CN-121997674-A - Accurate forming method of floor support plate

CN121997674ACN 121997674 ACN121997674 ACN 121997674ACN-121997674-A

Abstract

The invention belongs to the technical field of building carrier plate forming, and discloses a building carrier plate precise forming method, which is characterized in that dynamic parameter adaptation and hidden defect pre-judgment are integrated in the whole process from design drawing analysis to real-time production adjustment, hidden hazards such as stress concentration and local buckling are effectively avoided, core indexes such as ultimate bearing capacity and mid-span deflection of a building carrier plate are enabled to reach the standard stably, the product failure rate is reduced, the precision requirement of a high-end building is adapted, real-time linkage of virtual and physical equipment is realized through a 1:1 digital twin system, the multi-node data safety is ensured by a federal learning framework, collaborative optimization is realized, different production requirements can be flexibly adapted by a multi-working-condition simulation scene library and a parameterized model, intelligent optimizing of parameters, dynamic deviation correction and model iteration can be completed without manual intervention in the whole process, the process debugging period is shortened, meanwhile, the whole quantity traceability of production data is realized, the forming process is driven by empirically driven steering data, and the production controllability and the whole efficiency are improved.

Inventors

WEN ZHIXIANG
Hong Junhan
XU JIE

Assignees

杭州杰晟宝建筑围护系统有限公司

Dates

Publication Date: 20260508
Application Date: 20260306

Claims (8)

1. The accurate forming method of the floor support plate is characterized by comprising the following specific steps of: Constructing a multi-physical-field parameterized model of the floor support plate fused with dynamic feature perception, analyzing a design drawing to extract geometric parameters, collecting mechanical parameters of materials and environmental data to form a standardized data set, embedding an attention mechanism to realize dynamic weight assignment of the parameters, and completing model construction through multi-scale grid division; a simulation pre-judging stage, namely constructing a multi-working-condition simulation scene library based on a multi-physical-field fusion model, strengthening simulation precision, identifying forming hidden defects and outputting a multi-dimensional data set; The parameter optimization stage comprises the steps of constructing a mixed optimization model of a genetic algorithm driven by an attention mechanism and a neural network based on a simulation data set, setting an optimization target and a variable, and outputting an optimal parameter combination, a configuration file and a parameter sensitivity report after the verification precision is verified through collaborative optimization; Constructing a digital twin system of a building carrier plate production line of a federal learning architecture, carrying out proportional re-engraving to produce core components, establishing a real-time communication link between a virtual model and physical equipment, training and optimizing the twin model through an edge and cloud aggregation strategy, and realizing virtual-real linkage distributed collaborative optimization; The data preprocessing stage is used for acquiring forming key size data and process data based on a multisource sensor network deployed in the twinning linkage stage, and finishing data noise reduction, abnormal rejection and standardized conversion through an edge node preprocessing module to generate a real-time data stream; The closed-loop control stage comprises the steps of constructing a deviation recognition and dynamic correction module, comparing real-time production data with simulation prediction data to calculate deviation, triggering closed-loop control adjustment equipment parameters when the threshold value is exceeded, feeding back data to update a model, and uploading the model to a cloud; And in the model iteration stage, a knowledge graph and a distributed knowledge graph are constructed to store full-quantity production data, the data are mined to correlate and extract a forming rule, a model iteration trigger mechanism is established, and an incremental learning updating optimization model and a simulation model are adopted.
2. The precise building carrier plate forming method according to claim 1 is characterized in that in the multi-field modeling stage, building carrier plate design drawings are analyzed by adopting BIM technology to extract core geometric parameters and construct a basic three-dimensional model, mechanical parameters and multi-dimensional environment data of substrate materials are synchronously collected, a deep learning material characteristic dynamic predictor model is trained by combining historical material data, mechanical parameters which dynamically change along with working conditions are output, a parameterized framework is adopted to embed attention weight distribution modules, dynamic weight assignment is carried out on the geometric parameters, material dynamic parameters and environment parameters, and model construction is completed by means of differentiated grid division setting of a core stress area and a non-core area and combining dynamic parameters.
3. The precise building carrier plate forming method according to claim 2 is characterized in that in the simulation pre-judging stage, a multi-working-condition simulation scene library comprises combined working conditions of different rolling speeds, pressures and environment temperatures, a self-adaptive grid division algorithm is adopted to optimize grid density according to parameter weights output by an attention module, meanwhile, a material constitutive equation and a thermodynamic equilibrium equation are used as hard constraint conditions to be embedded into a finite element simulation model, a simulation system is used for analyzing key performance indexes, an anomaly detection algorithm is used for identifying hidden defects, and a multi-dimensional data set comprising working condition parameters, simulation results, defect risk grades and defect position coordinates is output.
4. The precise building carrier plate forming method according to claim 3 is characterized in that in the parameter optimization stage, a simulation data set is divided into a training set and a test set, rib height deviation precise control, wave distance uniformity improvement and bearing capacity standard reaching are taken as core optimization targets, rolling speed, pressure, die clearance and temperature compensation quantity are selected as key optimization variables, self-attention selection operators are embedded in a genetic algorithm, cross and variation probabilities are adaptively adjusted through association degrees of parameters and the optimization targets, optimized parameters are input into a multi-layer perception machine model to output forming quality prediction results, collaborative optimization is achieved through alternate training, and optimal forming parameter combinations, standardized parameter configuration files and parameter sensitivity analysis reports are output after accuracy standard reaching through verification of the test set.
5. The precise building carrier plate forming method according to claim 4 is characterized in that in the twinning linkage stage, a digital twinning model is used for producing a core component and relevant key characteristics including physical characteristics, motion characteristics and perception characteristics of the core component according to 1:1 re-carving, after an optimal parameter configuration file is imported, operation state data of low-delay real-time communication link synchronous equipment is established through an industrial Ethernet, a production line is divided into a plurality of edge nodes according to functions, the edge nodes build a data safety isolation area locally to train a local twinning model, model parameters are only encrypted and uploaded to a cloud, the cloud aggregates the local model through a weighted average algorithm to generate a global optimization twinning model, and the optimization parameters and control strategy are encrypted and issued to the edge nodes.
6. The precise building carrier plate forming method according to claim 5 is characterized in that in the data preprocessing stage, forming key size data and process data are collected in a forming key area based on a multi-source sensor network deployed in a twin linkage stage, a preprocessing module of an edge node integrates a noise reduction algorithm and an outlier rejection algorithm, after invalid data are filtered, the multi-source heterogeneous data are converted into standardized data in a unified format through a standardized algorithm, real-time data streams meeting interaction requirements are generated, and data transmission delay is monitored.
7. The precise building carrier plate forming method according to claim 6 is characterized in that in the closed-loop control stage, similarity of real-time production data and simulation prediction data is calculated by adopting a cosine similarity algorithm, whether deviation exceeds standard is judged by combining a preset similarity threshold, meanwhile, specific key indexes including size deviation and pressure deviation are obtained by adopting a difference calculation method, the preset deviation threshold is determined according to historical production data and quality standards, when the deviation exceeds the threshold, parameter correction is reversely deduced through a digital twin model, a correction instruction is generated by combining an optimization direction output by a federal learning global model, key equipment parameters are adjusted in real time, corrected production data are fed back to each edge node update model, and correction records are encrypted and uploaded to a cloud.
8. The precise building carrier plate forming method according to claim 7 is characterized in that in the model iteration stage, a distributed knowledge base is stored in an edge node and a cloud in a scattered mode by adopting a federal learning architecture, the edge node stores local full-scale production data, the cloud stores global commonality data and standardized knowledge patterns, the knowledge patterns are used for constructing a structured knowledge network by taking parameters, working conditions, defects and schemes as core association dimensions, data hidden association is mined through a graph neural network, forming rules are extracted, a model iteration trigger mechanism is constructed based on the knowledge patterns, the degree of difference between new batch data and the existing knowledge is analyzed after the new batch data are put in storage, the federal learning model iteration flow is triggered when a preset threshold value is exceeded, model parameter weights are updated by incremental learning, and new knowledge is supplemented to the knowledge patterns.

Description

Accurate forming method of floor support plate Technical Field The invention belongs to the technical field of floor support plate forming, and particularly relates to a precise floor support plate forming method. Background The building carrier plate is used as a core bearing member of an assembled building, and the forming precision of the building carrier plate directly relates to the safety, stability and construction efficiency of a building structure. At present, the floor support plate is molded by adopting a traditional mechanical rolling process, and the molding is finished by depending on a preset fixed die and technological parameters, so that the following technical problems exist. The parameter design of the traditional process is mostly based on manual experience, the influence of variables such as material characteristic fluctuation, environmental temperature change and the like on the molding precision cannot be fully considered, the quality problems such as rib height deviation, uneven wave distance and the like of products are easy to occur, the conventional error is large, and the severe requirements of super high-rise and large-span buildings on millimeter-level precision are difficult to meet. Along with the deep promotion of building industrialization, digital technologies such as BIM technology, finite element simulation and the like are gradually applied in the field of building, but the deep fusion in the field of building carrier plate forming still has obvious defects. In the prior art, the digital technology is only applied to the early design modeling link, the full-flow closed loop control from design, simulation to production execution and quality feedback cannot be realized, the finite element simulation model is developed for a single specification floor support plate, a reusable and iteratable parameterized model library is not formed, and the simulation result has larger deviation from actual molding, so that the actual production work is difficult to be effectively guided. Disclosure of Invention The invention aims to provide a precise building carrier plate forming method for solving the problems in the background technology. In order to achieve the purpose, the invention provides the following technical scheme that the accurate forming method of the floor support plate comprises the following specific steps: Preferably, the multi-field modeling stage constructs a multi-physical-field parameterized model of the building carrier plate fused with dynamic feature perception, wherein the multi-physical-field specifically comprises a structural mechanical field (describing stress and strain distribution in the process of forming a substrate) and a thermodynamic field (describing the influence of temperature change on material characteristics and forming precision), and the structural mechanical field and the thermodynamic field are fused through a parameter coupling relation, namely, temperature data of the thermodynamic field is used as input parameters of the structural mechanical field, and the mechanical characteristic parameters of the material are dynamically corrected to ensure that the model is attached to multi-field coupling effects in actual forming; The specific coupling logic is that when the temperature of the substrate changes by 5 ℃ in the environment temperature or the forming process, parameters such as elastic modulus, poisson ratio and the like in a mechanical field of the structure are synchronously corrected through a material characteristic dynamic predictor model, the correction amplitude is in linear correlation with the temperature change quantity, real-time linkage of multi-field data is ensured, and the parameter coupling effect in actual forming is attached. Adopting an attention mechanism to realize self-adaptive focusing of key parameters, analyzing a floor support plate design drawing through a BIM technology, extracting core geometric parameters such as rib height, wave distance, coverage width and the like, constructing a basic three-dimensional model, synchronously acquiring mechanical parameters such as stress and strain curves, temperature sensitivity coefficients, poisson's ratio and the like of substrate materials, and combining multidimensional environmental data such as real-time environmental temperature and humidity, atmospheric pressure and the like to form a standardized multidimensional parameter data set; Constructing a dynamic material characteristic predictor model based on deep learning, training the model through historical material data, inputting environmental parameters and material base parameters, and outputting mechanical parameters which dynamically change along with working conditions; The submodel adopts a 3-layer fully-connected neural network architecture, the number of neurons of an input layer is the total dimension of environmental parameters (temperature, humidity and atmospheric pressure) and material basic pa