Multi-channel complex building engineering dynamic management system and method

Abstract

The invention relates to the technical field of constructional engineering management, and discloses a multi-channel complex constructional engineering dynamic management system and method. The system comprises a multi-source data integration module, a dynamic progress prediction module, a resource optimization distribution module, a risk dynamic evaluation module, a cooperative control module and the like. The multi-source data integration module collects multi-dimensional real-time data to obtain structured engineering data; the dynamic progress prediction module generates a construction progress prediction result in combination with a specific algorithm; the resource optimization distribution module constructs a model to distribute resources; the risk dynamic evaluation module calculates the risk probability; and the cooperative control module adjusts construction process parameters. According to the invention, dynamic management of complex constructional engineering can be realized, the data processing accuracy, the construction progress prediction precision and the resource allocation rationality are improved, the risk is accurately evaluated, the construction process is optimized, the constructional engineering management efficiency and quality are effectively improved, and the engineering is ensured to be carried out smoothly.

Description

Technical Field

[0001] The present invention relates to the technical field of construction project management, and specifically to a multi-channel dynamic management system and method for complex construction projects. Background Art

[0002] With the acceleration of the urbanization process, the scale and complexity of construction projects have increased sharply, and the disadvantages of traditional management models have become increasingly prominent, making it difficult to meet the management needs of modern construction projects.

[0003] In terms of data collection and processing, complex construction projects involve multi-dimensional data, such as construction progress, resource consumption, environmental parameters, and personnel allocation. Previous data collection methods were scattered and inefficient, with different data formats between different systems and difficult to be compatible, forming a large number of "data islands". At the same time, the data processing technology was backward, unable to effectively remove noise and fill in missing values, resulting in poor data quality and unable to provide a reliable basis for management decisions. For example, the construction site environment is complex, sensors are easily interfered, and the raw data collected often contains a large number of outliers. If not processed, it will seriously affect the accuracy of subsequent analysis.

[0004] In construction progress management, traditional methods are mostly based on experience and static plans, lacking consideration of dynamic factors. Construction projects are affected by various factors such as seasons, weather, and material supply. Traditional progress plans are difficult to cope with these changes, often resulting in project delays. For example, during rainy season construction, if outdoor operations are still carried out according to the original plan, it will surely be blocked due to weather reasons, but the traditional management model is difficult to adjust the progress arrangement in time. Moreover, the accuracy of traditional progress prediction methods is low, unable to accurately estimate the construction progress of each stage in advance, making it difficult for all parties involved in the project to make preparations.

[0005] In the link of resource allocation, there is a lack of scientific planning. The allocation of human, material, and equipment resources often depends on subjective judgment, without fully considering the actual needs of the project and the finiteness of resources. This results in the coexistence of resource waste and shortage. For example, there is an overabundance of manpower in some work areas, while there is a shortage of manpower in key processes; the material procurement plan is unreasonable, resulting in overstock or shortage, increasing costs while affecting the construction progress. And during the process of resource scheduling, there is a lack of an effective mechanism for detecting and resolving resource conflicts, further reducing the resource utilization efficiency.

[0006] In terms of risk assessment, existing methods mainly rely on manual experience judgment, lacking systematicness and real-time nature. There are numerous and complex risk factors at the construction site, and it is difficult for manual work to comprehensively and timely identify and assess risks. For example, the potential risks brought by some new construction technologies may not be detected in time by manual work; for sudden environmental changes, such as strong winds and heavy rains, it is difficult to make risk assessments and responses in time. At the same time, there is a lack of an effective risk warning mechanism, and once an accident occurs, it often causes serious losses.

[0007] There are also problems with collaborative control of the construction process. The lack of effective communication and coordination between the various construction links and the poor information transmission have led to frequent problems such as confusion in the construction sequence and cross-operation conflicts. For example, due to information asymmetry between different construction teams, repeated construction in the same area or reversed construction sequence may occur, affecting the construction quality and progress. Moreover, when problems arise during the construction process, it is difficult to quickly adjust the construction process parameters and lacks dynamic control capabilities. Summary of the invention

[0008] The object of the present invention is to provide a multi-channel complex construction project dynamic management system and method to solve the problems raised in the above background technology.

[0009] To achieve the above object, the present invention provides the following technical solution: a multi-channel complex construction project dynamic management system, the system comprising:

[0010] Multi-source data integration module, used to collect multi-dimensional real-time data of construction projects and obtain structured engineering data;

[0011] A dynamic progress prediction module is used to input the structured engineering data into a prediction model based on time series decomposition, and generate multi-stage construction progress prediction results by combining seasonal trend decomposition and random forest regression algorithm;

[0012] A resource optimization allocation module is used to construct a multi-objective integer programming model based on the multi-stage construction progress prediction results, allocate human, material and equipment resources through a dynamic priority scheduling algorithm, and generate a resource allocation plan;

[0013] The risk dynamic assessment module is used to extract construction site image features using convolutional neural networks, perform real-time risk probability calculations in combination with the historical accident database, and generate a dynamic risk assessment matrix;

[0014] The collaborative control module is used to design a fuzzy logic controller to adjust construction process parameters based on the dynamic risk assessment matrix and output real-time collaborative control instructions.

[0015] Preferably, the prediction model based on time series decomposition includes:

[0016] Dividing the structured engineering data into time windows according to the construction phase, and extracting the progress indicators and resource consumption in each window;

[0017] Apply seasonal trend decomposition algorithm to separate trend term, seasonal term and residual term;

[0018] The decomposed components are input into the random forest regression model for independent prediction, and the prediction results of each component are weighted and fused to generate the multi-stage construction progress prediction results.

[0019] Preferably, the allocation of human, material, and equipment resources through the dynamic priority scheduling algorithm to generate a resource allocation plan includes:

[0020] Define a resource conflict detection function to identify time overlap conflicts between human resources and equipment;

[0021] Construct a two-objective optimization model that maximizes resource utilization and minimizes project duration delay;

[0022] Use the non-dominated sorting genetic algorithm to solve the Pareto front and screen the optimal resource allocation plan.

[0023] Preferably, the calculation of the real-time risk probability includes:

[0024] Extract multi-scale features from the construction site images to generate a spatial attention feature map;

[0025] Match the feature map with the spatio-temporal distribution of historical accident data to calculate the risk similarity;

[0026] Fuse the similarity and environmental parameters through a Naive Bayes classifier to output the dynamic risk probability value.

[0027] Preferably, the fuzzy logic controller adjusts the construction process parameters, including:

[0028] Define the input variables as the risk level and progress deviation, and the output variable as the adjustment amount of the operation priority;

[0029] Construct Gaussian membership functions to describe the fuzzy intervals of the input variables;

[0030] Use the centroid method for defuzzification to generate an accurate correction instruction for the operation execution order.

[0031] Preferably, the method for obtaining structured engineering data includes:

[0032] Collect multi-dimensional real-time data of the construction project, including construction progress, resource consumption, environmental parameters, and personnel allocation data;

[0033] Based on preset data cleaning rules, perform denoising and normalization processing on the multi-dimensional real-time data. The data cleaning rules include:

[0034] Detect abnormal outliers in the original data and use the isolation forest-based outlier removal algorithm;

[0035] For missing sensor data, use spatio-temporal Kriging interpolation method combined with data from adjacent work areas to fill in;

[0036] Convert the filled data into normalized data with a mean of zero and a variance of one according to the Z-score normalization method.

[0037] Preferably, the system further includes:

[0038] A feedback optimization module, configured to integrate multi-source feedback data, and use the Bayesian optimization algorithm to iteratively correct the collaborative control instructions to generate an optimized project management strategy;

[0039] The iterative correction of the Bayesian optimization algorithm includes:

[0040] Construct a Gaussian process surrogate model to describe the relationship between collaborative control instructions and construction efficiency;

[0041] Select the next set of candidate parameter combinations through the expected improvement function;

[0042] Update the surrogate model and perform iterative calculations until convergence to the optimal set of control parameters.

[0043] Preferably, the system further includes:

[0044] A three-dimensional visualization modeling module, configured to dynamically bind the construction progress prediction result to the BIM model to generate a real-time three-dimensional construction progress simulation animation;

[0045] Based on the deviation detection result of the simulation animation, trigger the secondary scheduling of the resource optimization allocation module.

[0046] Preferably, the system further includes:

[0047] Construct an anomaly response module based on reinforcement learning, train a decision-making agent through the Q-learning algorithm, and autonomously select an emergency response strategy according to the real-time risk probability;

[0048] Convert the strategy into a device shutdown instruction or a personnel evacuation instruction, and synchronously update the collaborative control module.

[0049] Preferably, the present invention further includes a multi-channel complex building project dynamic management method, which is applied to a multi-channel complex building project dynamic management system as described above. The method includes the following steps:

[0050] Step 1: Use a multi-source data integration module to collect multi-dimensional real-time data of a building project to obtain structured project data;

[0051] Step 2: Input the structured project data into a prediction model based on time series decomposition through a dynamic progress prediction module, and combine seasonal trend decomposition and random forest regression algorithm to generate a multi-stage construction progress prediction result;

[0052] Step 3: The resource optimization allocation module constructs a multi-objective integer programming model according to the multi-stage construction progress prediction result, and allocates human, material, and equipment resources through a dynamic priority scheduling algorithm to generate a resource allocation plan;

[0053] Step 4: The risk dynamic assessment module uses a convolutional neural network to extract the characteristics of the construction site images, combines with the historical accident database to calculate the real-time risk probability, and generates a dynamic risk assessment matrix.

[0054] Step 5: The collaborative control module designs a fuzzy logic controller to adjust the construction process parameters based on the dynamic risk assessment matrix, and outputs real-time collaborative control instructions.

[0055] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0056] The multi-channel dynamic management system and method for complex construction projects of the present invention have significant positive effects. In terms of data management, the multi-source data integration module can collect multi-dimensional real-time data of construction projects, and perform denoising and normalization processing through preset cleaning rules to obtain high-quality structured engineering data. The outlier removal algorithm based on Isolation Forest can accurately detect and remove abnormal outliers in the original data. The spatio-temporal Kriging interpolation method combines with data from adjacent work areas to fill in missing sensor data. The Z-score standardization method unifies the data scale, ensuring the accuracy and availability of the data, and providing a solid data foundation for subsequent management decisions.

[0057] In terms of construction progress management, the dynamic progress prediction module inputs the structured engineering data into a prediction model based on time series decomposition, combines seasonal trend decomposition and random forest regression algorithms to generate multi-stage construction progress prediction results. By dividing the time window according to the construction stage, extracting progress indicators and resource consumption, separating the trend term, seasonal term and residual term, and then predicting and weighted fusion respectively, it can fully consider various dynamic factors in the construction process, such as the impact of seasonal changes on the construction progress. Compared with traditional prediction methods, it greatly improves the accuracy of prediction. This helps all parties involved in the project to make construction plans in advance, allocate resources reasonably, and effectively avoid project delays.

[0058] The resource optimization allocation module constructs a multi-objective integer programming model according to the multi-stage construction progress prediction results, and allocates human, material and equipment resources through a dynamic priority scheduling algorithm. Defining a resource conflict detection function can identify time overlap conflicts between human and equipment. The constructed double-objective optimization model of maximizing resource utilization rate and minimizing project duration delay uses the non-dominated sorting genetic algorithm to solve the Pareto front and screen the optimal solution, realizing the scientific and reasonable allocation of resources. It not only improves the resource utilization efficiency, reduces resource waste, but also ensures the project is completed on time and reduces project costs.

[0059] The risk dynamic assessment module uses a convolutional neural network to extract the characteristics of construction site images, combines with the historical accident database to calculate the real-time risk probability, and generates a dynamic risk assessment matrix. The multi-scale feature extraction generates a spatial attention feature map, matches the spatio-temporal distribution of historical accident data to calculate the risk similarity, and then outputs the risk probability value by fusing environmental parameters through a Naive Bayes classifier, realizing the real-time and accurate assessment of construction site risks. This enables project managers to timely discover potential risks, take preventive measures in advance, effectively reduce the probability of accidents, and ensure the safety of construction workers and the smooth progress of the project.

[0060] The collaborative control module designs a fuzzy logic controller based on the dynamic risk assessment matrix to adjust the construction process parameters. The input variables are defined as the risk level and progress deviation, the output variable is the adjustment amount of the operation priority, a Gaussian membership function is constructed to describe the fuzzy interval, and the centroid method is used for defuzzification to generate accurate instructions. It can adjust the execution order of construction operations in real time according to the actual risk situation and progress of the construction site, optimize the construction process. Ensure that the construction process is carried out efficiently under the premise of safety, improve the construction quality and overall efficiency.

[0061] In addition, the feedback optimization module integrates multi-source feedback data, uses the Bayesian optimization algorithm to iteratively correct the collaborative control instructions, and continuously optimizes the project management strategy, enabling the system to continuously adapt to project changes. The 3D visualization modeling module dynamically binds the construction progress prediction results with the BIM model to generate a real-time 3D construction progress simulation animation, which is convenient for project personnel to intuitively understand the construction progress. Based on the deviation detection results, it can also trigger the secondary scheduling of the resource optimization allocation module to timely adjust the resource configuration. The anomaly response module based on reinforcement learning trains the decision-making agent through the Q-learning algorithm, and autonomously selects the emergency response strategy according to the real-time risk probability, enhancing the system's ability to respond to sudden risks and further ensuring the safety and stability of the project. Brief Description of the Drawings

[0062] Figure 1 It is the working principle diagram of the multi-channel complex building project dynamic management system described in the present invention;

[0063] Figure 2 It is the working flow chart of the prediction model based on time series decomposition;

[0064] Figure 3 It is the working flow chart of the resource optimization allocation module;

[0065] Figure 4 It is the principle diagram of the fuzzy logic controller adjusting the construction process parameters. Detailed Embodiment

[0066] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present invention.

[0067] Please refer to Figures 1 - 4 , the present invention provides a technical solution: a multi-channel complex construction project dynamic management system, and the implementation manner of this system includes:

[0068] The multi-source data integration module is responsible for collecting multi-dimensional real-time data of the construction project, covering various aspects such as construction progress, resource consumption, environmental parameters, and personnel allocation data. After collecting this data, noise reduction and normalization processing are performed on the multi-dimensional real-time data based on preset data cleaning rules, thereby obtaining structured project data. The processed data can more accurately reflect the actual situation of the project and provide a reliable basis for the operation of subsequent modules.

[0069] After obtaining the structured project data, the dynamic progress prediction module inputs this data into a prediction model based on time series decomposition. This model combines seasonal trend decomposition and random forest regression algorithms to generate multi-stage construction progress prediction results. This prediction result is of great significance for reasonably arranging the construction plan and making early resource allocation.

[0070] The resource optimization allocation module works based on the multi-stage construction progress prediction results generated by the dynamic progress prediction module. It constructs a multi-objective integer programming model and allocates human, material, and equipment resources through a dynamic priority scheduling algorithm, thereby generating a resource allocation plan. This plan can ensure the reasonable allocation of resources, improve resource utilization efficiency, and ensure the smooth progress of the project.

[0071] The risk dynamic assessment module uses a convolutional neural network to extract features from construction site images and combines a historical accident database to calculate real-time risk probabilities, generating a dynamic risk assessment matrix. This matrix can reflect the risk status of the construction site in real time and provide strong support for taking corresponding risk response measures.

[0072] The collaborative control module works based on the dynamic risk assessment matrix generated by the risk dynamic assessment module. It designs a fuzzy logic controller to adjust construction process parameters and outputs real-time collaborative control instructions. Through these instructions, the construction process is optimized and adjusted to ensure the safety and efficiency of the construction process.

[0073] The following further illustrates the present invention in combination with Embodiments 1 to 6:

[0074] Embodiment 1:

[0075] In this embodiment, the specific implementation process of the prediction model based on time series decomposition is further elaborated.

[0076] Divide the structured engineering data into time windows according to the construction stages. The division of construction stages is carried out according to the actual construction process and characteristics of the building project, such as the foundation construction stage, the main structure construction stage, the decoration construction stage, etc. The length of each time window is determined according to the scale and complexity of the project, generally in units of days, weeks or months. Within the divided time windows, extract the progress indicators and resource consumption within each window. The progress indicators can be the proportion of the completed engineering quantity to the total engineering quantity, the number of floors constructed, etc.; the resource consumption includes man-hours, the usage amount of building materials, the usage duration of construction equipment, etc.

[0077] Apply the seasonal trend decomposition algorithm to separate the trend term, seasonal term and residual term. The seasonal trend decomposition algorithm is a commonly used time series analysis method, and its principle is to decompose the time series data into different components to better understand and predict the change trend of the data. Assume that the time series data is y t , after seasonal trend decomposition, the trend term T t , seasonal term S t and residual term R t can be obtained, that is, y t = T t + S t + R t . Among them, the trend term T t reflects the long-term change trend of the data over time, the seasonal term S t reflects the seasonal fluctuations of the data within a fixed period, and the residual term R t contains other random factors that cannot be explained by the trend term and seasonal term.

[0078] Input the decomposed components into the random forest regression model for independent prediction respectively. The random forest regression model is an ensemble learning algorithm based on decision trees, which improves the accuracy and stability of prediction by constructing multiple decision trees and integrating their prediction results. For the trend term T t , seasonal term S t and residual term R t , train the corresponding random forest regression models respectively to obtain their respective prediction results.

[0079] Weightedly fuse the prediction results of each component to generate the multi-stage construction progress prediction results. In order to obtain more accurate prediction results, it is necessary to weightedly fuse the prediction results of each component. The weights can be determined according to the importance of each component in the original data and the accuracy of historical predictions. Assume that the prediction result of the trend term T t is Season item S t The prediction result of Residual item R t The prediction result of The weights are w1, w2, and w3 respectively, and w1 + w2 + w3 = 1. Then the multi-stage construction progress prediction result is: In this way, the information of each component can be fully utilized, the accuracy of construction progress prediction can be improved, and a strong guarantee can be provided for the smooth progress of the project.

[0080] Embodiment 2:

[0081] This embodiment details the specific implementation process of allocating human, material, and equipment resources through the dynamic priority scheduling algorithm to generate a resource allocation plan.

[0082] In actual construction projects, resource conflicts are common problems. Therefore, a resource conflict detection function is first defined to identify time overlap conflicts between human and equipment. The resource conflict detection function is designed based on the construction progress plan and resource usage records. Assume that the working time interval of human resources is [t h1 , t h2 , and the usage time interval of equipment resources is [t e1 , t e2 . When t h1 ≤ t e2 and t e1 ≤ t h2 , it is determined that there is a time overlap conflict. Through this function, possible resource conflict situations during the construction process can be quickly and accurately found.

[0083] Next, a two-objective optimization model of maximizing resource utilization rate and minimizing project duration delay is constructed. The goal of maximizing resource utilization rate aims to ensure that human, material, and equipment resources are fully utilized and avoid resource idle waste. Let the usage efficiency of human be u h , the usage efficiency of equipment be u e , and the consumption efficiency of materials be u m . The resource utilization rate objective function can be expressed as max(U) = w h u h + w e u e + w m u m , where w h , w e , w m are the corresponding weight coefficients, and w h + w e + w m= 1. The goal of minimizing construction delay is to ensure that the project can be completed according to the scheduled plan and minimize construction delays. Suppose the planned construction period of the project is T plan , the actual possible construction period is T actual , the construction period delay is D = T actual -T plan , the objective function of minimizing construction delay is min(D).

[0084] Then, the non-dominated sorting genetic algorithm is used to solve the Pareto frontier and screen the optimal resource allocation scheme. The non-dominated sorting genetic algorithm is a classic multi-objective optimization algorithm that simulates the biological evolution process and continuously iterates and optimizes the population through operations such as selection, crossover and mutation. In this embodiment, the resource allocation scheme is encoded as individuals, and the individuals in the population are divided into different levels through non-dominated sorting, and individuals with higher levels are given priority. At the same time, the crossover and mutation operations of the genetic algorithm are used to continuously explore new resource allocation schemes. After multiple iterations, the Pareto frontier, that is, a set of non-dominated optimal solutions, is obtained. The resource allocation scheme that best meets the actual needs of the project is screened from the Pareto frontier. The scheme achieves a good balance between resource utilization and construction period delay, and can effectively improve the efficiency of project management.

[0085] Embodiment 3:

[0086] In this embodiment, the specific implementation steps of real-time risk probability calculation are described in detail, including:

[0087] Perform multi-scale feature extraction on the construction site images to generate spatial attention feature maps. Utilize the powerful feature extraction capability of convolutional neural networks to perform convolution operations on construction site images at different scales. For example, set up multiple convolution layers, each of which uses convolution kernels of different sizes, such as 3×3, 5×5, 7×7, etc. Through these convolution operations of different scales, features of different levels and ranges in the image can be extracted. At the same time, the spatial attention mechanism is introduced, which can automatically learn the importance of different areas in the image and assign higher weights to important areas, thereby generating a more representative spatial attention feature map. Assume that the feature map obtained after convolution operation and spatial attention mechanism processing is F, and each element F in F ij It represents the feature value of the image at position (i, j), and its size reflects the importance of this position in the image.

[0088] Match the feature map with the spatio-temporal distribution of historical accident data to calculate the risk similarity. Historical accident data contains information such as the time, location, accident type, and related image features of various accidents that occurred in the past. Compare the feature map of the current construction site image with the image features in the historical accident data and calculate the similarity between them. Methods such as cosine similarity can be used for calculation. Let the similarity between the feature map F and a certain feature map F h in the historical accident data be sim(F, F h ), and the calculation formula is where F·F h is the dot product of the two feature maps, and |F| and |F h | are the norms of the two feature maps respectively. By calculating the similarity with the feature maps of multiple historical accident data, a risk similarity set is obtained.

[0089] Fuse the similarity and environmental parameters through a Naive Bayes classifier to output the dynamic risk probability value. The Naive Bayes classifier is a classification algorithm based on Bayes' theorem, which assumes that the features are independent of each other. Use the risk similarity and environmental parameters (such as weather conditions, terrain conditions of the construction site, etc.) as the input features of the Naive Bayes classifier. Let the risk similarity be S, the environmental parameter be E, and the dynamic risk probability value be P(R). According to Bayes' formula Obtain the parameters of the classifier through training, and then calculate the dynamic risk probability value. This value can reflect the risk level of the construction site in real time and provide an important decision-making basis for project managers.

[0090] Example 4:

[0091] This example details the specific implementation method of the fuzzy logic controller for adjusting the construction process parameters.

[0092] Specifically include:

[0093] Define the input variables as the risk level and the progress deviation, and the output variable as the adjustment amount of the operation priority. The risk level is determined according to the dynamic risk assessment matrix generated by the risk dynamic assessment module, and can be divided into different levels such as low risk, medium risk, and high risk. The progress deviation is calculated by comparing the current actual construction progress with the planned construction progress. If the actual progress lags behind the planned progress, the progress deviation is a positive value, otherwise it is a negative value. The adjustment amount of the operation priority is used to adjust the execution order of the construction operations. A positive value indicates increasing the operation priority, and a negative value indicates decreasing the operation priority.

[0094] Construct Gaussian membership functions to describe the fuzzy intervals of the input variables. The Gaussian membership function is a commonly used fuzzy function, and its formula is Where x is the value of the input variable, c is the center value of the membership function, and σ is the standard deviation. For risk levels, different risk levels correspond to different Gaussian membership function parameters. For example, the center value c1 of the Gaussian membership function of the low risk level is set to a lower risk assessment value, and the standard deviation σ1 is small to indicate that the range of the low risk level is relatively narrow; the center value c3 of the high risk level is set to a higher risk assessment value, and the standard deviation σ3 is large to cover a wider range of high-risk situations. For schedule deviation, the corresponding Gaussian membership function parameters are also set according to the actual engineering needs to accurately describe the fuzzy interval of schedule deviation.

[0095] The centroid method is used to defuzzify and generate accurate process execution sequence correction instructions. The centroid method is a commonly used defuzzification method. Its principle is to calculate the centroid of the fuzzy set as the accurate output value. Assume that the fuzzy output set is A and its membership function is μ A (x), the output value y after defuzzification is calculated as In actual calculation, the center of gravity is approximated by weighted averaging the discrete fuzzy output values. According to the calculated process priority adjustment amount, accurate process execution order correction instructions are generated, so as to reasonably adjust the construction process and ensure the smooth progress of the construction process.

[0096] Embodiment 5:

[0097] In this embodiment, the method of obtaining structured engineering data and the specific working process of the feedback optimization module are described in detail. Specifically, it includes:

[0098] In terms of obtaining structured engineering data, we first collect multi-dimensional real-time data of construction projects, including construction progress, resource consumption, environmental parameters and staffing data. Construction progress data is obtained through monitoring equipment and construction log records at the construction site, such as recording the amount of work completed each day and the floors that have been constructed. Resource consumption data is obtained through records of the use of building materials, the usage time of construction equipment, and the number of manpower hours. Environmental parameter data is collected with the help of various sensors, such as temperature sensors, humidity sensors, wind speed sensors, etc., to obtain real-time environmental conditions at the construction site. Staffing data includes information such as the number of people in each type of work at the construction site and the scheduling of personnel.

[0099] After collecting multi-dimensional real-time data, it is denoised and normalized based on the preset data cleaning rules. When detecting abnormal outliers in the original data, an outlier removal algorithm based on isolation forest is used. The algorithm calculates the outlier score of each data point by constructing an isolation forest model. The higher the outlier score, the more likely the data point is an outlier. Assuming that the outlier score of data point x is S(x), when S(x) is greater than the set threshold, x is determined to be an outlier and is removed.

[0100] For the missing sensor data, the spatio-temporal Kriging interpolation method is used to fill it in combination with the data of adjacent work areas. The spatio-temporal Kriging interpolation method is an interpolation method based on spatial autocorrelation, which takes into account the spatial position and time factors of the data. Suppose the sensor data at the position (x i , y i ) and time t j is missing. Through the known data of adjacent work areas at similar times and positions, the estimated value Z(x0, y0, t0) of the missing value is calculated using the spatio-temporal Kriging interpolation formula , where Z(x i , y i , t i ) is the known data of adjacent work areas, and λ i is the weight coefficient, which is obtained by solving the semi-variance function.

[0101] The filled data is converted into normalized data with a mean of zero and a variance of one according to the Z-score normalization method. The Z-score normalization formula is , where x is the original data, μ is the mean of the data, and σ is the standard deviation of the data. After the normalization process, the data has a unified scale, which is convenient for subsequent analysis and processing.

[0102] In terms of the feedback optimization module, it is used to integrate multi-source feedback data, and the Bayesian optimization algorithm is used to iteratively correct the collaborative control instructions to generate an optimized project management strategy. First, a Gaussian process surrogate model is constructed to describe the relationship between the collaborative control instructions and the construction efficiency. The Gaussian process surrogate model is a model based on probability distribution, which can model and predict unknown functions. Let the collaborative control instruction be x and the construction efficiency be y. The Gaussian process surrogate model assumes that y follows a Gaussian distribution N(m(x), k(x, x ′ ), where m(x) is the mean function and k(x, x ′ ) is the covariance function.

[0103] Then, the next set of candidate parameter combinations is selected through the expected improvement function. The expected improvement function is used to measure the expected improvement that can be brought by selecting a certain parameter combination under the current model. Let the current optimal observed value be y best . The calculation formula of the expected improvement function EI(x) is , where Φ is the cumulative distribution function of the standard normal distribution and φ is the probability density function of the standard normal distribution. By maximizing the expected improvement function, the next set of candidate parameter combinations is selected.

[0104] Finally, update the surrogate model and perform iterative calculations until convergence to the optimal set of control parameters. After obtaining new observed data each time, update the parameters of the Gaussian process surrogate model, recalculate the expected improvement function, select a new combination of candidate parameters, and iterate continuously until the convergence condition is met to obtain the optimal set of control parameters, thereby generating an optimized project management strategy and improving the effectiveness of project management.

[0105] Example 6:

[0106] The 3D visualization modeling module is responsible for dynamically binding the construction progress prediction results with the BIM model to generate a real-time 3D construction progress simulation animation. The BIM model is short for Building Information Model, which contains rich content such as geometric information, physical information, and construction information of the building project. Associate the multi-stage construction progress prediction results generated by the dynamic progress prediction module with the BIM model, and update the construction status of each component in the BIM model in real time according to the changes in the construction progress. For example, when it is predicted that the main structure of a certain floor will be completed at a specific time, correspondingly update the status of the main structure components of that floor in the BIM model from under construction to completed in the BIM model. Through this dynamic binding, use visualization technology to generate a real-time 3D construction progress simulation animation. Users can intuitively view information such as the construction progress of the project, the construction sequence of each component, and the spatial position relationship, which is convenient for timely discovery of possible problems in the construction process.

[0107] Based on the deviation detection results of the simulation animation, trigger the secondary scheduling of the resource optimization allocation module. In the simulation animation, detect the progress deviation in real time by comparing the actual construction progress with the planned construction progress. When the deviation exceeds the set threshold, it indicates that an abnormal situation has occurred in the construction process and the resources need to be reallocated. At this time, trigger the resource optimization allocation module, reconstruct the multi-objective integer programming model, and reallocate human, material, and equipment resources through the dynamic priority scheduling algorithm according to the current construction progress and resource usage situation to generate a new resource allocation plan to ensure that the project can resume normal construction progress as soon as possible.

[0108] Construct an anomaly response module based on reinforcement learning, train the decision-making agent through the Q-learning algorithm, and autonomously select emergency response strategies according to the real-time risk probability. The Q-learning algorithm is a reinforcement learning algorithm based on the value function. In the anomaly response module, the state of the decision-making agent consists of the real-time risk probability at the construction site and other relevant environmental information. The actions of the decision-making agent include emergency response operations such as equipment shutdown instructions and personnel evacuation instructions. Suppose the decision-making agent transfers to state s ′ , the obtained reward is r, the learning rate is α, the discount factor is γ, and the Q-value update formula is Q(s,a)=(1 - α)Q(s,a)+α(r + γmax a′ Q(s′ , a ′ ). By continuously interacting with the environment, the decision-making agent selects the optimal action according to the Q-value, thereby learning a reasonable emergency response strategy.

[0109] It should be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0110] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-channel dynamic management system for complex construction projects, characterized in that, The system includes: A multi-source data integration module, which is used to collect multi-dimensional real-time data of a construction project and obtain structured engineering data; A dynamic progress prediction module, which is used to input the structured engineering data into a prediction model based on time series decomposition, and combine seasonal trend decomposition and random forest regression algorithm to generate multi-stage construction progress prediction results; A resource optimization allocation module, which is used to construct a multi-objective integer programming model according to the multi-stage construction progress prediction results, and allocate human, material and equipment resources through a dynamic priority scheduling algorithm to generate a resource allocation plan; A risk dynamic assessment module, which is used to extract the image features of the construction site by using a convolutional neural network, and combine with a historical accident database to calculate the real-time risk probability and generate a dynamic risk assessment matrix; A collaborative control module, which is used to design a fuzzy logic controller to adjust the construction process parameters based on the dynamic risk assessment matrix and output real-time collaborative control instructions.

2. The multi-channel complex construction project dynamic management system according to claim 1, characterized in that, The prediction model based on time series decomposition includes: Dividing time windows for the structured engineering data according to construction stages, and extracting progress indicators and resource consumption amounts within each window; Applying a seasonal trend decomposition algorithm to separate the trend term, seasonal term and residual term; Inputting the decomposed components into a random forest regression model for independent prediction respectively, and weighted-fusing the prediction results of each component to generate multi-stage construction progress prediction results.

3. The multi-channel complex construction project dynamic management system according to claim 1, characterized in that, The process of allocating human, material and equipment resources through a dynamic priority scheduling algorithm to generate a resource allocation plan includes: Defining a resource conflict detection function to identify time overlap conflicts between human and equipment; Constructing a two-objective optimization model for maximizing resource utilization and minimizing project duration delay; Using a non-dominated sorting genetic algorithm to solve the Pareto front and screen the optimal resource allocation plan.

4. The multi-channel complex construction project dynamic management system according to claim 1, characterized in that, The real-time risk probability calculation includes: Performing multi-scale feature extraction on the construction site image to generate a spatial attention feature map; Matching the feature map with the spatio-temporal distribution of historical accident data to calculate the risk similarity; Fusing the similarity and environmental parameters through a Naive Bayes classifier to output a dynamic risk probability value.

5. The multi-channel complex construction project dynamic management system according to claim 1, characterized in that, The process of the fuzzy logic controller adjusting the construction process parameters includes: Defining the input variables as the risk level and progress deviation, and the output variable as the adjustment amount of the operation priority; Constructing Gaussian membership functions to describe the fuzzy intervals of the input variables; Using the centroid method for defuzzification to generate an accurate instruction for correcting the operation execution order.

6. The multi-channel complex construction project dynamic management system according to claim 1, characterized in that The method for obtaining structured engineering data includes: Collecting multi-dimensional real-time data of a construction project, including construction progress, resource consumption, environmental parameters and personnel allocation data; Performing denoising and normalization processing on the multi-dimensional real-time data based on preset data cleaning rules, and the data cleaning rules include: Detecting abnormal outliers in the original data and using an outlier removal algorithm based on isolation forest; For missing sensor data, using spatio-temporal Kriging interpolation method combined with data of adjacent work areas to fill in; Converting the filled data into normalized data with a mean of zero and a variance of one according to the Z-score standardization method.

7. The multi-channel complex construction project dynamic management system according to claim 1, wherein It also includes: A feedback optimization module, which is used to integrate multi-source feedback data, and uses the Bayesian optimization algorithm to iteratively correct the collaborative control instructions to generate an optimized project management strategy; The iterative correction of the Bayesian optimization algorithm includes: Constructing a Gaussian process surrogate model to describe the relationship between collaborative control instructions and construction efficiency; Selecting the next set of candidate parameter combinations through the expected improvement function; Updating the surrogate model and performing iterative calculations until convergence to the optimal regulation parameter set.

8. The multi-channel complex construction project dynamic management system according to claim 1, characterized in that, The system further includes: A 3D visualization modeling module, which is used to dynamically bind the construction progress prediction result to the BIM model to generate a real-time 3D construction progress simulation animation; Based on the deviation detection result of the simulation animation, trigger the secondary scheduling of the resource optimization allocation module.

9. The multi-channel complex construction project dynamic management system according to any one of claims 1 to 8, characterized in that The system further includes: Construct an anomaly response module based on reinforcement learning, train a decision-making agent through the Q-learning algorithm, and autonomously select an emergency response strategy according to the real-time risk probability; Convert the strategy into a device shutdown instruction or a personnel evacuation instruction, and synchronously update the collaborative control module.

10. A multi-channel dynamic management method for complex construction projects, which is applied to the multi-channel dynamic management system according to any one of claims 1 to 9, and is characterized in that, Including the following steps: Step 1: Use the multi-source data integration module to collect multi-dimensional real-time data of the construction project to obtain structured project data; Step 2: Input the structured project data into the prediction model based on time series decomposition through the dynamic progress prediction module, and combine seasonal trend decomposition and random forest regression algorithm to generate a multi-stage construction progress prediction result; Step 3: The resource optimization allocation module constructs a multi-objective integer programming model according to the multi-stage construction progress prediction result, and allocates human, material and equipment resources through the dynamic priority scheduling algorithm to generate a resource allocation plan; Step 4: The risk dynamic assessment module uses a convolutional neural network to extract the image features of the construction site, combines with the historical accident database to calculate the real-time risk probability, and generates a dynamic risk assessment matrix; Step 5: The collaborative control module designs a fuzzy logic controller to adjust the construction process parameters based on the dynamic risk assessment matrix, and outputs real-time collaborative control instructions.

Images