CN-121998275-A - Method and system for uniformly planning and allocating distributed resources of combustion chamber of solid engine

CN121998275ACN 121998275 ACN121998275 ACN 121998275ACN-121998275-A

Abstract

The invention discloses a method and a system for uniformly planning and allocating distributed resources of a combustion chamber of a solid engine, wherein a complex space-time coupling relation among multiple procedures and multiple resources is uniformly modeled through a time-varying hypergraph and a time expansion network; further constructing a four-target optimization model comprising finishing time, energy consumption, quality loss and distribution robustness CVaR risks, adopting Wasserstein distribution robustness opportunity constraint to process random disturbance, guaranteeing safety and quality probability threshold under uncertain distribution, solving by utilizing a co-evolution distribution robustness multi-target optimization algorithm, combining digital twin on-line calibration disturbance distribution, guaranteeing engineering feasibility of the solution through two-stage feasible domain projection, and finally outputting a robust pareto solution set and an interpretable rescheduling label. The method effectively solves the problems of low feasibility rate and weak risk control of the traditional method under the condition of distributed drift, and achieves multi-objective unified optimization of safety, beat, energy consumption and quality.

Inventors

MA CHI
LI MINGMING
HUANG WEIJIAN
CHEN ZHIXUAN
PU QIAN
MA GUOXUAN
LIU XIAOYU
Mou Borui
CAI WEIJIE

Assignees

重庆大学
西安航天化学动力有限公司

Dates

Publication Date: 20260508
Application Date: 20251104

Claims (10)

1. A method for uniformly planning and allocating distributed resources of a combustion chamber of a solid engine is characterized by comprising the following steps: step one, hypergraph modeling and time expansion network construction: mapping multi-source heterogeneous resources and tasks involved in a production line into a time-varying hypergraph Wherein the edge is exceeded Constructing a time expansion network based on the hypergraph, and dispersing the time dimension into a time slot set For each resource Construction node And pass through an arc Representing the time between adjacent time slots Forming a joint constraint characterization of the realization task in a time domain and a resource domain; step two, multi-objective optimization modeling: An optimization problem is constructed that includes four objectives: Wherein: Representing tasks Is a time of completion of (a); Representing the total energy consumption of the device; representing a mass loss function; Representing CVaR risk expectations under the wasperstein distribution uncertainty set; Representing a task; representing a control variable; Representing a device state variable; Representing a negative safety margin value; Representing the disturbance; Expressed at a confidence level Lower based on distribution families Conditional risk expectation operators of (2); Is a fuzzy distribution family; And step three, opportunity constraint and Wasserstein distribution robust processing: For production disturbance variable The Wasserstein spheres were constructed: Wherein: representing a family of fuzzy distributions; And Respectively representing a true distribution and an empirical distribution; Creating strong robust opportunity constraints: , Wherein: representing the quality; representing a quality threshold; representing a safety threshold; Representing a quality constraint weight coefficient; representing a security constraint weight coefficient; Step four, solving a co-evolution distribution robust multi-objective optimization algorithm: the CDRD-MO algorithm is adopted for solving, and the method comprises the following core steps: Decomposition strategy using reference vectors Decomposing a target space to construct a plurality of target sub-problems: Take over the preferred files, adopt The dominant external archive maintains non-dominant solutions and adjusts adaptively based on archive density A value; the risk perception speed is updated, namely an importance sampling is used for estimating a risk gradient, and the particle speed or the individual variation direction is updated; The feasible region network flow projection, namely restoring the solution to a feasible region through a two-stage feasible region projection operator; Step five, digital twin on-line calibration: The digital twin system is combined to collect the production line data in real time, disturbance distribution parameters are dynamically updated through Kalman filtering and multi-fidelity correction, and the Wasserstein sphere radius is adjusted accordingly And risk parameters ; Step six, outputting an interpretable scheduling scheme: outputting a robust pareto solution set and an interpretable rescheduling label vector Bottleneck resources, critical tasks and risk triggers are identified through sparse optimization.
2. The method for uniformly planning and allocating distributed resources of combustion chamber of solid engine as set forth in claim 1, wherein in said step one, time-varying hypergraph is obtained The resource synchronization occupation constraint is expressed as: Wherein: for the task At the moment of time Decision variables of whether to start execution; for the task At the moment of time Decision variables of whether to start execution; for the task At the moment of time Whether or not to be allocated to a resource Is a binary variable of (2); For the task Is a start time variable of (1); Is a discrete time index variable; is a continuous time index variable; for the task Duration of (2); For the task End time variable of (2); The capacity/mutual exclusion constraint is expressed as: Wherein: Is a resource An upper capacity or capacity limit of (2); The process context constraint is expressed as: Wherein: For the task Is a preamble task set; for the preamble task End time variable of (2); The AGV space-time flow conservation constraint is expressed as: Wherein: To be at the moment From station Direction station Binary decision variables for material or task transfer; For the last moment Slave node Directional node A material transfer state variable of (2); Is a starting point node; Is a target node; is the starting node for the previous transfer.
3. The method for uniformly planning and allocating distributed resources of a combustion chamber of a solid engine according to claim 1, wherein in the second step, CVaR risk items are constructed in a dual equivalent form: Wherein: lipschitz continuity constraints representing the loss function and obtained by jacobian spectral radius approximation; representing a smoothed excess loss function; Representing a sample Lower task At the disturbance of A loss function under conditions; representing an overall loss function; representing a risk cutoff threshold; Represent the first A plurality of disturbance samples; Representing risk parameters; Represents the Wasserstein sphere radius; Representing the number of samples.
4. The method for uniformly planning and allocating distributed resources of a combustion chamber of a solid engine according to claim 3, wherein in the fourth step, the risk gradient is estimated by importance sampling: Wherein: representing a fourth objective function modified by Wasserstein-CVaR; Representing a decision variable Gradient operators of (a); to smooth the step function; Representing an approximate update map based on gradient propagation.
5. The method for uniformly planning and allocating distributed resources of a combustion chamber of a solid engine according to claim 1, wherein in the fourth step, the solution is restored to the feasible region by a two-stage feasible region projection operator, comprising: First stage network flow projection, namely repairing discrete variables violating resource capacity or time sequence constraint by solving a minimum cost flow problem on a time expansion network: Wherein: Representing slave nodes Directional node Resource or task traffic of (1); Representing nodes And node Unit transmission cost or energy consumption coefficient; Representing nodes Net supply and demand of (3); Representing a total energy consumption or energy objective function of the system; and a second stage SOCP of mechanism repair, namely correcting continuous variables still violating temperature rise, energy consumption or fatigue mechanism constraint by solving a second-order cone programming problem: Wherein: representing a linear constraint coefficient matrix; representing a linear constraint constant vector; Represent the first Coefficient matrixes constrained by the second order cones; a translation vector representing a second order cone constraint; a linear direction vector representing cone constraints; Representing the offset scalar of the second order cone constraint.
6. The method for uniformly planning and allocating distributed resources of a combustion chamber of a solid engine according to claim 5, wherein the mechanism coupling constraint comprises an exothermic kinetic equation and an approximate equation of temperature rise of single casting/curing: Wherein: Represents the reaction activation energy; representing a reaction progress factor; Represents an Arrheni Wu Siyu index factor; representing the rate of heat release per unit volume; a product term representing a gas constant and a temperature; Represents equivalent specific heat capacity; representing the system temperature; indicating the rate of temperature change; Representing a system power input term; Representing a heat exchange function; Representing ambient temperature; A formula for defining an exothermic rate; Indicating the change in reaction enthalpy.
7. The method for uniformly planning and allocating distributed resources of a combustion chamber of a solid engine according to claim 1, wherein in the fourth step, a population update in a co-evolution distributed robust multi-objective optimization algorithm adopts a speed update strategy based on a reference vector: Wherein: updated particle velocity vectors are updated; a particle velocity vector before update that is before update; Is an inertial weight coefficient; And Individual learning factors; And Two random numbers uniformly distributed in the intervals of [0,1] respectively; Historical optimal positions for individual particles; the particle position vector is the particle position vector at the current iteration moment; Is an adaptive learning rate or step size coefficient; to decide variable Gradient operators of (a); for smoothing the risk objective function; the dominant archive controls the diversity of the solution sets, Self-annealing of values according to file density: Wherein: Is the attenuation coefficient.
8. The method for uniformly planning and allocating distributed resources of combustion chamber of solid engine according to claim 1, wherein in the fifth step, wasserstein sphere radius is According to the quantile dynamic update of the historical disturbance residual, the self-adaptive update law is satisfied: Wherein: is the updated Wasserstein sphere radius; to be the Wasserstein sphere radius before updating; the coefficients are adaptively adjusted; 95% quantile; Representation of Disturbance of time; representing a disturbance estimate or a smooth disturbance state.
9. The method for uniformly planning and allocating distributed resources of combustion chamber of solid engine according to claim 1, wherein in said step six, the interpretable rescheduling label vector Expressed as: Wherein: Representing a feature mapping function; Represent the first Decision variable vectors of the secondary iteration; representing a sparse weight matrix or a projection weight; representing sparse regularization coefficients; Representing vectors Is a normal number of L 1 .
10. A system for performing the solid engine combustor distributed resource unified planning and deployment method of any one of claims 1-9, comprising: the data acquisition module is used for acquiring the temperature, strain, power and AGV position data of the production line in real time through an OPC-UA or Modbus industrial protocol; The hypergraph modeling and time expansion network module is used for constructing and maintaining a time-varying hypergraph model and a time expansion network; a distributed robust optimization module configured with a reference vector generation unit, The archive maintenance unit and the risk gradient calculation unit are used for executing CDRD-MO algorithm; The feasible region restoration module is used for integrating a network flow solver and a second-order cone planning solver and executing two-stage feasible region projection; The digital twin module comprises a physical layer model, a data layer model and a fusion calibration unit and is used for correcting disturbance distribution in real time; an interpretable rescheduling module, which is used for outputting a bottleneck factor set of task-equipment-time dimension by adopting a sparse constraint optimization method; and the visualization module is used for graphically displaying the pareto front edge, the risk probability curve and the resource utilization rate.

Description

Method and system for uniformly planning and allocating distributed resources of combustion chamber of solid engine Technical Field The invention belongs to the technical field of production resource planning and scheduling, and particularly relates to a method and a system for uniformly planning and scheduling distributed resources of a combustion chamber of a solid engine. Background In the scheduling and resource planning of the conventional continuous charging production line, heuristic-based multi-objective optimization algorithm is mostly adopted, and cross-process resource coupling, risk disturbance and energy security constraint cannot be effectively described. Particularly in links such as AGV transportation, weighing mixing, pouring solidification and the like, safety and quality fluctuation caused by random disturbance are not systematically controlled. The existing method depends on static MOPSO or NSGA-II algorithms, and lacks distribution robustness and an interpretable rescheduling mechanism. The traditional NSGA-II and MOPSO algorithms have obvious limitations in processing distribution drift and mechanism constraint, including low random solution feasibility, excessively strong disturbance distribution assumption and lack of an interpretable rescheduling mechanism. Specifically, the prior art has the following disadvantages: 1) The cross-process coupling and safety are that the main mixing and pouring gate needs to occupy AGV/hoisting resources synchronously, and beat jitter is conducted to solidification quality and safety margin through temperature control/shear rate; 2) The algorithm and the model are split, namely heuristic/single-target approximation is difficult to process quality and safety probability constraint, disturbance distribution is strongly assumed, and robustness and audit of 'distribution drift' are lacked; 3) The feasible domain is fragile, namely the random solution under the constraint of multi-resource synchronization and time window is not feasible, and the global consistent projection based on a time expansion network is lacked; 4) The mechanism is missing, that is, the energy consumption, the temperature rise, the fatigue and the curing dynamics are not uniformly optimized, only the empirical threshold value is used for control, and the explanation of rearrangement is difficult; Disclosure of Invention In view of the above, the invention aims to provide a method and a system for uniformly planning and allocating distributed resources of a combustion chamber of a solid engine, which aim at solving the problem of multi-source heterogeneous resource coordination in continuous charging production of the combustion chamber of the solid engine, combine hypergraph, opportunity constraint and distributed robust multi-objective optimization, and realize uniform allocation of safety, energy consumption, quality and beat multi-objective through digital twin on-line calibration. In order to achieve the above purpose, the present invention provides the following technical solutions: the invention firstly provides a method for uniformly planning and allocating distributed resources of a combustion chamber of a solid engine, which comprises the following steps: step one, hypergraph modeling and time expansion network construction: mapping multi-source heterogeneous resources and tasks involved in a production line into a time-varying hypergraph Wherein the edge is exceededConstructing a time expansion network based on the hypergraph, and dispersing the time dimension into a time slot setFor each resourceConstruction nodeAnd pass through an arcRepresenting the time between adjacent time slotsForming a joint constraint characterization of the realization task in a time domain and a resource domain; step two, multi-objective optimization modeling: An optimization problem is constructed that includes four objectives: Wherein: Representing tasks Is a time of completion of (a); Representing the total energy consumption of the device; representing a mass loss function; Representing CVaR risk expectations under the wasperstein distribution uncertainty set; Representing a task; representing a control variable; Representing a device state variable; Representing a negative safety margin value; Representing the disturbance; Expressed at a confidence level Lower based on distribution familiesConditional risk expectation operators of (2); Is a fuzzy distribution family; And step three, opportunity constraint and Wasserstein distribution robust processing: For production disturbance variable The Wasserstein spheres were constructed: Wherein: representing a family of fuzzy distributions; And Respectively representing a true distribution and an empirical distribution; Creating strong robust opportunity constraints: , Wherein: Representing the quality; representing a quality threshold; representing a safety threshold; Representing a quality constraint weight coefficient; representi