Artificial intelligence-based brick masonry reinforcement construction progress optimization and management system

Abstract

The present application belongs to the technical field of project management, and relates to a brick masonry reinforcement construction progress optimization and management system based on artificial intelligence, comprising: an information constraint graph construction module, which acquires data to generate an information-physical constraint graph containing process control handles; a monitoring data flow establishment module, which cleans state signals to obtain monitoring data flows mapped to nodes; a delay risk quantification module, which compares target and data flow to obtain a time delay probability vector; a control strategy generation module, which calls handles to generate entity control strategy objects; a regulation and control instruction issuing module, which extracts threshold values to generate and issue targeted regulation and control instructions; an intervention feedback monitoring module, which combines feedback and boundary simulation to obtain a deterministic intervention completion time; and a construction sequence optimization module, which updates time parameters and deduces a globally optimized construction organization sequence. The present application solves the problems of site physical disturbance response lag, difficulty in forming active intervention, and large time deviation.

Description

Technical Field

[0001] This invention belongs to the technical field of project management and relates to an artificial intelligence-based optimization and management system for the construction progress of brick masonry reinforcement. Background Technology

[0002] Masonry structure reinforcement projects are widely used in existing building renovation and post-disaster structural reconstruction. These projects involve coordination of multiple processes and disciplines, and the construction process is highly sensitive to physical environmental conditions such as temperature and humidity. Currently, the main challenge in reinforcement project management lies in dealing with complex and ever-changing on-site physical environment disturbances, which can easily introduce schedule uncertainties into critical processes such as chemical grouting and mortar curing. Traditional schedule management models struggle to establish a closed-loop system for perception and control at the microscopic physical level, easily increasing the risk of schedule delays and cost overruns.

[0003] Currently, commonly used schedule control techniques in the industry mainly rely on static scheduling based on work breakdown structures (WBS) and the critical path method. Managers break down projects into independent tasks, use project management software to determine logical dependencies and resource constraints between tasks to generate a baseline plan, and manually track actual progress through manual inspections. If deviations are found, subsequent schedules are manually adjusted based on experience. In addition, some more advanced technologies have introduced risk prediction models. For example, Chinese invention patent CN116384756A proposes a deep learning-based method for predicting and evaluating the schedule risk of construction projects. This method collects historical data and uses machine learning algorithms to probabilistically assess the potential delay risks of specific tasks, thereby providing managers with decision support information for adjusting scheduling accordingly.

[0004] However, the aforementioned existing technologies have certain limitations in dealing with dynamic physical disturbances at real construction sites. On the one hand, the progress tracking mechanism relying on manual reporting suffers from information lag and struggles to capture sudden changes in micro-environmental parameters affecting the material curing rate in real time. Conventional shift scheduling adjustments are limited to passive responses at the information level and are not conducive to intervening at the physical level to address the root causes of delays. On the other hand, even the aforementioned comparative document technology incorporating deep learning and risk prediction primarily outputs probabilistic warnings, lacking a direct linkage mechanism with on-site physical execution equipment. It is difficult to automatically translate risk predictions into specific physical intervention actions, resulting in a certain technological gap between perceived information and actual control execution. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides an artificial intelligence-based system for optimizing and managing the construction progress of brick masonry reinforcement.

[0006] An AI-based brick masonry reinforcement construction progress optimization and management system includes: The information constraint graph construction module is used to acquire the decomposed structure data of the reinforcement project, extract independent task nodes and resource dependencies, and generate an information physical constraint graph containing process control handles. The monitoring data stream establishment module is used to acquire multi-source heterogeneous status signals collected by field sensing devices, perform time alignment and data cleaning, and obtain monitoring data streams mapped to independent task nodes. The delay risk quantification module is used to obtain the project schedule control target, compare the monitoring data stream with it, and obtain the schedule delay probability vector for the affected nodes. The control strategy generation module is used to obtain the project delay probability vector, call the built-in process control handle of the affected node, and generate an entity control strategy object. The control command issuance module is used to obtain entity control strategy objects, extract electrical control target thresholds, generate targeted control commands, and issue them to field execution nodes; The intervention feedback monitoring module is used to acquire the physical feedback characteristic signals of the field execution nodes based on the targeted control instructions, and to perform simulations and deductions in combination with environmental boundary conditions to obtain the deterministic intervention completion time. The construction sequence optimization module is used to obtain the deterministic intervention completion time, update the time node record parameters of the cyber-physical constraint diagram, perform local collaborative deduction, and obtain the globally optimized construction organization sequence.

[0007] A further aspect of the present invention generates a cyber-physical constraint graph containing process control handles, comprising the following steps: Extract reinforcement construction operations from the decomposed structural data of the reinforcement project, map them as independent task nodes in the topology graph, and establish an initial logical network structure for the project; Traverse all independent task nodes in the initial engineering logic network structure, embed industrial communication protocol interfaces into each independent task node, bind the hardware access identification code of the field physical control actuator, and generate process control handles. By integrating the initial engineering logic network structure with the control handles of each process, the logical dependency edges of start time between each independent task node are calculated, resource mutual exclusion constraint edges of shared construction equipment are established, and a cyber-physical constraint graph is generated.

[0008] A further aspect of this invention involves obtaining a monitoring data stream mapped to an independent task node, comprising the following steps: Temperature and humidity signal streams reflecting local microclimate changes are collected by temperature and humidity sensors installed in the brick masonry. The thermal image matrix of structural defects, which reflects internal defects or lesions of the wall, is obtained by infrared thermal imaging equipment. The temperature and humidity signal stream and the thermal image matrix of structural defects constitute a multi-source heterogeneous state signal. The network time synchronization protocol is used to align the timestamps of the temperature and humidity signal stream with the thermal map matrix of structural defects, and then perform filtering and extreme value removal operations to output the monitoring data stream.

[0009] A further aspect of this invention involves obtaining a project delay probability vector for affected nodes, comprising the following steps: Extract continuous time signal data segments from the monitoring data stream and input them into the trained spatiotemporal convolutional neural network model to extract the weight features of time delay influence; By importing the weighting characteristics of time lag into the Markov Monte Carlo statistical prediction model, the mean expected delay and the probability of extreme delay risk are calculated. The expected mean delay is correlated with the probability of extreme delay risk to the affected nodes in the cyber-physical constraint graph to generate a project delay probability vector.

[0010] A further aspect of the present invention generates an entity control strategy object, comprising the following steps: When the probability vector of project delay exceeds the time alarm threshold, the scheduling progress is suspended and the physical environment compensation strategy is activated. Read the process control handle built into the affected node, verify the type of hardware device mapped by the hardware access identifier code, and when it is confirmed to be an infrared heating device, parse and extract its highest output power parameter. Based on the maximum output power parameter, an inverse physical control equation is established, the compensated heating temperature rise gradient curve is calculated, and the compensated heating temperature rise gradient curve and the hardware execution timestamp are encapsulated into an entity control strategy object.

[0011] A further aspect of this invention involves generating targeted control commands and issuing them to field execution nodes, including the following steps: Intercept routine task delay scheduling signaling to verify the protocol compatibility between the entity control strategy object and the process control handle communication interface; If the verification passes, the communication driver runtime library mapped to the hardware access identifier code is loaded via the hardware bus. Extract the electrical control target threshold from the entity control strategy object, generate targeted control instructions, and send them to the field execution node.

[0012] A further aspect of the present invention, obtaining the deterministic intervention completion time, includes the following steps: The thermal radiation energy conversion output efficiency is obtained by reading the resistance temperature feedback signal generated by the field execution node through a high-frequency polling mechanism. The thermal radiation energy conversion output efficiency is input into the composite material condensation kinetics simulation module for prediction. Accumulate the actual time spent on physical interventions to calculate and generate the deterministic intervention completion time.

[0013] A further aspect of this invention involves obtaining a globally optimized construction organization sequence, comprising the following steps: In the time attribute block of the cyber-physical constraint diagram, the delayed delivery benchmark is updated using deterministic intervention completion time; Extract the associated data from the updated cyber-physical constraint graph, perform a forward traversal and trace downstream associated task nodes along the logical dependency edge of the start time, and filter the affected associated process constraint links. The Markov decision reinforcement optimization agent unit is activated to perform deduction on the constraint link of the related process and generate a globally optimized construction organization sequence that eliminates equipment resource conflicts.

[0014] A further aspect of the present invention involves calculating the compensated heating temperature rise gradient curve, including the following steps: The volume parameters, density parameters, and specific heat capacity parameters of the material to be heated are obtained. Combined with the delay time that needs to be shortened, the total compensation energy required to offset the delay is calculated based on the preset nonlinear function of material properties. Obtain the original planned construction period and set it as the target construction period for compensation intervention; Using the total compensation energy as the integration target value, the compensation intervention target period as the integration time limit, and the highest output power parameter as the instantaneous power constraint, the instantaneous power function is solved in reverse to obtain the compensation heating temperature rise gradient curve.

[0015] A further aspect of this invention involves calculating and generating the completion time of a deterministic intervention, including the following steps: The emissivity, total radiative surface area, and Stefan-Boltzmann constant of the material on the surface of the heating equipment are obtained. The absolute temperature of the surface of the heating equipment is obtained based on the resistance temperature feedback signal, and the absolute temperature of the ambient environment is also obtained. The difference between the fourth power of the absolute temperature of the heating equipment surface and the fourth power of the absolute temperature of the ambient environment is calculated. The difference is then multiplied by the material emissivity, total radiative surface area, and Stefan-Boltzmann constant to obtain the actual net radiative power. Divide the actual net radiation power by the input electrical power set by the targeted control command to obtain the thermal radiation energy conversion output efficiency. Obtain the estimated remaining completion time output by the composite material condensation kinetics simulation module, add the actual time spent to the estimated remaining completion time to obtain the deterministic intervention completion time.

[0016] In summary, the present invention has the following beneficial technical effects: 1. By constructing a cyber-physical constraint graph containing process control handles, abstract project management task nodes are mapped and bound to the underlying hardware access identifiers of specific physical control actuators on site. This feature transforms the traditional project logic network into a data system with physical layer intervention capabilities. When the upper-level scheduling system needs to intervene, it is not limited to adjusting the virtual timetable but can directly address and invoke the corresponding physical handles, establishing a direct data path from scheduling decisions to physical execution. This setup enables the project schedule control system to proactively intervene in sudden physical environmental disturbances on site, constructing a closed-loop mechanism of perception, decision-making, and execution, which helps improve the project management's responsiveness to on-site physical disturbances.

[0017] 2. By collecting multi-source heterogeneous state signals from the site and utilizing a spatiotemporal convolutional neural network model combined with Markov Monte Carlo statistical prediction, a quantified probability vector of project delay is generated. This method dynamically models the nonlinear impact of continuously changing physical environmental parameters on construction activities, extracts the spatiotemporal characteristics of environmental disturbances using a spatiotemporal convolutional network, and transforms them into a probability distribution prediction of the project schedule using Monte Carlo simulation, outputting characteristic values ​​such as expected delay and extreme risk probability. This approach changes the traditional qualitative judgment method that relies on human experience in management, transforming the risk of project schedule deviation into a dynamically quantified data indicator. This provides an objective input basis for subsequent intervention decisions, helps improve the accuracy and timeliness of risk identification, and reduces derivative project schedule losses caused by subjective and delayed judgment.

[0018] 3. By triggering feedforward correction based on the probability vector of schedule delays, an entity control strategy object is generated and targeted control instructions are issued. Simultaneously, high-frequency polling of physical feedback signals is used to monitor the execution status. This mechanism establishes an automated control process including risk quantification, inverse solution of control strategies, instruction issuance, physical execution, and closed-loop feedback. When a schedule deviation risk is detected, the system actively calculates the required compensation energy through inverse physical control equations, generating underlying hardware instructions. Based on this, the system corrects the internal physical simulation model in real time by collecting real physical feedback. By introducing actual physical feedback for closed-loop correction, this method reduces the random volatility of relying solely on probabilistic model predictions, obtains intervention completion times with engineering scheduling guidance significance, and thus improves the matching between the project plan and the actual on-site construction progress. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention.

[0020] Figure 1 This is a schematic diagram of the framework in the embodiments of this application.

[0021] Figure 2 This is a flowchart illustrating an embodiment of this application.

[0022] Figure 3 This is a graph showing the probability density distribution of project delays output by the Markov Monte Carlo statistical prediction model in this application.

[0023] Figure 4 This is a compensated heating temperature rise gradient curve generated by the inverse physical control equation in the embodiments of this application.

[0024] Figure 5 This is a characteristic curve of the actual net radiant heat power changing with the absolute surface temperature of the infrared heating device in the embodiments of this application. Detailed Implementation

[0025] The following is in conjunction with the appendix Figures 1-5 A preferred description of the present invention is provided below.

[0026] See attached document Figures 1-5 This invention proposes an artificial intelligence-based construction progress optimization and management system for brick masonry reinforcement, which includes the following modules: The information constraint graph construction module is used to acquire the decomposed structure data of the reinforcement project, extract independent task nodes and resource dependencies, and generate an information physical constraint graph containing process control handles. The monitoring data stream establishment module is used to acquire multi-source heterogeneous status signals collected by field sensing devices, perform time alignment and data cleaning, and obtain monitoring data streams mapped to independent task nodes. The delay risk quantification module is used to obtain the project schedule control target, compare the monitoring data stream with it, and obtain the schedule delay probability vector for the affected nodes. The control strategy generation module is used to obtain the project delay probability vector, call the built-in process control handle of the affected node, and generate an entity control strategy object. The control command issuance module is used to obtain entity control strategy objects, extract electrical control target thresholds, generate targeted control commands, and issue them to field execution nodes; The intervention feedback monitoring module is used to acquire the physical feedback characteristic signals of the field execution nodes based on the targeted control instructions, and to perform simulations and deductions in combination with environmental boundary conditions to obtain the deterministic intervention completion time. The construction sequence optimization module is used to obtain the deterministic intervention completion time, update the time node record parameters of the cyber-physical constraint diagram, perform local collaborative deduction, and obtain the globally optimized construction organization sequence.

[0027] In one embodiment of the present invention, the information constraint graph construction module is configured to perform the following steps: Extract reinforcement construction operations from the decomposed structural data of the reinforcement project, map them as independent task nodes in the topology graph, and establish an initial logical network structure for the project; Traverse all independent task nodes in the initial engineering logic network structure, embed industrial communication protocol interfaces into each independent task node, bind the hardware access identification code of the field physical control actuator, and generate process control handles. By integrating the initial engineering logic network structure with the control handles of each process, the logical dependency edges of start time between each independent task node are calculated, resource mutual exclusion constraint edges of shared construction equipment are established, and a cyber-physical constraint graph is generated.

[0028] Specifically, the process of acquiring the decomposed structural data of the reinforcement project, constructing an cyber-physical constraint diagram containing process control handles, and generating a progress baseline control network with physical layer intervention interfaces is executed by the central data processing server.

[0029] First, the central data processing server receives and parses a pre-formatted breakdown structure of reinforcement project data. This data, typically in the form of a spreadsheet or a file exported from project management software, clearly defines all construction activities for the brick masonry reinforcement project. The server iterates through this data list, instantiating each specific reinforcement work—such as removing the old plaster layer, drilling and cleaning holes in the wall, installing chemical anchors, installing and sealing formwork, and pressure grouting—in memory as an independent task node with a unique task identifier, mapping it to an independent task node in the hybrid logic topology graph. By aggregating all the generated independent task nodes, an initial engineering logic network structure is constructed, containing only nodes and without connecting edges.

[0030] The server iteratively processes each independent task node in the initial engineering logical mesh structure, injecting it with the ability to perceive and control the physical world. For each independent task node identified as requiring direct participation from physical equipment, such as a pressure grouting node, the server queries a pre-configured hardware resource mapping table, which records the correspondence between construction tasks and on-site physical control actuators.

[0031] Based on this mapping relationship, the server embeds a standardized industrial communication protocol interface definition for the node, such as defining its communication mode as a message queue-based telemetry transmission protocol, and extracts and binds a unique underlying hardware access identifier from the mapping table, such as an address string pointing to a specific programmable logic controller input / output port. These two together form a structured process control handle and are stored in the data structure of the independent task node.

[0032] After embedding the handles of all relevant nodes, the server performs a final graph assembly and full calculation of constraint relationships, merging the initial engineering logic network structure with all process control handles. The server then parses the task prerequisite relationships defined in the reinforcement engineering decomposition structure data again. For example, wall drilling and hole cleaning must be completed before the implantation of chemical anchors. Based on this, directed logical dependencies on start time are established between the corresponding independent task nodes.

[0033] At the same time, the server will analyze the resource requirements of all tasks. When it is identified that two or more tasks, such as grouting the east wall and grouting the west wall, need to share the same large grouting pump, and this equipment is listed in the list of key construction machinery and equipment, the server will establish undirected resource mutual exclusion constraint edges between these task nodes to characterize their non-parallelism in time.

[0034] The complete graph structure, which integrates all independent task nodes, logical dependencies on start time, and mutually exclusive resource occupancy constraints, is formally defined and generated as a cyber-physical constraint graph. This graph structure data is serialized and stored in the graph database, serving as the benchmark data model for all subsequent progress projections and dynamic control.

[0035] It should be noted that the decomposed structure data of the reinforcement project is a structured data table, which at least includes fields such as task number, task name, preceding task number, estimated construction period, and required key construction machinery and equipment, providing all the original input information for constructing the cyber-physical constraint diagram.

[0036] An independent task node is the basic unit in a graph data structure. Its data attributes include at least a task number, a task name, and a reserved field for storing a process control handle. The process control handle internally encapsulates two core pieces of information: The communication protocol type is used to specify the protocol standard used when interacting with physical devices, such as message queue telemetry transmission or open platform communication unified architecture; The underlying hardware access identifier is a string or address that ensures unique addressing of a specific physical actuator, such as the device serial number of a network relay or its station address on an industrial bus. Its setting is based on the network topology and device addressing rules of the field automation system.

[0037] The logical dependency edges for commencement time are directed edges in the graph, representing the mandatory sequence in the construction process. They are established based on the process constraints explicitly stated in national building construction codes or project construction organization designs. The mutual exclusion constraint edges for resources are undirected edges in the graph, representing the exclusivity of construction due to the sharing of indivisible critical resources. They are established based on the conflict detection results of the resource requirement field in the decomposed structure data of the reinforcement project.

[0038] Cyber-physical constraint graphs are comprehensive graph data structures generated through the above steps. They serve as the core data hub connecting virtual project plans with the physical construction site.

[0039] In one example of this embodiment, the structural breakdown data for the reinforcement project is a table containing five main tasks, which are as follows: Wall surface cleaning; Drill holes and insert reinforcing bars; Install grouting templates; Pressure grouting operation; : Curing and demolding. Their relationship is as follows: Regarding resource requirements, the task... and All of them need to share the same large pressure grouting machine, model PUMP-1000. The central data processing server first generates five independent task nodes based on this table. to .

[0040] Subsequently, the server traverses the nodes, and when it processes a node... At that time, a query of the hardware resource mapping table revealed that the physical actuator corresponding to the pressure grouting operation was an intelligent grouting pump unit, whose underlying hardware access identifier was the hardware access identifier of the target device, such as a network node address or device serial number. Therefore, it was determined that... A process control handle was generated and bound, containing a set of control attributes including the communication protocol and corresponding identifier codes. Based on the task logic, the server establishes four directed logical dependency edges for start time between nodes: Furthermore, the server identifies nodes through resource requirement analysis. and Since a large pressure grouting machine is shared, an undirected resource mutual exclusion constraint edge is established between these two nodes. The final generated cyber-physical constraint graph contains five nodes, four logical dependency edges, and one resource mutual exclusion edge, and is as follows: Key nodes have embedded process control handles pointing to physical devices, and the graph structure is stored in its entirety for use in subsequent steps.

[0041] In one embodiment of the present invention, the monitoring data stream establishment module is used to perform the following steps: Temperature and humidity signal streams reflecting local microclimate changes are collected by temperature and humidity sensors installed in the brick masonry. The thermal image matrix of structural defects, which reflects internal defects or lesions of the wall, is obtained by infrared thermal imaging equipment. The temperature and humidity signal stream and the thermal image matrix of structural defects constitute a multi-source heterogeneous state signal. The network time synchronization protocol is used to align the timestamps of the temperature and humidity signal stream with the thermal map matrix of structural defects, and then perform filtering and extreme value removal operations to output the monitoring data stream.

[0042] To activate the distributed sensor array deployed at the reinforcement construction site, collect multi-source heterogeneous status signals containing high-precision timestamps, and establish a monitoring data stream reflecting changes in the construction operation environment, the central data processing server will schedule and control the on-site sensing hardware based on the task execution status in the cyber-physical constraint diagram.

[0043] First, when the construction progress in the cyber-physical constraint diagram advances to or is about to enter an environmentally sensitive operational phase, such as pressure grouting, curing, or formwork removal, the central data processing server sends a start command to the temperature and humidity digital sensor array hardware pre-embedded and connected to designated measuring points on the brick masonry wall to be reinforced through its connected IoT gateway, thus turning on its power supply.

[0044] Once activated, these sensors continuously collect local surface microclimate parameters at a preset sampling frequency and send back structured data packets containing timestamps, temperature readings, and humidity readings via wireless protocols such as Bluetooth Low Energy or RFID technology, thus forming a continuously fluctuating temperature and humidity signal stream.

[0045] Simultaneously, the server sends scanning commands to an infrared thermal imaging scanner hardware terminal mounted on a tripod that can fully cover the critical construction work surface. Once activated, the terminal performs a complete infrared thermal radiation scan of the wall surface at set intervals, generating a two-dimensional thermal map of M×N pixels. The value of each pixel in the map represents the surface temperature of a corresponding tiny area of ​​the wall. By analyzing abnormally low or high temperature areas, hollow defects deep within the wall can be identified, manifesting as localized low-temperature zones caused by increased thermal resistance, or hidden defects caused by water seepage, manifesting as a significant low-temperature contour caused by the heat absorption of moisture evaporation. This two-dimensional thermal map data constitutes the structural defect characteristic thermal map matrix.

[0046] To integrate these two heterogeneous data streams, the server performs time alignment and data cleaning operations. It utilizes the master clock service of the time protocol deployed within the local area network of the construction site to forcibly align the system time reference stamps recorded during the acquisition of the temperature and humidity signal stream and the structural defect feature heat map matrix, ensuring that the time reference of all data points is consistent.

[0047] The server applies a moving average digital filter to smooth the high-frequency sampled temperature and humidity signal stream to eliminate glitches introduced by transient electromagnetic interference, and employs... The criteria identify and eliminate isolated outliers caused by accidental sensor failures. After the above processing, clean and time-synchronized data is integrated and packaged, and associated with the specific independent task node identifiers in the cyber-physical constraint diagram it monitors. The final output is a monitoring data stream directly mapped to the underlying task nodes.

[0048] A temperature and humidity digital sensor array is a set of electronic devices containing integrated temperature and humidity sensing chips installed on the surface of the wall to be reinforced according to a predetermined grid, for example, one per square meter. Its temperature measurement accuracy is usually required to be no less than ±0.5℃ and humidity measurement accuracy to be no less than ±3%RH, so as to accurately capture the minute environmental changes that affect the material curing process.

[0049] The structural defect feature heat map matrix is ​​a two-dimensional array with dimensions consistent with the resolution of the scanner's infrared detector, such as 320×240. Each element in the array is a floating-point number representing the absolute temperature value of the corresponding physical location, in degrees Celsius.

[0050] Network standard time synchronization protocol refers to a time protocol that can synchronize the clocks of all devices in the entire field network to the sub-millisecond level by exchanging timestamp information between network devices. This is a technical prerequisite for ensuring that multi-source heterogeneous data can be compared in the time dimension. The implementation of this protocol assumes that network switches and terminal devices that support the protocol have been deployed in the field.

[0051] The monitoring data stream directly mapped to the underlying task nodes is a structured collection of data, in which each data item is accompanied by a source identifier that can be parsed and uniquely associated with an independent task node in the cyber-physical constraint diagram, ensuring an explicit correspondence between the data and the construction activity it describes.

[0052] In one example of this embodiment, when the project progress is about to enter the task... During pressure grouting, the central data processing server activates the sensors deployed for the task area. First, the server sends a command to the numbered temperature and humidity digital sensor array on the wall, which begins reporting data at a frequency of 1Hz. The server receives the raw temperature and humidity signal stream, a segment of which may contain noisy data. At a certain time t1, normal temperature and humidity sampling data is received, such as a temperature of 15.2℃ and humidity of 65.0%. Subsequently, at the adjacent time t2, abnormal extreme values ​​of data due to environmental noise interference are received, such as a sudden change in humidity reading to 99.9%. Subsequent data returns to normal. Simultaneously, the server instructs an infrared thermal imaging scanner to scan at the corresponding synchronization time, acquiring a 160×120 thermal map matrix of structural defect features. Analysis reveals a 5×5 pixel area near matrix coordinates [80,60] with an average temperature of 12.5℃, significantly lower than the surrounding area's average temperature of 15.0℃. This is marked as a potential water seepage hazard. Next, the server processes the data and confirms that all timestamps ts are synchronized via a time protocol. During the cleaning phase, the server applies... The criteria analysis of the humidity data series identified and removed readings of 99.9 as isolated outliers. Finally, the server packaged the cleaned temperature and humidity time series, along with a structural defect feature heatmap matrix containing information on potential defect locations and extents, and labeled its associated tasks. This was assigned a mission. The tagged, processed data sets constitute a monitoring data stream that is directly mapped to the underlying task nodes.

[0053] In one embodiment of the present invention, the delay risk quantification module is used to perform the following steps: Extract continuous time signal data segments from the monitoring data stream and input them into the trained spatiotemporal convolutional neural network model to extract the weight features of time delay influence; By importing the weighting characteristics of time lag into the Markov Monte Carlo statistical prediction model, the mean expected delay and the probability of extreme delay risk are calculated. The expected mean delay is correlated with the probability of extreme delay risk to the affected nodes in the cyber-physical constraint graph to generate a project delay probability vector.

[0054] The server executes a risk quantification process based on deep learning and statistical simulation, compares the monitoring data stream with the optimal expected project schedule control target, identifies and evaluates deterministic on-site physical disturbance factors that can induce schedule deviation losses, and outputs a schedule delay probability vector for a specific local construction task.

[0055] The server continuously receives monitoring data streams that are directly mapped to the underlying task nodes. For each monitored individual task node, it segments and extracts fixed-length, continuously sliding time signal data segments containing historical and current data at fixed time steps, such as data from the past 60 minutes. This data segment includes all temperature and humidity signal streams within that time period, as well as the latest structural defect feature heatmap matrix.

[0056] After the server preprocesses this data fragment, it inputs it into a spatiotemporal convolutional neural network model that has been pre-trained in the cloud based on a large amount of historical data from similar projects. The input to this model is a three-dimensional tensor. Where T is the time step, H and W are the spatial resolution of the heatmap, and C is the number of feature channels, including temperature, humidity, and defect area markers.

[0057] This model, through its internal three-dimensional convolutional layer structure, specifically including three cascaded 3D convolutional layers, two three-dimensional max pooling layers, and a global average pooling layer, can simultaneously capture the changing trends of data in the time dimension, such as the continuous decrease in temperature, as well as the distribution characteristics in the spatial dimension, such as the range of cold spots or humid areas.

[0058] During the model training phase, the label of historical samples is defined as the ratio of the actual completion time of the corresponding process to the standard completion time. The model uses mean squared error as the loss function for backpropagation gradient descent training. This allows the model to learn or extract the nonlinear influence of environmental factors on the physicochemical reaction rate of materials. The direct output of the model inference is a numerical vector, defined as the time-lag influence weight feature, which quantifies the deterioration rate of the current physical environment relative to the ideal curing environment.

[0059] The server imports and passes the extracted time-delay influence weight features into the Markov Monte Carlo statistical prediction model. This model is a probabilistic model simulating the transition process of grouting materials from a liquid to a solid state, and the time-delay influence weight features are used as key parameters to adjust the state transition probabilities in the model.

[0060] The server executes tens of thousands of Monte Carlo simulations, each generating possible project completion times based on the transition probabilities affected by disturbances. Through statistical analysis of the results of these tens of thousands of simulations, the server calculates the expected mean delay, i.e., the average time that all simulated project durations exceed the planned duration, and the probability of extreme delays, i.e., the frequency with which the project duration exceeds a certain severe risk threshold.

[0061] The server tracks and matches the calculated average expected delay with the probability of extreme delay risk, encapsulates these two values ​​as a two-dimensional vector, and associates them with the independent task node mapped by the monitoring data stream that generated the data source, i.e. the target affected node in the cyber-physical constraint diagram. This quantifies and ultimately generates a delay probability vector that reveals the degree of damage caused by the delay risk in the target operation.

[0062] The probability vector of project delay can be represented by the following form: ; In the formula, This is a probability vector for project delays, used to encapsulate a quantitative assessment of the project delay risk for a specific task node. This represents the expected mean delay. This value is calculated by averaging all possible project durations simulated using the Markov Monte Carlo method, i.e., the average completion time. Subtract the standard planned duration of the task under ideal conditions. The calculations are based on the statistical central tendency of a large number of random simulations, reflecting the most likely delay duration under the current disturbance.

[0063] This represents the probability of extreme delays. Among them, It is a single project duration sample generated by Markov Monte Carlo simulation. This is a preset extreme delay time threshold. The probability is calculated by taking all simulated samples that meet this threshold. The proportion of the sample size to the total number of simulated samples is used to determine the sample size. The threshold is typically set based on the latest start time of key milestones in the project contract or subsequent critical path tasks. Exceeding this threshold will trigger a chain reaction of delays or default risks. For example, based on project management experience, it may be set at 150% of the standard duration.

[0064] The spatiotemporal convolutional neural network model is a deep learning model designed to process data with spatial and temporal dimensions, such as video streams or time series sensor data in this solution. By training on the relationship between environmental parameter sequences in historical data and the final actual project duration, this model learns how to extract effective predictive features from new environmental data.

[0065] The time lag effect weight feature is one or more floating-point numbers output by the model. The value range is normalized, for example, between 0.5 and 1.5, and is used to quantify the acceleration or deceleration effect of the current environment on the standard curing rate. For example, 0.8 means that the curing rate drops to 80% of the normal rate.

[0066] Markov Monte Carlo statistical prediction models are computational methods based on random sampling. They estimate the probabilistic characteristics of complex systems by constructing Markov chains that describe the evolution of system states and conducting extensive simulations. Here, they are used to predict the distribution of final completion time under the influence of uncertain environments.

[0067] The project delay probability vector condenses complex risk assessment results into standardized data objects containing expected values ​​and extreme risk probabilities, which are directly associated with nodes in the cyber-physical constraint graph.

[0068] In one example of this embodiment, the central data processing server continuously analyzes and performs tasks. The monitoring data stream associated with pressure grouting operations is directly mapped to the underlying task nodes. The server extracts data from the most recent hour as a continuous sliding time signal data segment, reflecting an average wall temperature maintained at 15.2℃, with localized low-temperature zones at 12.5℃. This data segment is fed into a spatiotemporal convolutional neural network model. The model infers the adverse effects of the current environment on the curing of the grouting material based on the input data and outputs a time-lag impact weight feature of 0.80. This value means that the model's evaluation result is that the current environment will cause the curing rate to decrease to 80% of that under ideal conditions. Subsequently, the server feeds this weight 0.80 into a Markov Monte Carlo statistical prediction model. Assuming the task... The standard planned duration is 24 hours, and the extreme delay time threshold is... The time was set to 30 hours. The server ran 50,000 simulations, and the results showed an average completion time of 29.5 hours. Therefore, the expected average delay is... It was calculated to be 5.5 hours. In 50,000 simulations, 9,500 showed completion times exceeding 30 hours. Therefore, the probability of extreme delays... It was calculated to be 0.19. The completion time distribution and various threshold indicators predicted by the Markov Monte Carlo statistical model can be found in [reference needed]. Figure 3 As shown. Ultimately, the server encapsulates these two results into a project delay probability vector. This vector is then used as a dynamic risk attribute to update the cyber-physical constraint graph and the task. corresponding nodes superior.

[0069] In one embodiment of the present invention, the control strategy generation module is configured to perform the following steps: When the probability vector of project delay exceeds the time alarm threshold, the scheduling progress is suspended and the physical environment compensation strategy is activated. Read the process control handle built into the affected node, verify the type of hardware device mapped by the hardware access identifier code, and when it is confirmed to be an infrared heating device, parse and extract its highest output power parameter. Based on the maximum output power parameter, an inverse physical control equation is established, the compensated heating temperature rise gradient curve is calculated, and the compensated heating temperature rise gradient curve and the hardware execution timestamp are encapsulated into an entity control strategy object.

[0070] To trigger the feedforward correction intervention decision-making mechanism of the underlying equipment based on the probability vector of project delay, the process control handle built into the target affected node is called to generate an entity control strategy object that actively intervenes in the physical operation environment process. The central data processing server will execute a closed-loop intervention decision and control instruction generation process.

[0071] The central data processing server continuously monitors the project delay probability vector associated with each activity node in the cyber-physical constraint diagram and compares it in real time with the preset system time buffer safety tolerance alarm limit.

[0072] Once the server detects that an element in the project delay probability vector of any node, such as the expected average delay, exceeds the threshold set by the alarm limit, it immediately triggers a corrective intervention decision mechanism. This mechanism first sends an interrupt signal to the application software layer's regular virtual scheduling progress rescheduling algorithm, suspending and freezing any subsequent scheduling operations on the affected node. Then, it activates and loads the physical environment optimization and compensation response strategy algorithm library located at the interface layer of the underlying hardware devices on-site.

[0073] The server matches and reads the process control handles bound inside the target affected node, and sends a status query request to the physical device represented by the corresponding underlying hardware access identifier code according to the device type defined in the handle, such as an environmental temperature control unit.

[0074] The server receives and verifies the confirmation information returned by the device, confirms that the actual control device hardware type is an external infrared heating and curing flange device, and parses and extracts the highest output power parameter contained in its device metadata.

[0075] Based on the device's capability boundaries and the expected delay time to be offset, the server establishes an inverse physical control equation. This equation is based on a material solidification kinetics model, takes the planned solidification completion time as input, and solves in reverse the energy input curve required to achieve the goal.

[0076] By solving this equation, the server calculates a compensating heating temperature rise gradient curve that can compensate for and shorten the expected lag time gap. Finally, the server encapsulates the calculated compensating heating temperature rise gradient curve with the execution trigger hardware timestamp obtained from the time protocol server, assembles them together into a structured entity control strategy object, and places it into the instruction queue to be sent.

[0077] To calculate the compensated heating temperature rise gradient curve, the core of the inverse physics governing equations is to establish a simplified energy balance model. The total energy to be compensated... With the delay time that needs to be shortened There exists a functional relationship determined by the material properties. This energy must be available within the new target timeframe. The internal heating is provided by heating equipment. The compensated heating temperature rise gradient curve is the power function. .

[0078] ; In the formula, The total compensation energy required to offset the delay, in units of . It is a nonlinear function fitted based on materials science experimental data, which shortens the delay time as needed. Volume of the material to be heated ,density and specific heat capacity The additional heat energy required to accelerate the curing process was calculated. The target duration for compensation intervention to maintain the original schedule baseline is equal to the original planned duration. The system ensures completion of curing within the target timeframe by inputting the total compensation energy required to offset delays. It should be noted that during this inverse integration calculation, the integration time variable t and the target timeframe for compensation intervention are... All of these must be switched and converted to s by the system. To compensate for the heating temperature rise gradient curve, an instantaneous power function varying with time is used, with units of W. Its functional form is solved by the control algorithm based on the principle of optimal energy utilization efficiency. For example, it can be a piecewise constant power function or a smoothly varying function, but its value at any given time t... All must be less than or equal to the maximum power determined by the highest output power parameter. .

[0079] The preset system time buffer safety tolerance alarm limit is a configurable threshold pair, such as {max_delay:2.0h, max_prob:0.15}. An alarm is triggered when any component in the project delay probability vector exceeds the corresponding threshold.

[0080] The Physical Environment Optimization Compensation Response Strategy Algorithm Library is a series of algorithm modules stored on the server. Each module corresponds to a physical intervention method, such as heating, dehumidification, and ventilation, and contains the corresponding physical process model and control algorithm.

[0081] The inverse physical control equation is the core of this algorithm library. It maps the engineering goal of shortening the construction period to physical control parameters such as the power curve through inverse solution. The compensated heating temperature rise gradient curve is the solution to this equation, which is represented in digital form as a sequence of timestamps and power setpoints.

[0082] Entity control strategy objects are standardized data structures, typically in XML or JSON format, containing all the information needed to perform a complete physical intervention, ensuring unambiguous communication between the decision-making and execution layers.

[0083] In one example of this embodiment, the central data processing server captures the node The project delay probability vector is [5.5h, 0.19]. The server's internal preset system time buffer safety tolerance alarm limit is set to ensure that the average expected delay does not exceed 2 hours. Since the detected delay of 5.5h is greater than 2 hours, the intervention mechanism is triggered. The system suspends operations on the task. The regular schedule was activated, and the hardware intervention process was initiated. The server reads the node. The process control handle was identified as being associated with a target infrared heating device. The server sent a query command to the device via the network, confirming it as an infrared heating device and obtaining its maximum output power parameter of 3.3kW. At this point, the system needs to compensate for the expected 5.5-hour delay caused by environmental degradation to ensure timely completion within the original planned 24-hour period. Therefore, the compensation intervention target period is set to 24 hours. The server invokes the inverse physical control equations, with the input parameters being... =5.5h, compensation intervention target project duration After solving the equation, a compensated heating temperature rise gradient curve was generated. The specific trend of this curve can be found in [reference needed]. Figure 4 As shown, the curve is defined as follows: from 0 to 12 hours after the intervention, heating is carried out continuously at 800W; from 12 to 24 hours, the power is reduced to 400W for heat preservation. Combined with... Figure 4 It can be seen that both power values ​​are within the capability range of the 3.3kW device. Finally, the server encapsulates the power curve and the execution trigger hardware timestamp obtained from the network time server into an entity control policy object, ready for distribution.

[0084] In one embodiment of the present invention, the control command issuing module is used to perform the following operations: Intercept routine task delay scheduling signaling to verify the protocol compatibility between the entity control strategy object and the process control handle communication interface; If the verification passes, the communication driver runtime library mapped to the hardware access identifier code is loaded via the hardware bus. Extract the electrical control target threshold from the entity control strategy object, generate targeted control instructions, and send them to the field execution node.

[0085] In order to transmit the entity control strategy object to the execution control manager, enter the forced intervention mode, generate targeted control instructions to change the temperature and humidity on site, the execution control manager inside the central data processing server will perform a series of verification and instruction conversion and distribution operations.

[0086] First, the manager, acting as the central hub for command issuance, proactively intercepts routine task delay scheduling signaling data generated by the regular scheduling subsystem for the same affected node, ensuring the highest priority of physical intervention commands. Simultaneously, the manager verifies the data string format of received entity control policy objects. It checks whether the structure of the data object conforms to a predefined JSON or XML format pattern and verifies that all fields, such as trigger time and power curve, are complete and of correct data type.

[0087] The manager further verifies the compatibility of the data string format settings with the communication interface protocol of its target device, i.e., the device pointed to by the process control handle, ensuring that the policy object can be correctly parsed by the target device's driver or firmware. After receiving final confirmation of successful verification, the manager invokes the server's underlying operating system hardware bus trunk, and through the input / output control interface, loads and activates the industrial communication controller hardware device communication driver runtime library, which has a unique device mapping relationship with the underlying hardware access identifier code contained in the entity control policy object. Once this driver library is activated, a dedicated communication link is established between the server and the target physical device.

[0088] The manager extracts the electrical control target threshold, defined by the power time series of the compensation heating temperature rise gradient curve, from the data structure encapsulated within the entity control strategy object. It then digitally encodes this high-level logic power curve using conversion functions built into the driver library, generating low-level binary data representing the switching frequency of the control equipment—the targeted control command. The manager sends this targeted control command to the field execution nodes via the established communication link.

[0089] The execution control manager is responsible for coordinating upper-level logic scheduling and lower-level hardware control, implementing hierarchical decision-making, and ensuring that the most appropriate action is taken in different scenarios, whether it is routine scheduling adjustments or physical environment intervention. Forced intervention mode means that after receiving an entity control policy object, the manager's system control priority will be prioritized, allowing it to pause or override scheduling instructions from other non-urgent software modules.

[0090] The targeted control command is binary data, and its format follows the message structure of a specific industrial communication protocol, such as Modbus TCP or Profinet. It contains the target device address, the address of the register to be written, and a hexadecimal number representing the power setting value.

[0091] Field execution nodes typically refer to industrial IoT gateways, programmable logic controllers, or embedded controllers deployed at construction sites. They directly connect to and control physical actuators, such as infrared heating equipment, and are entities that receive and ultimately execute targeted control commands.

[0092] In one example of this embodiment, the execution control manager receives a request for a task. The target device is an infrared heating device, and the physical control strategy object immediately enters the hardware-priority intervention mode. It first intercepts the task-related messages sent by the regular scheduling system. The plan adjustment instruction is postponed by 5.5 hours. The manager begins to verify the received entity control strategy object, confirming that its data structure is in valid JSON format, containing two control fields: a timestamp and a power curve. The power curve is a sequence composed of time offsets and corresponding power values. Verification confirms that this data structure is compatible with the communication protocol required by the target heating device's communication driver. After successful verification, the manager calls the operating system interface to load the Modbus TCP communication driver runtime library for controlling the device. Next, the manager parses the entity control strategy object, extracting the compensated heating temperature rise gradient curve, i.e., the corresponding segmented power setting time series, where the time offset is in seconds. The manager digitally encodes the first power setting value of 800W. Assuming that setting the device's power requires writing to a specific control register address, the manager converts the power value into hexadecimal control words in the corresponding protocol format and assembles them into a targeted control instruction containing the device address, function code, register address, and data value according to the Modbus TCP protocol format. Finally, when the predetermined trigger time specified in the entity control strategy object arrives, the manager sends this byte stream to the programmable logic controller (PLC) that controls the infrared heating device in the field via industrial Ethernet. The PLC then acts as the field execution node, thereby starting the first stage of the 800W heating program.

[0093] In one embodiment of the present invention, the intervention feedback monitoring module is used to perform the following steps: The thermal radiation energy conversion output efficiency is obtained by reading the resistance temperature feedback signal generated by the field execution node through a high-frequency polling mechanism. The thermal radiation energy conversion output efficiency is input into the composite material condensation kinetics simulation module for prediction. Accumulate the actual time spent on physical interventions to calculate and generate the deterministic intervention completion time.

[0094] To monitor and record the physical action response parameters of the targeted control command, collect and analyze the physical feedback characteristic signals, output the deterministic intervention completion time, and execute the control manager to immediately enter the closed-loop monitoring and simulation correction state after issuing the command.

[0095] Upon receiving the targeted control command, the manager immediately initiates a high-frequency polling mechanism, sending read commands to the field execution node at a frequency of, for example, five times per second, to retrieve real-time data from the temperature sensor associated with the heating wire integrated within the infrared heating device. The hardware node executes the action according to the command, causing the heating wire to heat up and its temperature to rise. This change is captured by the internal platinum resistance temperature sensor and converted into a resistance temperature feedback signal representing the current core temperature of the heating element, which is then transmitted back in response to the manager's polling request.

[0096] After receiving this signal stream, the server, combining the known surface area of ​​the heating equipment and the emissivity parameters of the material, calculates the thermal radiation energy conversion output efficiency of the hardware during the physical work process using a thermodynamic model. Then, the server inputs this real-time updated thermal radiation energy conversion output efficiency, along with real-time external ambient temperature data acquired from the temperature and humidity digital sensor array hardware, into the composite material coagulation kinetics simulation module deployed on the server. This software module simulates the coagulation and hydration reaction process of mortar based on the actual energy input efficiency and dynamically changing environmental boundary conditions. This effectively filters and attenuates the oscillating divergence effect on the actual construction progress variance data that may be caused by sudden on-site events, such as a sudden cold air convection. This prevents the chain reaction of delays caused by small physical disturbances in the project plan.

[0097] The manager uses a high-precision clock built into the host computer to calculate the actual elapsed time, starting from the microsecond-level timestamp of the initiation of the closed-loop active physical intervention response. This elapsed time is then added to the estimated remaining completion time output by the simulation software module, corrected for real physical parameters, to calculate and ultimately generate a high-confidence deterministic intervention completion time. This is used to directly overwrite and discard previously generated random delay guesses based on probability extrapolation by the system.

[0098] This step involves calculating the thermal radiation energy conversion output efficiency and the final deterministic intervention completion time.

[0099] Thermal radiation energy conversion output efficiency The calculation formula is: ; Deterministic intervention completion time The calculation formula is: ; In the formula, The thermal radiation energy conversion output efficiency is a dimensionless value, typically between 0 and 1. This refers to the net radiant heat power that the heating equipment actually outputs to the external environment, measured in W. The input electrical power is set by the targeted control command, and the unit is W. It is the emissivity of the material on the surface of the heating device, which is a known constant. It is the Stefan-Boltzmann constant, whose typical value is A is the total radiant surface area of ​​the heating equipment, a known parameter, in units of... . It is the absolute surface temperature of the heating equipment calculated from the resistance temperature feedback signal, and the unit is K. It is the absolute temperature of the ambient environment obtained from the digital temperature and humidity sensor array, and the unit is K. The time required to complete a definitive intervention is measured in hours (h). The accumulated time elapsed from the start of the intervention to the present moment. This is the remaining completion time predicted by the composite material condensation kinetics simulation module based on the current state; it represents efficiency. and environmental conditions The function.

[0100] It should be noted that the high-frequency polling mechanism is a communication mode in which the controller sends data read requests to field devices at a fixed high frequency, such as 5Hz, to achieve near real-time status monitoring.

[0101] The composite material condensation kinetics simulation module is a numerical simulation program based on partial differential equations. It simulates the curing process of materials under specific temperature and humidity conditions by solving transient heat conduction equations and chemical reaction kinetic equations that include internal heat sources.

[0102] The specific partial differential equation for internal heat conduction is defined as follows: ; In the formula, and These are the density and specific heat capacity of the composite material, respectively. Thermal conductivity, For the temperature field Laplace operator, The baseline exothermic rate for the material's hydration reaction is given. This is the thermal radiation energy conversion output efficiency calculated above, which is substituted into the source term as an energy boundary condition. The system solves this equation discretically using the finite difference method, iterating to obtain the remaining completion time.

[0103] The deterministic intervention completion time is the final product of this step. It is a floating-point number representing the total time that the construction operation is expected to take from start to finish under active physical intervention. Its determinism is much higher than any probability-based prediction.

[0104] In one example of this embodiment, after the manager issues a command to the target infrared heating device to heat it at 800W, closed-loop monitoring is immediately initiated. One hour after the intervention, i.e., 3600 seconds have elapsed and the system automatically converts and assigns the result to a value... The manager obtains the surface temperature of the device through high-frequency polling. It stabilized at 65℃, or 338.15K. Meanwhile, the on-site digital temperature and humidity sensor array reported the ambient temperature. The temperature is 15℃, or 288.15K. The emissivity of the surface material of this device is known. =0.9, radiative surface area . Reference Figure 5 The thermal radiation energy output conversion characteristic curve shown is used by the server to calculate the actual net radiation power using the formula. Therefore, the thermal radiation energy conversion output efficiency is calculated as follows: This efficiency value was immediately sent to the composite material solidification kinetics simulation module. Based on the actual energy input resulting from this efficiency and considering the current environmental conditions, the module resimulated the remaining curing process, concluding that it would take an estimated 22.8 hours to complete. Finally, the server adds the elapsed time to the remaining time to calculate the deterministic intervention completion time. This 23.8-hour deterministic duration data will be used to update the cyber-physical constraint graph in the next step.

[0105] In one embodiment of the present invention, the construction sequence optimization module is used to perform the following steps: In the time attribute block of the cyber-physical constraint diagram, the delayed delivery benchmark is updated using deterministic intervention completion time; Extract the associated data from the updated cyber-physical constraint graph, perform a forward traversal and trace downstream associated task nodes along the logical dependency edge of the start time, and filter the affected associated process constraint links. The Markov decision reinforcement optimization agent unit is activated to perform deduction on the constraint link of the related process and generate a globally optimized construction organization sequence that eliminates equipment resource conflicts.

[0106] To extract the deterministic intervention completion time, update the time node record parameters of the cyber-physical constraint diagram, perform local collaborative deduction and reconstruction calculations, and publish a globally optimized construction organization sequence with disturbance resistance resilience, the central data processing server will perform the final plan update and global optimization operations.

[0107] The server retrieves the deterministic intervention completion time generated in the previous stage from memory and retrieves the time attribute storage block of the target affected node in the location information physical constraint diagram. The server then uses the loaded deterministic intervention completion time to perform a write operation on the storage block, updating the original delayed delivery baseline.

[0108] The server extracts the associated data from the updated cyber-physical constraint graph and, starting from the updated node, performs a forward traversal along the start time logical dependency edge to trace downstream associated task nodes to recalculate the earliest start time and earliest finish time of all downstream tasks.

[0109] During this traversal, the system will automatically filter and identify those process nodes whose planned buffer time has been severely squeezed due to changes in the project schedule of upstream nodes, and which even face the risk of exceeding the time limit of the critical delivery path. These nodes and their dependency chains will be used to form the affected extended related process constraint links.

[0110] Once the system detects the formation of this link, it activates the Markov decision reinforcement agent unit in the background. This agent program specifically performs a multi-resource collaborative scheduling simulation for the physical space range involved in the extracted affected extended associated process constraint links, which are typically confined and limited in distribution.

[0111] Specifically, the reinforcement learning environment for this agent unit is defined as follows: State space is defined as ,in This represents the current estimated start-up time delay for each node in the link. This is a matrix showing the occupancy status of key equipment in each time window. This provides a buffer for the remaining time in downstream processes.

[0112] Action space is defined as This refers to the number of hours by which the start time of non-critical path node i in the affected link is delayed or advanced.

[0113] The reward function is defined as follows: ,in For the total construction period, To ensure the bottom line of delivery, This is a time penalty for overlapping use of equipment resources. These are the weighting coefficients.

[0114] By maximizing the cumulative reward, the agent program can efficiently generate a globally optimized construction organization sequence that eliminates equipment resource conflicts. This sequence is then distributed by the system to management terminals at all levels of the project.

[0115] The time attribute storage block of the target affected node refers to the data address in a graph database or relational database that stores the duration or deadline attribute of a specific task node. Its location ensures the accuracy of data updates.

[0116] The affected extended associated process constraint link is a dynamically generated data set that includes all downstream task chains affected by time propagation, starting from the updated node and continuing until the project ends or until there are no more subsequent tasks.

[0117] Markov decision reinforcement agent is a software intelligent agent based on reinforcement learning. It models the construction scheduling problem as a Markov decision process and learns the optimal scheduling strategy through interaction with the environment, i.e., the simulated project plan, to maximize long-term returns, such as minimizing the total project duration or resource conflicts.

[0118] Multi-resource collaborative scheduling simulation is an advanced decision-making algorithm used by this agent program. It treats the tasks in the affected links as collaborative players who need to coordinate the allocation of shared resources. It conducts collaborative game simulation on a chessboard of limited time and space resources, and quickly finds the global approximate optimal solution that can eliminate resource conflicts and minimize the overall project duration loss, i.e., the optimal scheduling scheme.

[0119] The globally optimized construction organization sequence is a construction task schedule that has been dynamically optimized and has no internal logic or resource conflicts. It can be directly issued to guide on-site operations.

[0120] In one example of this embodiment, the central data processing server extracts the deterministic intervention completion time calculated for task T4, which is 23.8 hours. The server then locates node N4 in the cyber-physical constraint graph and updates its internal attribute value representing the project duration from the originally planned 24 hours to the physically corrected 23.8 hours. Since the actual completion time of T4 is 0.2 hours ahead of schedule, the server initiates a forward traversal to update the earliest start time of its downstream task T5, enabling it to start 0.2 hours earlier. However, during the update process, the system detects that even though... Even with only a slight advance in starting work, the mobile scaffolding resources required will still overlap with tasks on another parallel task chain with a tight schedule. The wall construction in adjacent areas overlaps due to differences in usage time, causing a conflict. The system will therefore involve nodes... , , The links with resource constraints are identified as affected extended associated process constraint links. This situation automatically activates the Markov decision reinforcement agent unit. This agent program will... and A rapid game theory simulation was conducted, treating the two collaborating players as needing to coordinate and allocate mobile scaffolding resources. The agent program's strategy library, after evaluation, found that... Delaying the start time by 4 hours will consume its own float buffer time, but it will not affect the overall critical path of the project, and will instead allow... By fully utilizing the advance time window, the overall construction period can be shortened. Therefore, the agent program generates a globally optimized construction organization sequence, in which... The construction period was confirmed to be 23.8 hours. The start time was 0.2 hours ahead of schedule. The start time was delayed by 4 hours compared to the original plan. This new construction sequence eliminated resource conflicts, achieved global optimization, and was immediately distributed to the construction team's task management system.

[0121] Each of the modules can be implemented in whole or in part through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0122] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An AI-based brick masonry reinforcement construction progress optimization and management system, characterized in that, include: The information constraint graph construction module is used to acquire the decomposed structure data of the reinforcement project, extract independent task nodes and resource dependencies, and generate an information physical constraint graph containing process control handles. The monitoring data stream establishment module is used to acquire multi-source heterogeneous status signals collected by field sensing devices, perform time alignment and data cleaning, and obtain monitoring data streams mapped to independent task nodes. The delay risk quantification module is used to obtain the project schedule control target, compare the monitoring data stream with it, and obtain the schedule delay probability vector for the affected nodes. The control strategy generation module is used to obtain the project delay probability vector, call the built-in process control handle of the affected node, and generate an entity control strategy object. The control command issuance module is used to obtain entity control strategy objects, extract electrical control target thresholds, generate targeted control commands, and issue them to field execution nodes; The intervention feedback monitoring module is used to acquire the physical feedback characteristic signals of the field execution nodes based on the targeted control instructions, and to perform simulations and deductions in combination with environmental boundary conditions to obtain the deterministic intervention completion time. The construction sequence optimization module is used to obtain the deterministic intervention completion time, update the time node record parameters of the cyber-physical constraint diagram, perform local collaborative deduction, and obtain the globally optimized construction organization sequence.

2. The artificial intelligence-based brick masonry reinforcement construction progress optimization and management system according to claim 1, characterized in that, Generating a cyber-physical constraint graph containing process control handles includes the following steps: Extract reinforcement construction operations from the decomposed structural data of the reinforcement project, map them as independent task nodes in the topology graph, and establish an initial logical network structure for the project; Traverse all independent task nodes in the initial engineering logic network structure, embed industrial communication protocol interfaces into each independent task node, bind the hardware access identification code of the field physical control actuator, and generate process control handles. By integrating the initial engineering logic network structure with the control handles of each process, the logical dependency edges of start time between each independent task node are calculated, resource mutual exclusion constraint edges of shared construction equipment are established, and a cyber-physical constraint graph is generated.

3. The artificial intelligence-based brick masonry reinforcement construction progress optimization and management system according to claim 1, characterized in that, Obtaining the monitoring data stream mapped to individual task nodes includes the following steps: Temperature and humidity signal streams reflecting local microclimate changes are collected by temperature and humidity sensors installed in the brick masonry. The thermal image matrix of structural defects, which reflects internal defects or lesions of the wall, is obtained by infrared thermal imaging equipment. The temperature and humidity signal stream and the thermal image matrix of structural defects constitute a multi-source heterogeneous state signal. The network time synchronization protocol is used to align the timestamps of the temperature and humidity signal stream with the thermal map matrix of structural defects, and then perform filtering and extreme value removal operations to output the monitoring data stream.

4. The artificial intelligence-based brick masonry reinforcement construction progress optimization and management system according to claim 1, characterized in that, Obtaining the schedule delay probability vector for the affected nodes includes the following steps: Extract continuous time signal data segments from the monitoring data stream and input them into the trained spatiotemporal convolutional neural network model to extract the weight features of time delay influence; By importing the weighting characteristics of time lag into the Markov Monte Carlo statistical prediction model, the mean expected delay and the probability of extreme delay risk are calculated. The expected mean delay is correlated with the probability of extreme delay risk to the affected nodes in the cyber-physical constraint graph to generate a project delay probability vector.

5. The artificial intelligence-based brick masonry reinforcement construction progress optimization and management system according to claim 1, characterized in that, Generating entity control policy objects includes the following steps: When the probability vector of project delay exceeds the time alarm threshold, the scheduling progress is suspended and the physical environment compensation strategy is activated. Read the process control handle built into the affected node, verify the type of hardware device mapped by the hardware access identifier code, and when it is confirmed to be an infrared heating device, parse and extract its highest output power parameter. Based on the maximum output power parameter, an inverse physical control equation is established, the compensated heating temperature rise gradient curve is calculated, and the compensated heating temperature rise gradient curve and the hardware execution timestamp are encapsulated into an entity control strategy object.

6. The artificial intelligence-based brick masonry reinforcement construction progress optimization and management system according to claim 1, characterized in that, Generating targeted control instructions and issuing them to field execution nodes includes the following steps: Intercept routine task delay scheduling signaling to verify the protocol compatibility between the entity control strategy object and the process control handle communication interface; If the verification passes, the communication driver runtime library mapped to the hardware access identifier code is loaded via the hardware bus. Extract the electrical control target threshold from the entity control strategy object, generate targeted control instructions, and send them to the field execution node.

7. The artificial intelligence-based brick masonry reinforcement construction progress optimization and management system according to claim 1, characterized in that, To determine the duration of a definitive intervention, the following steps are required: The thermal radiation energy conversion output efficiency is obtained by reading the resistance temperature feedback signal generated by the field execution node through a high-frequency polling mechanism. The thermal radiation energy conversion output efficiency is input into the composite material condensation kinetics simulation module for prediction. Accumulate the actual time spent on physical interventions to calculate and generate the deterministic intervention completion time.

8. The artificial intelligence-based brick masonry reinforcement construction progress optimization and management system according to claim 1, characterized in that, Obtaining the globally optimized construction organization sequence includes the following steps: In the time attribute block of the cyber-physical constraint diagram, the delayed delivery benchmark is updated using deterministic intervention completion time; Extract the associated data from the updated cyber-physical constraint graph, perform a forward traversal and trace downstream associated task nodes along the logical dependency edge of the start time, and filter the affected associated process constraint links. The Markov decision reinforcement optimization agent unit is activated to perform deduction on the constraint link of the related process and generate a globally optimized construction organization sequence that eliminates equipment resource conflicts.

9. The artificial intelligence-based brick masonry reinforcement construction progress optimization and management system according to claim 5, characterized in that, The calculation of the compensated heating temperature rise gradient curve includes the following steps: The volume parameters, density parameters, and specific heat capacity parameters of the material to be heated are obtained. Combined with the delay time that needs to be shortened, the total compensation energy required to offset the delay is calculated based on the preset nonlinear function of material properties. Obtain the original planned construction period and set it as the target construction period for compensation intervention; Using the total compensation energy as the integration target value, the compensation intervention target period as the integration time limit, and the highest output power parameter as the instantaneous power constraint, the instantaneous power function is solved in reverse to obtain the compensation heating temperature rise gradient curve.

10. The artificial intelligence-based brick masonry reinforcement construction progress optimization and management system according to claim 7, characterized in that, Calculating the completion time for a deterministic intervention includes the following steps: The emissivity, total radiative surface area, and Stefan-Boltzmann constant of the material on the surface of the heating equipment are obtained. The absolute temperature of the surface of the heating equipment is obtained based on the resistance temperature feedback signal, and the absolute temperature of the ambient environment is also obtained. The difference between the fourth power of the absolute temperature of the heating equipment surface and the fourth power of the absolute temperature of the ambient environment is calculated. The difference is then multiplied by the material emissivity, total radiative surface area, and Stefan-Boltzmann constant to obtain the actual net radiative power. Divide the actual net radiation power by the input electrical power set by the targeted control command to obtain the thermal radiation energy conversion output efficiency. Obtain the estimated remaining completion time output by the composite material condensation kinetics simulation module, add the actual time spent to the estimated remaining completion time to obtain the deterministic intervention completion time.

Images