A smart factory modular foundation planning method

Abstract

The application belongs to the field of intelligent factory construction and is an intelligent factory modular foundation planning method, comprising the following steps: S1, based on the full life cycle demand of the intelligent factory, multi-dimensional demand modeling is completed, a digital twin foundation framework is constructed, regional planning is designed, and intelligent positioning devices and adaptive connection devices are pre-buried in each foundation; S2, the basic functions are selected, and the matching of the functions and production devices is completed through an AI algorithm; S3, the estimated area of the foundation and the equipment installation distribution map are calculated according to the selected devices, the planning layout and connection mode design are completed, and simulation verification is performed by using the digital twin foundation framework; S4, whether the production land area, connection mode and simulation result meet the standards is judged; S5, the intelligent positioning device is used to guide the automatic splicing of the foundation according to the connection mode, dynamic calibration is completed and is synchronized to the digital twin foundation framework; S6, the full life cycle adaptation verification is completed through the digital twin foundation framework, an operation and maintenance plan is generated, and the planning is completed if the standards are met.

Description

Technical Field

[0001] This invention belongs to the field of smart factory construction, and in particular, it is a modular foundation planning method for smart factories. Background Technology

[0002] In existing technologies, the foundation construction of smart factories mostly adopts traditional fixed structure design. Once the foundation layout and function are determined, they are difficult to adjust and cannot meet the needs of later expansion, function upgrade or production process optimization of smart factories.

[0003] While some modular foundation technologies enable the splitting and combination of foundations, the planning process lacks a systematic consideration of the needs throughout the entire life cycle, focusing only on the functional realization of the initial construction phase and failing to fully take into account the convenience of operation and maintenance and the flexibility of future expansion.

[0004] Meanwhile, the matching of functions and production equipment relies heavily on manual experience or simple parameter comparisons, lacking the support of intelligent optimization algorithms. This can easily lead to problems such as mismatch between equipment and foundation functions and low space utilization. During the splicing process, path planning lacks precise control over smoothness and collision risks. The calibration accuracy is limited by a single sensor, making it difficult to meet the requirements of high-precision installation. Furthermore, it lacks deep integration with digital twin technology, resulting in asynchronous virtual and real data, which affects the verification and iteration of planning schemes.

[0005] With the advancement of Industry 4.0, smart factories are developing towards flexibility, intelligence, and full life cycle management, which places higher demands on the modularity, planning accuracy, adaptability, and ease of operation and maintenance of the foundation.

[0006] Traditional fixed foundations can no longer meet the needs of frequent functional adjustments and capacity upgrades in smart factories. Existing modular foundation planning methods have significant shortcomings in terms of demand quantification, functional matching, path planning, accuracy calibration, and virtual-physical collaboration, resulting in high foundation construction costs, difficulty in adjustment, and low operation and maintenance efficiency, which restricts the overall operational efficiency and development potential of smart factories. Summary of the Invention

[0007] This invention proposes a modular foundation planning method for intelligent factories.

[0008] A modular foundation planning method for a smart factory includes the following steps: S1. Based on the full life cycle requirements of the smart factory, complete multi-dimensional requirement modeling, construct a digital twin foundation framework, design regional planning with multiple foundations, and pre-embed intelligent positioning devices and adaptive connection devices in each foundation. S2. Based on the digital twin foundation framework, select basic functions and use AI algorithms to achieve optimal matching between functions and production equipment; S3. Calculate the estimated foundation area and equipment installation distribution diagram based on the selected device, complete the planning layout and connection method design through multi-objective optimization algorithm, and verify the design using the digital twin foundation frame simulation. S4. Determine whether the production land area, connection method and simulation results meet the standards. If they do not meet the standards, feedback is sent to S2. S5. Based on the connection method, the foundation is automatically spliced ​​through an intelligent positioning device, and dynamic calibration is completed by combining multi-sensor data and synchronized to the digital twin foundation frame. S6. Complete the full lifecycle adaptation verification and generate operation and maintenance plans through the digital twin foundation framework. If the standards are not met, feedback is sent to S2. If the standards are met, the planning is completed.

[0009] Preferably, step S1 includes the following sub-steps: S11. A multi-scenario demand weight allocation algorithm is adopted to decompose and prioritize production capacity, ease of operation and maintenance, and future expansion needs, and to construct a dynamic mapping model between demand parameters and foundation physical parameters to achieve accurate conversion of demand parameters to foundation physical parameters. S12. Lightweight digital twin modeling technology is used to construct a digital twin of the foundation module, and a standardized Internet of Things interface is developed to achieve low-latency data transmission and interface adaptation for intelligent positioning devices and adaptive connection devices.

[0010] Preferably, the optimal matching of functions and production devices in S2 is achieved by: constructing a function and device matching dataset containing function parameters, device parameters and adaptability scores; training a matching model using a random forest algorithm; inputting selected basic function parameters; and outputting a list of production devices with the highest adaptability scores by fusing the prediction results and confidence weights of multiple decision trees.

[0011] Preferably, step S3 includes the following sub-steps: S31. Based on the dynamic adjustment and optimization of target weights for production load in smart factories, establish a coupled model of foundation area and connection stability to achieve coordinated optimization of production capacity, energy consumption, and logistics efficiency. S32. Introduce a time dimension to construct a dynamic simulation scenario, simulate the load changes and connection reliability of the foundation at different production stages, calculate the simulation error and use corresponding algorithms to correct the planning parameters, thereby improving the planning accuracy.

[0012] Preferably, in step S4, it is determined whether the production land area, connection method, and simulation results meet the standards. If they do not meet the standards, feedback is provided. This step specifically includes the following sub-steps: S41. Feed back the unapproved expansion and maintenance requirements data to the requirement modeling module to drive it to requantify the requirement parameters and supplement the maintenance parameters. S42. Feed back the data on area gap and insufficient connection stability to the functional device selection module to drive it to re-match the device. S43. Feed back the simulation error data to the multi-objective optimization planning module to drive it to solve again.

[0013] Preferably, step S5 includes the following sub-steps: S51. A path planning algorithm that balances cost and smoothness is used to process the location data collected by the intelligent positioning device to generate the optimal splicing path. The adaptive connection device collects docking pressure data in real time and achieves adaptive compensation for docking deviation through a deviation compensation algorithm to ensure high-precision installation standards of the foundation. S52 integrates multi-sensor data from GPS, laser positioning, and inertial navigation, and uses a precise calibration algorithm to achieve high-precision calibration of the foundation position. The calibration data is then input back into the digital twin foundation framework to complete the iterative update of the digital twin.

[0014] Preferably, in step S6, the full lifecycle adaptation verification includes: completing it by simulating operation and maintenance failure scenarios, expansion device addition scenarios, and structural adjustment scenarios; the operation and maintenance plan is generated by fault tree analysis method, including fault warning thresholds, module replacement process and parameter adjustment scheme, and determining the key fault causes and countermeasures based on real-time monitoring data of the digital twin foundation framework.

[0015] Preferably, the digital twin foundation framework and the multi-objective optimization planning module include real-time interaction: the digital twin foundation framework synchronizes the foundation positioning accuracy and connection interface status data to the multi-objective optimization planning module, providing real data input for the optimization algorithm; the planning scheme output by the multi-objective optimization planning module is transmitted to the digital twin foundation framework for simulation verification, and after simulation, parameter correction instructions and interference problem data are generated and transmitted back to drive the optimization algorithm iteration and replanning of the splicing path.

[0016] Preferably, the functional device selection module and the multi-objective optimization planning module include bidirectional feedback: the functional device selection module transmits device size, energy consumption, and production capacity parameters to the multi-objective optimization planning module as planning boundary constraints; after identifying device compatibility issues, the multi-objective optimization planning module feeds back the problem type and compatibility deviation data to the functional device selection module, driving it to re-match devices, and at the same time transmits the optimal energy consumption solution to the functional device selection module to guide it in adjusting energy consumption priority weights.

[0017] As a preferred embodiment, the multi-dimensional satisfaction judgment module and the intelligent splicing module include feedback interaction: the detection data of insufficient connection stability and production land area not meeting the standard are fed back to the intelligent splicing module, driving it to adjust the connection pressure parameters, calibrate the accuracy threshold, and re-plan the foundation splicing combination method; the passed position accuracy data is transmitted to the intelligent splicing module as standard parameters to solidify the calibration process.

[0018] The present invention has the following beneficial effects: 1. This invention achieves accurate conversion of demand parameters into foundation physical parameters through the deep integration of full life cycle demand modeling and digital twin framework. Combined with intelligent optimization algorithms, it completes the optimal matching of functions and devices, enabling the planning scheme to dynamically adapt to changes in production load and demand adjustments at different stages. This significantly improves the adaptability and accuracy of foundation planning and reduces the cost of later adjustments.

[0019] 2. This invention significantly improves the accuracy and stability of foundation splicing through multi-sensor fusion precision calibration technology and a path planning method that balances cost and smoothness, ensuring that the foundation installation quality meets millimeter-level requirements. At the same time, by leveraging digital twin technology, it achieves virtual-real synchronization of the entire process of planning, splicing, and verification, providing reliable support for the simulation verification of planning schemes and the generation of operation and maintenance plans, effectively improving operation and maintenance efficiency and ensuring the long-term stable operation of the smart factory. Attached Figure Description

[0020] Figure 1 This is a schematic diagram illustrating the steps of a modular foundation planning method for an intelligent factory according to the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions in the embodiments of this invention will be clearly described below in conjunction with the examples.

[0022] Example 1 like Figure 1 As shown, this invention proposes a modular foundation planning method for smart factories, which specifically includes the following steps: S1. Demand modeling and digital twin foundation framework construction: Based on the full life cycle requirements of the smart factory, complete multi-dimensional demand modeling, construct a digital twin foundation framework, design regional planning with multiple foundations, and pre-embed intelligent positioning devices and adaptive connection devices with integrated IoT modules in each foundation. S2. Functional device matching optimization and intelligent selection: Based on the digital twin foundation framework, basic functions are selected, and AI algorithms are used to achieve the optimal matching between functions and production devices. S3. Multi-objective optimization planning and simulation verification: Calculate the estimated foundation area and equipment installation distribution diagram based on the selected device, complete the planning layout and connection method design through multi-objective optimization algorithm, and verify the digital twin foundation frame through simulation. S4. Multi-dimensional satisfaction judgment and feedback optimization: judge whether the production land area, connection method and simulation results meet the standards. If they do not meet the standards, feedback is sent to S2. S5. Intelligent splicing and dynamic calibration: Based on the connection method, the intelligent positioning device guides the automatic splicing of the foundation, and combines multi-sensor data to complete dynamic calibration and synchronize it to the digital twin foundation frame. S6. Full lifecycle adaptation verification and operation and maintenance plan generation: The full lifecycle adaptation verification is completed and the operation and maintenance plan is generated through the digital twin foundation framework. If the standard is not met, feedback is sent to S2. If the standard is met, the planning is completed.

[0023] Furthermore, step S1 also includes S11 multi-dimensional demand modeling and parameter quantification, specifically: using a multi-scenario demand weight allocation algorithm to decompose and prioritize the production capacity, operation and maintenance convenience, and future expansion needs of the smart factory, constructing a dynamic mapping model between demand parameters and foundation physical parameters, and realizing the accurate conversion of demand parameters to foundation physical parameters. The demand weight allocation uses the analytic hierarchy process (AHP) to construct a judgment matrix. ,in: This is a matrix for judging the relative importance of demand. To determine the importance of the i-th requirement relative to the j-th requirement, Number of demand types; Calculate the weight vector using an improved eigenvalue method. ,satisfy ,in: This is the demand weight vector. These are the weight values ​​for each requirement. This is the demand decay factor, which characterizes the degree of decay of demand throughout its entire life cycle. For the validity period of the demand, To determine the largest eigenvalue of a matrix; Consistency test ,in: As a consistency indicator, To determine the largest eigenvalue of a matrix, Number of demand types; ,in: For consistency ratio, As a consistency indicator, The average random consistency index is used; the final weight is then obtained.

[0024] Furthermore, step S1 also includes S12 initialization of the digital twin foundation framework, specifically: constructing a digital twin of the foundation module using lightweight digital twin modeling technology to achieve real-time synchronization of foundation physical properties, connection status, and equipment operation data; Develop standardized IoT interfaces to enable low-latency transmission of positioning data and connection status of the intelligent positioning device and adaptive connection device, and complete interface adaptation with the digital twin foundation framework.

[0025] Furthermore, the optimal matching of functions and production devices in step S2 is achieved through an AI algorithm, specifically: constructing a function and device matching dataset, including the function parameters, device parameters and adaptability scores of historical matching cases, training the matching model using the random forest algorithm, inputting the selected basic function parameters, and outputting a list of production devices with the highest adaptability scores; The predicted output of the random forest model is: ,in: The overall score for device adaptability is given. For the number of decision trees, The confidence weight of the k-th decision tree is determined by its prediction accuracy on historical datasets. The prediction result of the k-th decision tree, i.e., the device adaptability score, For input function parameters;

[0026] Furthermore, step S3 also includes S31 deployment of a multi-objective optimization planning algorithm, specifically: based on the production load of the smart factory, dynamically adjust the optimization target weights of capacity, energy consumption, and logistics efficiency, establish a coupled model of foundation area and connection stability, and realize the coordinated optimization of foundation area demand and connection stability. Wherein, the coupling optimization objective function is: in: To optimize the overall objective value for multiple objectives, These are planning variables, including foundation dimensions and connection point locations. These are dynamic weights for capacity, energy consumption, and logistics efficiency, which are adjusted in real time according to production load. These are the capacity deviation function, energy consumption quantification function, and logistics efficiency deviation function, respectively. The coupling coefficient between the i-th and j-th objectives represents the degree of mutual influence between the objectives, such as the positive correlation between production capacity and energy consumption.

[0027] Furthermore, step S3 also includes S32 digital twin simulation verification and iteration, specifically: introducing a time dimension to construct a dynamic simulation scenario to simulate the load changes and connection reliability of the foundation at different production stages; The simulation error is calculated using the virtual-to-real comparison data of the digital twin, and the planning parameters are automatically corrected to improve planning accuracy; wherein, the simulation error is calculated as follows: ,in: For simulation error, For the number of data samples, Let be the stage weight for the i-th sample, with higher weights for samples during peak production periods than during off-peak periods. For physical entities, measured data, such as foundation bearing capacity and connection stiffness, For simulation data; Based on error Gradient descent method ,in: These are the planning parameters after iteration. For current planning parameters, For learning rate, For the objective function in The gradient at the point; adjust the planning parameters.

[0028] Furthermore, step S5 also includes S51 intelligent stitching path planning and execution, specifically: using the A* algorithm to process the location data collected by the intelligent positioning device, generating the optimal stitching path for foundation movement, and avoiding path redundancy and collisions. The adaptive connection device integrates a pressure sensing module to collect docking pressure data in real time and achieves adaptive compensation for docking deviation through a deviation compensation algorithm; wherein, the evaluation function of the A* algorithm is: ,in: This is the path evaluation value. For path nodes, From the starting point to the node The actual cost of movement, including energy consumption and time. This is the path smoothing coefficient, with a value ranging from 0.1 to 0.3. For nodes The path curvature at a given point represents the smoothness of the path. For nodes The estimated cost to the destination is calculated using an improved Euclidean distance method.

[0029] Furthermore, step S5 also includes S52 dynamic calibration and twin synchronization update, specifically: fusing multi-sensor data from GPS, laser positioning, and inertial navigation, and using a Kalman filter algorithm to achieve millimeter-level precision calibration of the ground position; The calibration data is then input back into the digital twin foundation framework to complete the iterative update of the digital twin, improving the consistency between the virtual and real systems; wherein, the state update equation of the Kalman filter is: ,in: The optimal state estimate at time k is the precise location of the foundation. The weights for prior state confidence are determined by the sensor accuracy level, with the highest laser positioning accuracy corresponding to the largest weight. Let k be the prior state estimate. For Kalman gain, For multi-sensor fusion observations, This is the observation matrix.

[0030] Furthermore, step S6 constructs a full lifecycle adaptation verification system to simulate fault scenarios in the operation and maintenance phase, device addition scenarios in the expansion phase, and structural adjustment scenarios in the transformation phase, thereby completing the adaptation verification; based on the real-time monitoring data of the digital twin foundation framework, an operation and maintenance plan is generated using the fault tree analysis method, including fault warning thresholds, module replacement procedures, and parameter adjustment schemes.

[0031] Furthermore, the digital twin foundation framework synchronizes the foundation positioning accuracy and connection interface status data to the multi-objective optimization planning module in real time, providing real data input for the multi-objective optimization algorithm. The planning scheme output by the multi-objective optimization planning module is transmitted to the digital twin foundation framework for simulation verification. After the simulation is completed, parameter correction instructions are generated and sent back to the multi-objective optimization planning module to drive the optimization algorithm iteration. After identifying foundation interference problems during the simulation, the digital twin foundation framework feeds back the interference position and range data to the planning module, and the planning module replans the splicing path based on the data.

[0032] Furthermore, the functional device selection module transmits the selected device size, energy consumption, and production capacity parameters to the multi-objective optimization planning module as planning boundary constraints. After identifying device compatibility issues during the planning process, the multi-objective optimization planning module feeds back the problem type and compatibility deviation data to the functional device selection module, driving the selection module to re-match devices. After calculating the optimal energy consumption solution, the multi-objective optimization planning module transmits the result to the functional device selection module, guiding the selection module to adjust the energy consumption priority weights.

[0033] Furthermore, the multi-dimensional satisfaction judgment module feeds back the data related to unapproved expansion and operation and maintenance requirements to the requirement modeling module, driving the requirement modeling module to requantify the corresponding requirement parameters; the multi-dimensional satisfaction judgment module outputs the requirement satisfaction data of the planning scheme to the requirement modeling module, and the requirement modeling module updates the requirement priority weights based on the data; when the operation and maintenance requirements are not met, the multi-dimensional satisfaction judgment module outputs the missing operation and maintenance requirement parameter information to the requirement modeling module, driving the requirement modeling module to supplement the operation and maintenance parameters, thereby optimizing the operation and maintenance monitoring dimensions of the digital twin foundation framework.

[0034] Furthermore, the intelligent stitching module synchronizes the real-time location data of the stitching process to the digital twin foundation frame, which then constructs a virtual-real synchronized stitching visualization scene based on this data. When the digital twin foundation frame predicts potential collision risks during the simulation of the stitching process, it transmits the risk nodes and warning level data to the intelligent stitching module, which then adjusts the stitching path based on this data. After completing the stitching, the intelligent stitching module transmits the physical connection status data to the digital twin foundation frame for remote verification via the digital twin.

[0035] Furthermore, the full lifecycle adaptation verification module feeds back the foundation position offset data after long-term operation to the intelligent splicing module, which optimizes calibration parameters based on this data. In the operation and maintenance plan generated by the full lifecycle adaptation verification module, the data related to module replacement requirements is transmitted to the intelligent splicing module, driving the intelligent splicing module to preset disassembly and reinstallation paths. When the expansion adaptation verification fails, the full lifecycle adaptation verification module feeds back the size and installation location data of the new device to the intelligent splicing module, which readjusts the foundation layout based on this data.

[0036] Furthermore, the multi-dimensional satisfaction judgment module feeds back detection data indicating insufficient connection stability to the intelligent splicing module, which adjusts the connection pressure parameters and calibration accuracy threshold based on this data. When the multi-dimensional satisfaction judgment module identifies that the production land area does not meet the standard, it feeds back the area gap data to the intelligent splicing module, driving the intelligent splicing module to re-plan the foundation splicing combination method to expand the usable area. The multi-dimensional satisfaction judgment module transmits the passed position accuracy data to the intelligent splicing module, which uses this data as a standard parameter to solidify the calibration process.

[0037] Example 2 This embodiment focuses on the full-process planning and implementation of modular foundations for smart factories, achieving precise planning, intelligent assembly, and full lifecycle adaptation of foundation modules. The specific technical implementation process is as follows: The first step is to perform step S1, demand modeling and digital twin foundation framework construction. Based on the full lifecycle requirements of the smart factory, multi-dimensional demand modeling and the initial construction of the digital twin foundation framework are completed. The original content clearly states that this step includes two sub-steps: S11 multi-dimensional demand modeling and parameter quantification, and S12 digital twin foundation framework initialization. It also requires the design of regional planning with multiple foundations and the pre-installation of corresponding devices.

[0038] To accurately capture multi-dimensional needs, the demand survey module first collects information on the production capacity needs, operation and maintenance convenience needs, and future expansion needs of the smart factory to form an initial demand set; based on this initial demand set, the S11 multi-dimensional demand modeling and parameter quantification sub-step is executed.

[0039] The demand weight allocation uses the analytic hierarchy process (AHP), first constructing a judgment matrix. Here, n is set to 3, corresponding to the three core requirements of production capacity, ease of operation and maintenance, and future expansion. The importance of factors such as production capacity relative to ease of operation and maintenance is determined through expert scoring. In terms of production capacity, ease of operation and maintenance is more important. This is a matrix for judging the relative importance of demand. To determine the importance of the i-th requirement relative to the j-th requirement, In this embodiment, n=3, which represents the number of demand types.

[0040] To adapt to the dynamic changes in the needs of a smart factory throughout its entire lifecycle, a demand time decay coefficient is introduced based on the traditional eigenvalue method for calculating the weight vector. and validity period The improved formula for calculating the weight vector is obtained. ,in: This is the demand weight vector. The weighted values ​​are production capacity, ease of operation and maintenance, and future expansion needs, respectively. This is the demand decay factor, which characterizes the degree of decay of demand throughout its entire life cycle. For the validity period of the demand, To determine the largest eigenvalue of a matrix.

[0041] Calculate the weight vector Afterwards, a consistency check is required, and it must pass. Calculate the consistency index ,in: As a consistency indicator, To determine the largest eigenvalue of a matrix, In this embodiment, n=3, representing the number of demand types. Combined with the average random consistency index In this embodiment, when n=3 Calculate the consistency ratio ,in: For consistency ratio, As a consistency indicator, The average random consistency index; when If the matrix satisfies the consistency requirement, the weight vector is valid; otherwise, it needs to be readjusted. The value is taken and the calculation is repeated.

[0042] Based on the effective weight vector, a dynamic mapping model between demand parameters and foundation physical parameters is constructed, which transforms the quantified demand parameters into feasible foundation physical parameters.

[0043] Among them, the production capacity demand parameters are converted into physical parameters such as foundation bearing pressure and foundation area; the operation and maintenance convenience demand parameters are converted into physical parameters such as foundation module disassembly space and connection device operation space; and the future expansion demand parameters are converted into physical parameters such as the number of reserved foundation interfaces and reserved space dimensions.

[0044] After completing sub-step S11, execute sub-step S12, which initializes the digital twin foundation framework, and uses lightweight digital twin modeling technology to construct a digital twin of the foundation module.

[0045] Specifically, firstly, geometric feature data of the foundation module is collected using 3D laser scanning technology, and then a 3D geometric model of the foundation module is constructed using BIM technology. Then, based on the Industrial Internet protocol, the sensing data interfaces of the intelligent positioning device and adaptive connection device pre-embedded in the foundation are associated with the 3D geometric model, and physical attribute mapping relationships are added, so that the digital twin can synchronize the physical attributes, connection status and equipment operation data of the foundation in real time.

[0046] Simultaneously, a standardized IoT interface is developed, and the MQTT communication protocol is used to realize low-latency data transmission between intelligent positioning devices, adaptive connection devices and digital twin ground-based frameworks. During the interface adaptation process, the communication protocol negotiation, data format unification and transmission rate debugging need to be completed to ensure that the transmission latency of positioning data and connection status data is controlled within 100ms.

[0047] Based on the above requirements modeling results and digital twin foundation framework, a regional plan with multiple foundations is designed. The initial position of each foundation module is determined according to the site foundation conditions and the physical parameters transformed by the requirements. The pre-embedded intelligent positioning device integrates GPS, laser positioning, and inertial navigation multi-sensor components, and the adaptive connection device integrates pressure sensing module and drive component.

[0048] Based on the demand quantification parameters and digital twin foundation framework output in step S1, step S2, functional device matching optimization and intelligent selection, is executed. The basic functions of the smart factory are selected in conjunction with the digital twin foundation framework, and the optimal matching between functions and production devices is achieved through AI algorithms.

[0049] The original content clearly states that the core of this step is the training and application of a function and device matching model based on the random forest algorithm. It is necessary to construct a function and device matching dataset and output a list of production devices with the highest suitability scores.

[0050] In the specific implementation process, based on the basic functional requirements determined in step S1, the core basic functions of the smart factory are first selected, including material transfer function, processing and production function, testing and inspection function, etc.; then, a function and device matching dataset is constructed. The samples in the dataset are historical function and device matching cases, and each sample contains data in three dimensions: functional parameters, device parameters and adaptability score.

[0051] The functional parameters include function type, processing capacity, and operating environment requirements, while the device parameters include device size, energy consumption index, production capacity parameters, and compatible function types. The compatibility score is obtained from the evaluation of historical application effects, with a value range of 0 to 10. The higher the score, the stronger the compatibility.

[0052] A random forest matching model was trained based on this dataset. The training process is as follows: the dataset was divided into training and test sets in a 7:3 ratio, and the number of decision trees was set accordingly. The bootstrap sampling method is used to extract 100 sample subsets from the training set, and each sample subset corresponds to a decision tree for training. During the construction of the decision tree, the Gini coefficient is used as the feature selection criterion. The splitting feature of each node is determined by calculating the Gini coefficient gain of different features until the decision tree reaches the preset depth or the number of node samples is less than the preset threshold.

[0053] After training, the prediction accuracy of each decision tree on the test set is calculated, and the prediction accuracy is used as the confidence weight of that decision tree. ,Right now , Let be the test set accuracy of the k-th decision tree.

[0054] In the model application phase, the basic functional parameters selected in step S1 are... Input the trained random forest model, and use the formula The device compatibility score is calculated, where: The overall score for device adaptability is given. In this embodiment, the number of decision trees is [number]. , Let the confidence weights of the k-th decision tree be denoted as . The prediction result of the k-th decision tree, i.e., the device adaptability score, The input parameters are the basic functional parameters; the output is the top 3 production units with the highest scores as a candidate list.

[0055] The candidate list is transmitted to the digital twin foundation framework, and the installation feasibility of each candidate device in the foundation module is simulated through the digital twin, so as to finally determine the list of production devices with the strongest adaptability.

[0056] After the functional devices are selected, step S3, multi-objective optimization planning and simulation verification, is executed. Based on the selected devices, the estimated foundation area and equipment installation distribution diagram are calculated. The planning layout and connection method design are completed through multi-objective optimization algorithms, and simulation verification is performed using a digital twin foundation frame.

[0057] This step includes two sub-steps: S31 deployment of multi-objective optimization planning algorithm and S32 digital twin simulation verification and iteration. The original content clearly requires the introduction of dynamic weights and coupling models of production load, as well as dynamic simulation in the time dimension.

[0058] In practice, the production equipment list output in step S2 is obtained first, and the core parameters such as size, energy consumption, and production capacity of each equipment are extracted. The estimated area of ​​a single foundation module is calculated, and the number and distribution range of foundation modules are initially determined in combination with the number of equipment and the requirements for compact layout.

[0059] Based on smart factory production load data, the S31 multi-objective optimization programming algorithm is deployed in a sub-step to dynamically adjust the optimization objective weights of capacity, energy consumption, and logistics efficiency. .

[0060] Production load data is collected in real time through a digital twin foundation framework. When the production load exceeds 80%, the capacity weight is increased. Reduce the energy consumption weight to 0.4. Up to 0.3, logistics efficiency weight Up to 0.3; When production load is below 50%, increase the energy consumption weight. Reduce capacity weight to 0.4. Up to 0.3, logistics efficiency weight Up to 0.3; When the production load is between 50% and 80%, the weights of all three are set to 0.33.

[0061] Simultaneously, a coupled model of foundation area and connection stability is established, and a multi-objective optimization objective function is constructed. in: To optimize the overall objective value for multiple objectives, These are planning variables, including parameters such as foundation dimensions, connection point locations, and device placement coordinates. These are the dynamic weights for production capacity, energy consumption, and logistics efficiency, respectively. Let this be the capacity deviation function. Let quantize energy consumption. Let the logistics efficiency deviation function be... Let be the coupling coefficient between the i-th objective and the j-th objective, where production capacity and energy consumption are positively correlated. Production capacity and logistics efficiency are negatively correlated. Energy consumption and logistics efficiency are negatively correlated. ;in, , In order to plan production capacity, To meet demand and production capacity, , For the number of devices, Let m be the energy consumption of the m-th device. , To plan the path length, This represents the optimal path length.

[0062] The genetic algorithm was used to solve the optimization objective function. The population size was set to 50, the number of iterations was set to 100, the crossover probability was set to 0.8, and the mutation probability was set to 0.1. Finally, the optimal foundation planning layout and connection method design scheme was obtained.

[0063] The optimal planning scheme is input into the digital twin foundation framework, and the S32 digital twin simulation verification and iterative sub-steps are executed to introduce the time dimension to construct a dynamic simulation scenario.

[0064] The simulation scenarios cover different production stages throughout the entire lifecycle of a smart factory, including the initial production phase, stable operation phase, peak production phase, and maintenance phase. Each stage has corresponding production loads and equipment operating status parameters. During the simulation, simulation data from the digital twin is collected in real time. Simultaneously, physical entity measurement data for the corresponding stage are obtained through physical experiments. Factors such as foundation bearing capacity, connection stiffness, and logistics transportation time.

[0065] Based on this data, through the formula Calculate simulation error ,in: For simulation error, For the number of data samples, The stage weight for the i-th sample, during peak production periods. Stable operation period In the early stages of production Maintenance period , For physical entity measured data, This is simulation data.

[0066] When simulation error When using the gradient descent method Adjust the planning parameters, where: These are the planning parameters after iteration. For current planning parameters, The learning rate is set to 0.01 in this embodiment. For the objective function in the current planning parameters The gradient at a given point is calculated using numerical differentiation; the above simulation and correction process is repeated until the simulation error is corrected. The approved planning scheme was obtained.

[0067] During the simulation verification process in step S3, the digital twin foundation framework and the multi-objective optimization planning module maintain real-time interaction.

[0068] The digital twin foundation framework synchronizes the foundation positioning accuracy and connection interface status data to the multi-objective optimization planning module in real time via a standardized IoT interface, providing real physical constraint data for the multi-objective optimization algorithm. The initial planning scheme output by the multi-objective optimization planning module is transmitted to the digital twin foundation framework for simulation verification. After the simulation is completed, parameter correction instructions are generated, including the planning variables to be adjusted and the adjustment range, and sent back to the multi-objective optimization planning module to drive the optimization algorithm iteration. If the digital twin foundation framework identifies foundation interference problems during the simulation, it feeds back the three-dimensional coordinates of the interference location and the size data of the interference range to the planning module. Based on this data, the planning module readjusts the position parameters of the foundation module and replans the splicing path.

[0069] After completing the simulation verification in step S3, step S4, multi-dimensional satisfaction judgment and feedback optimization, is executed to determine whether the production land area, connection method and simulation results meet the standards.

[0070] The original content clearly states that the core of this step is multi-dimensional satisfaction judgment and feedback optimization, and the judgment results need to be fed back to the corresponding module.

[0071] In practical implementation, a multi-dimensional satisfaction judgment index system is constructed, including the production land area satisfaction index, connection method satisfaction index, and simulation result satisfaction index. The production land area satisfaction index is determined by comparing the planned total foundation area with the area parameters transformed from the demand in step S1. When the absolute value of the difference between the planned area and the demand area is ≤10%, it is judged as satisfactory. The connection method satisfaction index is determined by checking whether the bearing capacity and connection stiffness of the connection points reach the design threshold. When the bearing capacity of the connection points is ≥ the demand bearing capacity and the connection stiffness is ≥ the design stiffness, it is judged as satisfactory. The simulation result satisfaction index is the simulation error in step S3. .

[0072] The multi-dimensional satisfaction judgment module comprehensively judges the above indicators. If any indicator is not satisfied, the specific data of the unsatisfied indicator is fed back to the corresponding module: if the production land area or connection method is not satisfied, the area gap data or connection stiffness insufficient data is fed back to the functional device selection module in step S2, driving the selection module to rematch the device. For example, when the area gap is large, a smaller alternative device is selected; if the simulation result is not satisfied, the simulation error distribution data is fed back to the multi-objective optimization planning module in step S3, driving the optimization algorithm to solve again.

[0073] If all indicators are met, the planning scheme is deemed feasible, and the process proceeds to step S5: intelligent splicing and dynamic calibration.

[0074] Meanwhile, the multi-dimensional satisfaction judgment module interacts with the demand modeling module, feeding back data on unapproved expansion and maintenance requirements to the demand modeling module, driving it to requantify the corresponding demand parameters; outputting demand satisfaction data of the planning scheme to the demand modeling module, which then updates the demand priority weights based on this data; and outputting missing maintenance parameter information to the demand modeling module when maintenance requirements are not met, driving it to supplement maintenance parameters and thereby optimize the maintenance monitoring dimensions of the digital twin foundation framework.

[0075] Based on the satisfaction judgment result of step S4, step S5, intelligent splicing and dynamic calibration, is performed. The foundation is automatically spliced ​​according to the connection method through an intelligent positioning device. Dynamic calibration is completed by combining multi-sensor data and synchronized to the digital twin foundation frame.

[0076] This step includes two sub-steps: S51 intelligent stitching path planning and execution, and S52 dynamic calibration and twin synchronization update. The original content clearly states that the A* algorithm should be used to plan the path and the Kalman filter algorithm should be used to achieve accurate calibration.

[0077] In practice, the real-time location data of each base module is first collected by the intelligent positioning device and transmitted to the intelligent splicing module. Based on the connection method determined in step S3, the intelligent splicing module executes the intelligent splicing path planning and execution sub-step S51.

[0078] The A* algorithm is used to process the location data and generate the optimal stitching path for foundation movement. The evaluation function of the A* algorithm is: ,in: This is the path evaluation value. For path nodes, From the starting point to the node The actual movement cost is calculated by measuring the distance from the starting point to the node of the foundation module. The energy consumption is obtained by weighting the time, i.e. , For mobile energy consumption, For the time of travel, The path smoothing coefficient is determined based on the weight of the foundation module; when the weight exceeds 50 tons... When the weight is between 20 and 50 tons When the weight is less than 20 tons , For nodes The path curvature at the point, through the node The coordinates of its immediate and adjacent nodes are calculated. For nodes The estimated cost to the destination is calculated using a modified Euclidean distance, namely: in, This is the distance correction factor, with a value of 1.2.

[0079] After path planning is completed, the foundation module moves along the planned path under the action of the drive device. The adaptive connection device collects docking pressure data in real time. When the docking pressure deviates from the preset threshold, the compensation amount is calculated through the deviation compensation algorithm, and the fine-tuning component of the drive connection device is used to correct the deviation, so as to realize the adaptive compensation of docking deviation.

[0080] After the foundation modules are assembled, the S52 dynamic calibration and twin synchronization update sub-step is executed. This involves fusing multi-sensor data from GPS, laser positioning, and inertial navigation, and using a Kalman filter algorithm to achieve millimeter-level precision calibration of the foundation position.

[0081] Multi-sensor data fusion employs a weighted average method, with weights determined based on sensor accuracy levels. Laser positioning, with the highest accuracy, has a weight of 0.6; GPS positioning is next, with a weight of 0.2; and inertial navigation, with the lowest accuracy, has a weight of 0.2. The fused observations... .

[0082] The state update equation for Kalman filtering is: ,in: The optimal state estimate at time k is the precise location of the foundation. The prior state confidence weights are determined based on the accuracy of the prior estimate, derived from the laser positioning data. Prior estimates derived from other sensor data , Let k be the prior state estimate. For Kalman gain, via calculate, Let be the prior error covariance matrix. To observe the noise covariance matrix, For multi-sensor fusion observations, The observation matrix is ​​determined based on the sensor's observation model.

[0083] The calibrated foundation precise location data The digital twin foundation framework is input in reverse, and the digital twin foundation framework updates the position parameters of the digital twin of the foundation module based on this data, completing the iterative update of the digital twin and ensuring the real-time consistency between the digital twin and the physical entity.

[0084] During the intelligent stitching and dynamic calibration process in step S5, the intelligent stitching module and the digital twin foundation frame maintain a virtual-real linkage interaction: the intelligent stitching module synchronizes the real-time location data of the stitching process to the digital twin foundation frame via the MQTT protocol. Based on this data, the digital twin foundation frame constructs a virtual-real synchronized stitching visualization scene, displaying the movement trajectory and stitching status of each foundation module in real time. When the digital twin foundation frame simulates the stitching process, it predicts potential collision risks through a collision detection algorithm and transmits the coordinates and warning level data of the risk nodes to the intelligent stitching module. The intelligent stitching module adjusts the stitching path based on the warning level. A Level 1 warning with a collision probability >80% immediately stops movement and replans the path. A Level 2 warning with a collision probability of 50%~80% slows down movement and fine-tunes the path. A Level 3 warning with a collision probability <50% maintains the path but increases the monitoring frequency. After the intelligent stitching module completes the stitching, it transmits the physical connection status data, such as connection pressure and connection stiffness, to the digital twin foundation frame for remote verification through the digital twin. The differences between the virtual connection status and the physical connection status are compared to ensure the stitching quality.

[0085] After completing the intelligent stitching and dynamic calibration in step S5, step S6, full lifecycle adaptation verification and operation and maintenance plan generation, is executed. This step utilizes a digital twin foundation framework to complete the full lifecycle adaptation verification and generate operation and maintenance plans. The original content clearly states that this step requires building a full lifecycle adaptation verification system and using fault tree analysis to generate operation and maintenance plans. In practice, the full lifecycle adaptation verification system is first constructed. This system, based on the digital twin foundation framework, simulates typical scenarios throughout the smart factory's lifecycle, including fault scenarios in the operation and maintenance phase, equipment addition scenarios in the expansion phase, and structural adjustment scenarios in the modification phase. Fault scenarios during the operation and maintenance phase include wear and tear of connecting devices, drift in positioning device accuracy, and overload of the foundation. By setting corresponding fault parameters in the digital twin, such as a 30% decrease in connection stiffness, an increase in positioning error to 10mm, and a bearing pressure exceeding the design threshold by 20%, the operational stability of the foundation module is verified. During the expansion phase, the device addition scenario simulates the installation process of a new production device to verify whether the reserved foundation interface and reserved space meet the installation requirements of the new device. During the renovation phase, the structural adjustment scenario simulates the recombination process of the foundation module to verify the reconfiguration flexibility and connection reliability of the foundation module.

[0086] During the adaptation verification process, the full lifecycle adaptation verification module and the intelligent stitching module maintain iterative interaction: if the adaptation verification discovers ground position offset data after long-term operation, the offset data is fed back to the intelligent stitching module, which then optimizes the prior state confidence weights of the Kalman filter algorithm based on this data. This improves calibration accuracy. The maintenance plan generated by the full lifecycle adaptation verification module includes data related to module replacement requirements, such as the module type and replacement time window. This data is transmitted to the intelligent splicing module, which then presets a disassembly and reassembly path. The path planning uses the A* algorithm in step S5 to ensure the efficiency and safety of the disassembly and reassembly process. If the expansion adaptation verification fails or the reserved space is insufficient, the dimensions and installation location data of the new device are fed back to the intelligent splicing module. Based on this data, the intelligent splicing module readjusts the foundation layout and optimizes the combination of foundation modules.

[0087] Meanwhile, based on real-time monitoring data from the digital twin foundation framework, fault tree analysis is used to generate operation and maintenance plans.

[0088] The top event of the fault tree is the failure of the foundation module, the intermediate events include connection failures, positioning failures, load failures, etc., and the bottom events are the specific causes of the failure, such as aging of the connection device, damage of the sensor, overload, etc.

[0089] Fault tree analysis is used to determine the probability of occurrence of each bottom event, calculate the probability of occurrence of the top event, and identify the causes of key failures. Based on the causes of key failures, fault warning thresholds are set, such as setting the connection pressure warning threshold to 80% of the design threshold, the positioning error warning threshold to 5mm, and the bearing pressure warning threshold to 90% of the design threshold. A module replacement process is developed, clarifying the preparatory work before replacement, the operation steps during replacement, and the verification process after replacement. Parameter adjustment schemes are determined, and for different failure types, the adjustment range and methods for positioning device accuracy calibration parameters, connection device pressure adjustment parameters, and foundation bearing buffer parameters are developed.

[0090] If the full lifecycle adaptation verification passes and the operation and maintenance plan is generated, the smart factory foundation planning is considered complete. If the adaptation verification fails, such as the foundation module failing to operate normally in a certain scenario, or the operation and maintenance plan not covering key fault types, the failed verification results and missing fault type data are fed back to the functional device selection module in step S2, driving the selection module to re-match the device, and repeating the process from step S2 to step S6 until the full lifecycle adaptation verification passes and the operation and maintenance plan is complete. Furthermore, the multi-dimensional satisfaction assessment module and the intelligent splicing module maintain precise feedback interaction: if the multi-dimensional satisfaction assessment module detects insufficient connection stability (connection stiffness < design threshold), it feeds this data back to the intelligent splicing module. Based on this data, the intelligent splicing module adjusts the connection pressure parameters and calibration accuracy threshold, increasing the connection pressure by 10%~20% and raising the calibration accuracy threshold from 3mm to 2mm. If the multi-dimensional satisfaction assessment module identifies that the production land area is below standard (e.g., area gap > 10%), it feeds the area gap data back to the intelligent splicing module, driving it to replan the foundation splicing combination method, using staggered splicing or superimposed splicing to expand the usable area. If the multi-dimensional satisfaction assessment module transmits the passed position accuracy data to the intelligent splicing module, the intelligent splicing module uses this data as a standard parameter to solidify the calibration process, and subsequent splicing processes directly use this standard parameter for calibration.

[0091] The functional device selection module and the multi-objective optimization planning module maintain a two-way feedback interaction: the functional device selection module transmits the selected device size, energy consumption, and production capacity parameters to the multi-objective optimization planning module as planning boundary constraints. For example, the device size determines the minimum side length of the foundation module, and the energy consumption parameters determine the minimum target for energy consumption optimization. During the planning process, the multi-objective optimization planning module identifies device compatibility issues, such as a mismatch between the device size and the foundation module size, and feeds back the problem type and compatibility deviation data to the functional device selection module, driving the selection module to re-match devices. After calculating the optimal energy consumption solution, the multi-objective optimization planning module transmits the optimal energy consumption value to the functional device selection module, guiding the selection module to adjust the energy consumption priority weights and increase the selection priority of low-energy-consumption devices.

[0092] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A modular foundation planning method for intelligent factories, characterized in that, Includes the following steps: S1. Based on the full life cycle requirements of the smart factory, complete multi-dimensional requirement modeling, construct a digital twin foundation framework, design regional planning with multiple foundations, and pre-embed intelligent positioning devices and adaptive connection devices in each foundation. S2. Based on the digital twin foundation framework, select basic functions and use AI algorithms to achieve optimal matching between functions and production equipment; S3. Calculate the estimated foundation area and equipment installation distribution diagram based on the selected device, complete the planning layout and connection method design through multi-objective optimization algorithm, and verify the design using the digital twin foundation frame simulation. S4. Determine whether the production land area, connection method and simulation results meet the standards. If they do not meet the standards, feedback is sent to S2. S5. Based on the connection method, the foundation is automatically spliced ​​through an intelligent positioning device, and dynamic calibration is completed by combining multi-sensor data and synchronized to the digital twin foundation frame. S6. Complete the full lifecycle adaptation verification and generate operation and maintenance plans through the digital twin foundation framework. If the standards are not met, feedback is sent to S2. If the standards are met, the planning is completed.

2. The modular foundation planning method for a smart factory as described in claim 1, characterized in that, S1 includes the following sub-steps: S11. A multi-scenario demand weight allocation algorithm is adopted to decompose and prioritize production capacity, ease of operation and maintenance, and future expansion needs, and to construct a dynamic mapping model between demand parameters and foundation physical parameters to achieve accurate conversion of demand parameters to foundation physical parameters. S12. Lightweight digital twin modeling technology is used to construct a digital twin of the foundation module, and a standardized Internet of Things interface is developed to achieve low-latency data transmission and interface adaptation for intelligent positioning devices and adaptive connection devices.

3. The modular foundation planning method for a smart factory as described in claim 1, characterized in that, The optimal matching of functions and production devices in S2 is achieved in the following way: a function and device matching dataset containing function parameters, device parameters and adaptability scores is constructed, a matching model is trained using the random forest algorithm, the selected basic function parameters are input, and the prediction results and confidence weights of multiple decision trees are fused to output a list of production devices with the highest adaptability scores.

4. The modular foundation planning method for a smart factory as described in claim 1, characterized in that, S3 includes the following sub-steps: S31. Based on the dynamic adjustment and optimization of target weights for production load in smart factories, establish a coupled model of foundation area and connection stability to achieve coordinated optimization of production capacity, energy consumption, and logistics efficiency. S32. Introduce a time dimension to construct a dynamic simulation scenario, simulate the load changes and connection reliability of the foundation at different production stages, calculate the simulation error and use corresponding algorithms to correct the planning parameters, thereby improving the planning accuracy.

5. The modular foundation planning method for a smart factory as described in claim 1, characterized in that, In step S4, it is determined whether the production land area, connection method, and simulation results meet the standards. If they do not meet the standards, feedback is provided. This includes the following sub-steps: S41. Feed back the unapproved expansion and maintenance requirements data to the requirement modeling module to drive it to requantify the requirement parameters and supplement the maintenance parameters. S42. Feed back the data on area gap and insufficient connection stability to the functional device selection module to drive it to re-match the device. S43. Feed back the simulation error data to the multi-objective optimization planning module to drive it to solve again.

6. The modular foundation planning method for a smart factory as described in claim 1, characterized in that, Step S5 includes the following sub-steps: S51. A path planning algorithm that balances cost and smoothness is used to process the location data collected by the intelligent positioning device to generate the optimal splicing path. The adaptive connection device collects docking pressure data in real time and achieves adaptive compensation for docking deviation through a deviation compensation algorithm to ensure high-precision installation standards of the foundation. S52 integrates multi-sensor data from GPS, laser positioning, and inertial navigation, and uses a precise calibration algorithm to achieve high-precision calibration of the foundation position. The calibration data is then input back into the digital twin foundation framework to complete the iterative update of the digital twin.

7. The modular foundation planning method for a smart factory as described in claim 1, characterized in that, In step S6, the full lifecycle adaptation verification includes: completing the verification by simulating operation and maintenance failure scenarios, expansion device addition scenarios, and structural adjustment scenarios; the operation and maintenance plan is generated by fault tree analysis method, including fault warning thresholds, module replacement process and parameter adjustment scheme, and determining the key fault causes and countermeasures based on real-time monitoring data of the digital twin foundation framework.

8. The modular foundation planning method for a smart factory as described in claim 1, characterized in that, The digital twin foundation framework and the multi-objective optimization planning module include real-time interaction: the digital twin foundation framework synchronizes the foundation positioning accuracy and connection interface status data to the multi-objective optimization planning module, providing real data input for the optimization algorithm; The planning scheme output by the multi-objective optimization planning module is transmitted to the digital twin foundation framework for simulation verification. After simulation, parameter correction instructions and interference problem data are generated and transmitted back to drive the optimization algorithm iteration and replanning of the splicing path.

9. The modular foundation planning method for a smart factory as described in claim 1, characterized in that, The functional device selection module and the multi-objective optimization planning module include bidirectional feedback: the functional device selection module transmits device size, energy consumption, and production capacity parameters to the multi-objective optimization planning module as planning boundary constraints; after identifying device adaptability issues, the multi-objective optimization planning module feeds back the problem type and adaptability deviation data to the functional device selection module, driving it to rematch devices, and at the same time transmits the optimal energy consumption solution to the functional device selection module to guide it to adjust energy consumption priority weights.

10. The modular foundation planning method for a smart factory as described in claim 1, characterized in that, The multi-dimensional satisfaction judgment module and the intelligent splicing module include feedback interaction: the detection data of insufficient connection stability and production land area not meeting the standard are fed back to the intelligent splicing module, driving it to adjust the connection pressure parameters, calibrate the accuracy threshold, and re-plan the foundation splicing combination method; the passed position accuracy data is transmitted to the intelligent splicing module as standard parameters to solidify the calibration process.

Images