which brings many difficulties to the solvability of the model. The introduction of massive parameters and equations has greatly improved the accuracy of the model, but it also brings high time and computational costs. Each sub-process is modeled based on its own physical constraints, and it is difficult to ensure that the system model is solvable when integrated. This directly restricts the application of the bond graph modeling method in the modeling of multi-domain complex industrial processes. At present, the model solution problem usually achieves consistency with the expression of the basic elements of the traditional bond graph through linear fitting methods such as least squares regression to reduce the computational cost, but it largely sacrifices the accuracy of the model. At the same time, although the data-driven local proxy method can effectively realize the refined expression of the mechanism, due to the internal high-order and segmented mechanism characteristics, it cannot be directly converted into the basic bond graph element in definition, which makes this type of method unable to play its advantages in the traditional bond graph modeling process. Therefore, when modeling complex industrial processes. The agent model is introduced into the bond graph modeling method. Through the graphical representation of the extended mechanism elements based on the unit body constraints, the local agent model is established using cubic spline interpolation, and some high-order complex processes are equivalently modeled to solve the model reduction and solution. A solvability detection mechanism is established in the local agent model to ensure the solvability of the model. The graphical modeling method based on the local agent model for complex industrial processes has greatly reduced the computational cost of the model and ensured the optimal solution of the model. It not only expands the compatibility, reusability and reconfiguration capabilities of the model, but also separates the coupling relationship between the model mechanism and modeling. Through graphical interactive modeling, the entry threshold and model mechanism requirements of modelers are reduced, and the efficiency and accuracy of modeling are improved. Summary of the invention

[0004] In order to solve the problem of optimal solution of models in multi-domain modeling in complex industrial process modeling, the present invention provides a graphical modeling method based on local proxy model. This method can greatly reduce the model order of high-order nonlinear models and the problem of solving the model during integration. The local proxy model accurately represents the complex physical constraints inside the component to achieve the purpose of order reduction and solution. At the same time, the local proxy model maximizes the connection between the separation model mechanism and the modeler, reducing the modeler's requirements for understanding the fine mechanism, and greatly improving the efficiency of modeling and the accuracy of the model.

[0005] The graphical system modeling method of the nonlinear process mechanism is modeled according to the V-shaped concept, which decomposes and concretizes the complex industrial process model from top to bottom, forms a rough model at the top level and a fine model of each component, and then integrates it from bottom to top to form a fine model at the system level. By continuously refining the graphical model, the performance requirements of functional modules, functional subsystems, and functional components are continuously concretized; by continuously combining the graphical model, the design analysis results of each modeler in the assembly process are continuously transmitted and integrated, thereby realizing the integration and performance optimization of models at different structural levels in the assembly-subassembly-final assembly process, and achieving efficient and accurate modeling.

[0006] The characteristics of the present invention are:

[0007] (1) Through the abstract definition of each functional unit in the energy transfer process, the ontological constraints and graphical design specifications of the functional elements in the user-defined construction process are constructed, which effectively separates the functional business of mechanism description personnel and system modeling personnel in the redevelopment process, and lays a convenient architectural foundation for the embedding of local agent models.

[0008] (2) The high-fidelity local proxy function of complex mechanisms is realized through cubic spline interpolation, which effectively reduces the difficulty of mechanism description based on experimental data for professionals, thereby improving the accuracy of the mechanism model in describing the mechanism of real industrial systems at the component level.

[0009] (3) An automatic completion mechanism for implicit differential relations based on Newman conditions was constructed. Further, based on this mechanism, a positive definiteness test mechanism for complex nonlinear differential equations was established to ensure that the global solution uniqueness of the system model is maintained during the user's customized embedding of local mechanisms. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 Steps for graphical system modeling of nonlinear process mechanisms

[0011] Figure 2 Graphical representation of registration information for extended mechanism elements

[0012] Figure 3 High-fidelity local proxy characterization steps for complex mechanisms

[0013] Figure 4 Automatic completion and positive definiteness test steps for system differential equations DETAILED DESCRIPTION

[0014] The graphical system modeling method for nonlinear process mechanisms in process industries provided by the present invention is described in detail below with reference to the accompanying drawings.

[0015] The present invention provides a graphical modeling method for process industry technology. Figure 1 As shown, the graphical system modeling method for nonlinear process mechanisms in the process industry first starts from the upper-level functional architecture and gradually differentiates the lower-level functional elements, completes the decomposition design of the complex industrial process and the concreteness of the functional element model, and then completes the registration of extended mechanism elements based on unit body constraints. For nonlinear complex mechanism elements, a high-fidelity local proxy model with cubic spline fitting is used for equivalence. Finally, from the bottom to the top, the functional components of the lower layer are integrated into the complex industrial process of the upper layer. Through the positive definiteness test of the system differential equation and the automatic parameter completion mechanism, the model is ensured to be solvable, and finally a graphical model of the process industry technology is established. It includes three core steps: registration of extended mechanism elements based on unit body constraints, high-fidelity local proxy characterization of complex mechanisms based on third-order spline interpolation, and automatic completion and positive definiteness test of system differential equations based on Newman conditions:

[0016] 1. Extended mechanism component registration based on unit ontology constraints

[0017] The extended mechanism element registration step is based on the ontological constraint characteristics of each mechanism unit in the energy transfer process, and defines the extended mechanism element registration information as four standardized contents: element name, element legend, element interface anchor point, and element interface description, such as Figure 2As shown. The specific unit body constraint information is bound to the type name of the extended mechanism element. The information binding standardizes the basic element interface standard by adjusting the data format according to the two generalized variables of potential and flow required for the energy transfer mechanism characterization process. The specific format is {name: model name [string type], id: component instance identifier [string type], legend: image path [string type], input interface: [{name: input name [string type], order: input sequence number [unsigned integer type], value: immediate value [floating point type]}, ...], state variable: [{name: parameter name [string type], key: parameter identifier used inside the function [integer type], value: immediate value [floating point type / local proxy model data], isFunc: whether to define the proxy model [Boolean type]}, ...], output interface: [{name: output name [string type], order: output sequence number [unsigned integer type], value: immediate value [floating point type]}, ...]}. The component name is the identifier of the extended mechanism component in the model, and is usually named according to the component function; the component legend is the graphical representation of the component in the model, and try to choose a graphic that can represent the graphical structure or function; the component interface anchor is the place where energy is transferred between components, and the representation is to be marked, and the component interface description is used to ensure that the modeler is clear about the interface function; the state variable of the component defines the performance characteristic parameters of the component instance. If isFunc is set to false, value is filled in with the corresponding numerical value (for example: the resistance value of a general resistor is the parameter isFunc, and value is the resistance value); otherwise, if isFunc is set to true, value is filled in with the supporting data of the local proxy model, in the format of {IN_1: the value selected by input 1 [floating point list], IN_2: the value selected by input 2 [floating point list], …, IN_K: the value selected by input K [floating point list], value: …[K-dimensional list under multiple input combinations]}, so as to support the model to obtain the grid data required by the local proxy during online simulation with the minimum memory space occupied. The registration of extended mechanism components based on unit ontology constraints not only greatly facilitates the reusability and extensibility of the model, but also better unifies the compatibility of the mechanism model and the local proxy model interface.

[0018] 2. High-fidelity local proxy characterization of complex mechanisms based on third-order spline interpolation

[0019] The complex mechanism high-fidelity local proxy characterization adopts the third-order spline fitting method, which is mainly responsible for the complex nonlinear mechanism within the functional unit, completing the data-based mechanism characterization, and embedding it into the component mechanism description through the state variable interface registered by the extended mechanism element, realizing the mixed mechanism characterization mechanism of data and formula. The multivariate cubic spline model is mainly aimed at complex high-order nonlinear functional elements with unclear mechanisms. The specific construction process is as follows Figure 3 Steps to proceed:

[0020] ① Construct a local proxy model of cubic spline fitting. Cubic spline interpolation fitting is achieved by using the cubic function a i x 3 +b i x 2 +c i x+d i To describe the nonlinear and complex mechanism of functional elements.

[0021] ② The input and output data of the components that are to be used in the local proxy model are normalized. The data of the input interface of the graphical component of the local proxy model is m yuan. According to the registration specification of the extended mechanism component, the format of the input interface is consistent with the value format defined by the local proxy model in the state parameter, that is: {IN_1: the value selected by input 1 [floating point list], IN_2: the value selected by input 2 [floating point list], ..., IN_K: the value selected by input K [floating point list], value: ... [K-dimensional list under multiple input combinations]}. The normalization processing step is to adaptively select the corresponding multivariate spline interpolation method by counting the number of key names starting with "IN", and select the key name starting with "IN" and sort them by the serial number starting from the third digit of the key name, so as to calibrate the multivariate input combination corresponding to the gridded data in the value.

[0022] ③ Use the previous t frames of data for fitting. The cubic spline fitting function is shown in the following formula:

[0023] S i (x i )=a i (xx i ) 3 +b i (xx i ) 2 +c i (xx i )+d i ,i=1,2,…,t

[0024] Where: a i , b i , c i , d iRespectively represent the cubic term coefficient, quadratic term coefficient, linear term coefficient and constant term of the ith segment of cubic spline. i , S i (x i ) represent the input data and output data of the i-th frame respectively.

[0025] Using t-frame data to establish the cubic spline function and its first-order derivative, second-order derivative equations and boundary conditions, no less than 4 (t-1) equations can be established. The specific formula is as follows

[0026] Function value: s i (x i )=y i ,i=1,2,...,t-1;s t-1 (x t )=y t ;s i-1 (x i )=y i ,i=2,...,t;s1(x1)=y1

[0027] First derivative: s′ i (x i ) = s′ i-1 (x i ),i=2,3,...,t-1

[0028] Second-order derivative: s″ i (x i )=s″ i-1 (x i ),i=2,3,...,t-1

[0029] Boundary conditions: s′1(x1),s″1(x1),s″′1(x1),s′ t-1 (x t ),s″ t-1 (x t ),s″′ t-1 (x t )

[0030] According to the multivariate linear equation formed above, the coefficients of each order of the cubic spline fitting function can be solved. The solvability of the equation is judged by step 3, and the differential relationship is completed when it is unsolvable, and the cubic spline local proxy model with 4 (t-1) parameters is obtained by solving.

[0031] ④ After the local proxy model is modeled, it is further registered as an extended mechanism element through the registration specification, and the state variables of the extended mechanism element are used as state variables in the form of functions to participate in the mechanism characterization process in the form of formulas, thereby realizing the mixed modeling function of complex process based on data and formulas. For example: the resistance of the thermistor is related to the square integral of the current, so the thermistor can use this mechanism to construct a local proxy model of the square integral of the current and the resistance, and register it as the state variable R, thereby defining the thermistor mechanism as:

[0032]

[0033] Where: e and f represent voltage (potential) and current (flow) respectively; q f2 represents the square integral of the current; R represents the local proxy model of the square integral of the current and the resistance.

[0034] 3. Automatic completion and positive definiteness test of system differential equations based on Newman conditions

[0035] The automatic completion and positive definiteness test steps of the system differential equations are based on the two generalized variables of potential and flow that the energy propagation process depends on. By mapping the relationship between the interfaces of the extended mechanism elements and the dynamic binding of the internal physical constraints, all local dynamic equilibrium relationships are integrated to construct the constraint equations of the system. The solvability of the extended mechanism elements determines the solvability of the model. Figure 4 As shown in Figure 2, the steps to test the solvability of the local proxy model are:

[0036] ① Calculate the number of cubic spline fitting parameters of the local proxy model. The number of input data in step 2 is t, so the number of fitting parameters is 4(t-1);

[0037] ② Calculate the number of cubic spline fitting equations of the local proxy model, and use the equal relationship between the function values ​​on both sides at the endpoints of the internal interval to establish 2(t-2) equations. The first-order derivatives of the function values ​​on both sides at the endpoints of the internal interval are equal to establish t-2 equations. The second-order derivatives of the function values ​​on both sides at the endpoints of the internal interval are equal to establish t-2 equations. The two endpoints of the sample must pass through the spline function to establish 2 equations. In the absence of initial conditions, the number of equations is less than the number of parameters, and the model is unsolvable. Since there is a Newman condition in cubic spline fitting, it is necessary to complete the differential terms through initial conditions or boundary conditions.

[0038] ③ Completion of differential relations. Construct Newman boundary conditions, establish the first-order derivative, second-order derivative, and third-order derivative equations of the endpoints. If the system is given an initial condition, the differential term completes the first-order derivative; if the system is not given an initial condition, the second-order derivative of the endpoint can be specified to be equal to 0, and the second-order derivative is used to complete the differential term; when there is not enough information to determine the first- and second-order derivative values ​​of the endpoint, set the third-order derivative to 0, and use the third-order derivative to complete the differential term. When high-order derivatives are used to complete differential terms, the equations constructed by low-order derivatives are not involved in the calculation. At this time, the high-order derivative of the state at the previous moment is taken to complete the value of the low-order derivative at the current moment, ensuring the state uniqueness requirements of the complex process model for simulation.

[0039] ④ Test for the positive definiteness of the system equations. After completing the differential relations in ③, the number of equations obtained is greater than or equal to the number of parameters, the system is positive definite, the model is solvable, and the cubic spline fitting parameters of the local surrogate model are obtained by solving the multivariate linear equations to obtain the local surrogate model.

Claims

1. A graphical system modeling method for nonlinear process mechanisms in process industries, characterized by: The graphical system modeling method of the nonlinear process mechanism is unitized according to the energy transfer relationship, and supports the hybrid modeling requirements of local data and formula mechanisms. It mainly includes three functions: an extended mechanism element registration step based on unit body constraints, a high-fidelity local proxy characterization step of complex mechanisms based on third-order spline interpolation, and automatic completion and positive definiteness test of system differential equations based on Newman conditions.

2. The graphical system modeling method for nonlinear process mechanisms in process industries according to claim 1 is characterized by: The extended mechanism element registration step of the graphical system modeling method is based on the ontological constraint characteristics of each mechanism unit in the energy transfer process, and defines the extended element registration information as four standardized contents: component name, component legend, component interface anchor point, and component interface description. Through automatic parsing of the standardized data format, the extended element is embedded in the energy transfer framework where the basic element is located, and the consistency of the form is achieved, thereby supporting the expansion needs of personalized mechanism units.

3. The graphical system modeling method for nonlinear process mechanisms in process industries according to claim 1 is characterized by: The complex mechanism high-fidelity local proxy characterization step of the graphical system modeling method adopts the cubic spline method, adaptively constructs a multivariate cubic spline model of the local complex mechanism based on the normalized mechanism data access format, and further embeds it into the energy transfer process model based on the parameter interface of the component body, thereby realizing the hybrid modeling function of complex process based on data and formulas.

4. The graphical system modeling method for nonlinear process mechanisms in process industries according to claim 1 is characterized by: The system differential equation automatic completion and positive definiteness verification step of the graphical system modeling method is responsible for calculating the relationship between the number of parameters and the number of equations. When the number of equations is less than the number of parameters, it constructs Newman boundary conditions, establishes endpoint first-order derivative, second-order derivative, and third-order derivative equations, completes the system differential relationship, and verifies the positive definiteness of the overall system equation, thereby ensuring the state uniqueness requirements of the complex process model for simulation.

Images