Artificial intelligence-based engineering progress management system

Abstract

The application discloses an engineering progress management system based on artificial intelligence and relates to the technical field of engineering progress management. Multidimensional parameters such as the structure type, process complexity, building area, contract duration and climate condition of a target engineering are extracted, are standardized and are weighted by an entropy weight method to construct a comprehensive feature vector, and a reference engineering set with high matching degree is intelligently screened from a historical engineering library based on the Euclidean distance. A standardized construction difficulty coefficient is generated by an analytic hierarchy process by fusing a first-class difficulty index and combining a historical delay rate of the reference engineering and a second-class difficulty index. Meanwhile, expected material demand is calculated according to a theoretical material quantity and a reference engineering loss rate, material supply and timeliness are locked through actual procurement quantity, and a warning mechanism is triggered. The expected construction period of a current node task is calculated by comprehensively considering various parameters, and closed-loop correction is performed through actual period deviation after the task is completed. Weight parameters are dynamically adjusted, and the intelligentization and self-adaptive iteration of engineering progress management are realized.

Description

Technical Field

[0001] This invention belongs to the field of engineering progress management technology, specifically, it relates to an artificial intelligence-based engineering progress management system. Background Technology

[0002] Traditional project schedule management relies heavily on manual experience to formulate construction plans. When faced with multiple overlapping processes, resource allocation, and uncontrollable environmental factors, problems such as plans deviating from reality and delayed responses often occur.

[0003] Application CN116934274A discloses an artificial intelligence-based engineering progress management system. This application sets up an engineering case matching module to obtain reference engineering cases, providing a data source for setting relevant values ​​of the target project. It sets up a node task construction difficulty comparison module and a node material supply control module to take into account the problem of construction cycle imbalance caused by different construction difficulties and quantities of different projects, as well as the problem of construction delay caused by untimely material supply. It sets up an expected node construction cycle prediction module to calculate the node expected cycle compensation coefficient and expected node construction cycle of the target project. It sets up an engineering expected node construction cycle correction module to correct the predicted expected node construction cycle based on the ratio of the actual construction cycle of the previous project node to the expected construction cycle, further improving the accuracy of data prediction. In the actual implementation of this application, the surface oxygen release rate of soil was introduced as the core matching condition. However, this indicator is only related to the environmental safety of radon prevention in buildings, and has no direct causal relationship with the project progress, construction period, or construction difficulty. This caused the matching results to deviate from the actual characteristics of the project. A large number of historical engineering cases with real reference value were excluded because the indicator did not match. The matched cases may have huge differences from the target project in key dimensions such as structural type, construction technology, and project scale. The cycle prediction based on this was built on a distorted reference from the beginning. Secondly, in terms of construction difficulty assessment, the application relies entirely on the subjective rating of engineers. Different engineers may have completely different judgments on the difficulty of the same project. There is a lack of unified scoring standards and reliability testing mechanisms. Moreover, the application cannot be implemented for new projects or unconventional engineering types without historical scoring data, because its highly subjective assessment method makes the construction difficulty coefficient unrepeatable and unverifiable. Furthermore, the material supply timeliness calculation formula in this application uses the theoretical material quantity rather than the actual procurement quantity as the numerator, which results in the calculated supply rate being almost always less than 1 and failing to truly reflect whether the material procurement meets the construction needs. If this formula is implemented, the material supply control will be completely ineffective. Finally, this application is a static open-loop system. Although it records the actual construction cycle, it does not design any feedback mechanism to use this real data to optimize the model parameters. The prediction deviation will accumulate step by step as the project progresses and cannot be corrected. The error of the previous node will directly cause the prediction of the next node to deviate more from reality.

[0004] To address the aforementioned issues, this invention proposes an artificial intelligence-based engineering progress management system. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an AI-based engineering progress management system, which solves the problems of static prediction models, inability to adaptively adjust, over-reliance on subjective scoring, and lack of key variables in existing technologies.

[0006] The objective of this invention can be achieved through the following technical solutions: An AI-based project progress management system, the system comprising: The multi-dimensional matching end locks in the multi-dimensional feature parameters of the target project, constructs the multi-dimensional fusion features associated with the target project, generates a comprehensive feature vector, and locks in the comprehensive matching degree between each historical project and the target project from the historical project database based on the Euclidean distance between the comprehensive feature vectors, thus determining the reference project set. On the construction difficulty assessment side, a first-class difficulty index of the current node task of the target project is obtained based on the construction design drawings, and a weighted fusion is performed by combining the second-class difficulty index of each reference project to generate a standardized construction difficulty coefficient of the current node task. On the material supply control side, the theoretical material quantity of the current node task is determined, the expected material quantity is calculated by combining the material loss rate of the reference project during the same node task, and the material supply timeliness rate is locked in by combining the actual material procurement quantity of the target project for the current node task. On the construction cycle prediction end, the construction cycle of the reference project at the corresponding node task is extracted, and the standardized construction difficulty coefficient, material supply timeliness rate and adaptive weight parameters of the target project are combined to calculate the expected construction cycle of the current node task of the target project. The closed-loop correction end determines the actual construction period after the completion of the current node task of the target project in real time, locks the period deviation, generates a construction period anomaly report, and dynamically adjusts the adaptive weight parameters of the expected construction period.

[0007] As a further aspect of the present invention, in the multidimensional matching end, the specific method for locking the multidimensional feature parameters of the target project and constructing the multidimensional fusion features associated with the target project is as follows: Determine the target project Q and its construction design drawings QM; Multidimensional feature parameters of the target project Q are extracted from QM, including project structure type, construction process complexity level, total building area, contract period, and climate classification of construction region. Among them, the construction process complexity level is obtained by technical personnel assessment. Standardize the multidimensional feature parameters to construct a multidimensional fusion feature vector FQ=(f1,f2,f3,f4,f5) for the target project Q, where f1 is the project structure type encoding value, which is obtained by converting the project structure type based on a preset project structure type mapping table; f2 is the normalized value of the construction process complexity level, which is obtained by mapping the construction process complexity level assessed by technical personnel to the interval [0,1]. f3 is the logarithmically standardized value of the total building area of ​​the project, obtained by performing Z-score standardization after natural logarithmic transformation; f4 is the relative ratio of contract duration, which is expressed as the ratio of the contract duration of the target project to the average duration of similar benchmark projects. The similar benchmark projects are projects pre-marked by technical personnel. f5 is the climate classification code value for the construction area, which is obtained by converting the climate classification of the construction area based on the preset climate zoning mapping table.

[0008] As a further aspect of the present invention, the specific method for generating the comprehensive feature vector in the multi-dimensional matching end is as follows: Differential weight coefficients are assigned to the feature parameters of each dimension in the multidimensional fusion feature vector FQ, namely w1, w2, w3, w4, w5. The differential weight coefficients are determined based on the entropy weight method analysis of the contribution of each dimension feature parameter in the historical project database to the construction cycle, and the sum of the differential weight coefficients is 1. By weighting, we obtain FQ'=(w1f1,w2f2,w3f3,w4f4,w5f5), which is labeled as the comprehensive feature vector FQ' of the target project Q.

[0009] As a further aspect of the present invention, in the multi-dimensional matching end, the specific method for determining the reference project set by locking the comprehensive matching degree between each historical project and the target project from the historical project database based on Euclidean distance is as follows: Extract the engineering structure type of the target project Q, and denote it as QA; Using QA as the search criteria, a precise search is performed from the historical project database to extract candidate historical projects, which are then sorted in reverse chronological order to obtain the candidate historical project sequence Q1, Q2, ..., Qj, where j is the total number of candidate historical projects; Extract the comprehensive feature vector of each candidate historical project, and align the candidate historical project sequence to form the historical comprehensive feature vector FQ1', FQ2', ..., FQj'; Calculate the Euclidean distance between each historical comprehensive feature vector and the comprehensive feature vector FQ' of the target project Q in sequence to obtain the Euclidean distance sequence D1',D2',...,Dj'; Extract the operator's preset Euclidean distance threshold, filter j Euclidean distances, remove Euclidean distances greater than the threshold, and then extract the historical projects corresponding to the smallest α% Euclidean distances from the remaining Euclidean distances, denoted as the reference project set CQ={CQ1,CQ2,...,CQm}, where m is the total number of reference projects and α% is the preset percentage.

[0010] As a further aspect of the present invention, the construction difficulty assessment end also includes a construction design drawing QM based on the target project Q, and the node tasks of the target project Q are determined according to the construction time sequence, denoted as the node task sequence R1, R2, ..., Ro, where o is the total number of node tasks; Determine the task Ru of the current node, 1 ≤ u ≤ o; Extract a type of difficulty index for the current node task Ru, including the number of job types per unit area, the number of processes per unit area, and the proportion of high-risk operations; Next, obtain the two types of difficulty indicators for m reference projects in the reference project set CQ={CQ1,CQ2,...,CQm} when they are on the same node task, specifically the average delay rate; The Analytic Hierarchy Process (AHP) is used to compare each difficulty index in a class of difficulty indices pairwise to determine the index weights, and then weighted and fused to generate the basic value of construction complexity, Su. The two types of difficulty indicators are integrated with the basic value of construction complexity Su to obtain the standardized construction difficulty coefficient η_u of the current node task Ru of the target project Q, where η_u = Su × (1 + average delay rate).

[0011] As a further aspect of the present invention, in the material supply control end, the specific method for locking the material supply timeliness rate by determining the actual procurement quantity of materials for the current node task of the joint target project is as follows: Based on the construction design drawing QM of the target project Q, extract the theoretical material quantity Vu required for the target project Q at the current node task Ru; Extract the material loss rate ρ of m reference projects in the reference project set CQ={CQ1,CQ2,...,CQm} when they are at the same node task; The expected material quantity YVu of the target project Q at the current node task Ru is calculated by YVu=Vu×(1+ρ); Obtain the actual material purchase quantity GVu of the target project Q at the current node task Ru, and calculate the material supply timeliness rate γ_u=min(1,GVu / YVu).

[0012] As a further aspect of the present invention, in the material supply control system, when the material supply timeliness rate γ_u < 1, a material early warning is triggered: The material shortage amount ΔVu is calculated using ΔVu=YVu-GVu, and a corresponding material early warning report is generated to notify the operator to perform material replenishment. The replenishment amount is ΔVu+ΔX, where ΔX is the preset redundant material amount.

[0013] As a further aspect of the present invention, the specific method for calculating the expected construction period of the current node task of the target project in the construction period prediction terminal is as follows: Extract the actual construction period Ti of each reference project in the reference project set CQ={CQ_1,CQ_2,...,CQ_m} at the corresponding node task R_u, 1≤i≤m, and calculate the average construction period T_avg=(1 / m)×ΣT_i; Obtain the standardized construction difficulty coefficient η_u of the current node task R_u of the target project Q, and obtain the timely material supply rate γ_u; Extract the adaptive weight parameter λ, and the initial value of the adaptive weight parameter λ is a preset constant λ_0; The expected construction period Tu of the current node task R_u of the target project Q is calculated by Tu=T_avg×(1+η_u)×(2-γ_u)×λ.

[0014] As a further aspect of the present invention, the specific method for locking the period deviation and generating a construction period anomaly report in the closed-loop correction terminal is as follows: Extract the actual construction period Tu_re after the completion of the current node task R_u of the target project Q; Extract the expected construction period Tu of the current node task R_u of the target project Q, calculate the period deviation ΔTu=Tu_re-Tu, and calculate the period deviation rate δ_u=ΔTu / Tu; Compare the preset period deviation rate tolerance threshold δ_th with the period deviation rate δ_u: If δ_u≤δ_th, record ΔTu in the record of the current node task R_u of the target project Q in the historical cycle deviation database; If δ_u > δ_th, a construction cycle anomaly report is generated, reminding operators that the actual construction cycle of the current node task R_u is abnormal.

[0015] As a further aspect of the present invention, the specific method for dynamically adjusting the adaptive weight parameter of the expected construction period in the closed-loop correction terminal includes: The closed-loop correction end performs iterative adjustments based on the actual periodic deviation after each node task is completed. The adjustment method is λ_new=λ_old×(1+θ×deviation rate), where θ is the preset learning rate, λ_old is the adaptive weight parameter before iterative adjustment, and λ_new is the adaptive weight parameter after iterative adjustment.

[0016] The beneficial effects of this invention are: This invention improves the accuracy of reference project selection by deeply mining historical engineering data through a multi-dimensional matching end, providing a reliable benchmark for subsequent evaluation. The construction difficulty assessment end integrates design drawing indicators with historical experience, making the standardized construction difficulty coefficient both objective and adaptable, avoiding the bias of a single data source. The material supply control end couples theoretical usage with historical loss rates and dynamically links them to actual procurement quantities, quantifying the on-time supply rate in real time, injecting a constraint dimension of material security into progress forecasting. The construction cycle forecasting end introduces adaptive weight parameters, allowing the expected cycle to be flexibly adjusted according to the material supply status, enhancing the responsiveness to actual disturbances. The closed-loop correction end actively triggers the dynamic optimization of weight parameters through actual cycle feedback, forming an intelligent closed loop, generating anomaly reports in a timely manner to warn of risks, and continuously enhancing the predictive adaptability in continuous iteration, realizing the leap from static estimation to dynamic self-evolution in project progress management, reducing the probability of project delays and improving the agility of resource allocation. This invention improves the screening accuracy of historical engineering reference samples and the adaptability of construction difficulty assessment by accurately quantifying multidimensional feature parameters and dynamically configuring weights. It transforms unstructured attributes such as engineering structure type and construction process complexity into a unified feature vector through mapping and normalization. Combined with the differentiated weights determined by the entropy weight method, it avoids subjective weighting bias and makes the comprehensive feature vector more realistically reflect the actual impact of each factor on the construction cycle. This invention, based on a matching mechanism of Euclidean distance and threshold filtering, ensures the homogeneity of reference projects. By dynamically selecting the most similar historical cases through proportional segmentation, it enhances the reliability of the reference objects. At the node task level, it introduces a hierarchical fusion of a type of difficulty index and a type of average delay rate. It uses the analytic hierarchy process to scientifically quantify the index weights and outputs them in the form of standardized construction difficulty coefficients. This allows the difficulty assessment to reflect both the structural complexity of the project itself and the actual delay risks of historical nodes, thus providing quantitative support with both theoretical basis and practical guidance for construction schedule planning, resource allocation, and risk warning. This invention corrects theoretical demand by introducing the loss rate of a reference project at the material supply control end, and calculates the on-time supply rate by combining it with the actual procurement volume. This makes the material shortage early warning more realistic and can effectively buffer sudden risks by setting a redundancy, thereby avoiding project delays caused by material shortages. The construction cycle prediction end integrates the average construction cycle, difficulty coefficient and material supply on-time rate, and introduces adaptive weight parameters, so that cycle estimation is transformed from a static empirical value to a quantitative result that is sensitive to real-time site conditions. Finally, the closed-loop correction end triggers anomaly reports by the actual cycle deviation rate and iteratively adjusts the weight parameters based on the deviation rate. This allows the system to continuously optimize its prediction model in continuous node tasks, gradually improving the prediction accuracy of subsequent node tasks, and providing intelligent decision support with both robustness and evolution for the entire life cycle of the project. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 This is a schematic diagram of the system described in this invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, this application provides an artificial intelligence-based engineering progress management system, which is deployed on a cloud server or local computing device through software modules, and interacts with the construction project management system, building information model (BIM) database, enterprise resource planning (ERP) system and material procurement platform through application programming interfaces. The technical solution of the present invention will be described in detail below with reference to specific embodiments.

[0021] As an embodiment 1 of this application, it specifically includes: The multi-dimensional matching end locks in the multi-dimensional feature parameters of the target project, constructs the multi-dimensional fusion features associated with the target project, generates a comprehensive feature vector, and locks in the comprehensive matching degree between each historical project and the target project from the historical project database based on the Euclidean distance between the comprehensive feature vectors, thus determining the reference project set. On the construction difficulty assessment side, a first-class difficulty index of the current node task of the target project is obtained based on the construction design drawings, and a weighted fusion is performed by combining the second-class difficulty index of each reference project to generate a standardized construction difficulty coefficient of the current node task. On the material supply control side, the theoretical material quantity of the current node task is determined, the expected material quantity is calculated by combining the material loss rate of the reference project during the same node task, and the material supply timeliness rate is locked in by combining the actual material procurement quantity of the target project for the current node task. On the construction cycle prediction end, the construction cycle of the reference project at the corresponding node task is extracted, and the standardized construction difficulty coefficient, material supply timeliness rate and adaptive weight parameters of the target project are combined to calculate the expected construction cycle of the current node task of the target project. The closed-loop correction end determines the actual construction period after the completion of the current node task of the target project in real time, locks the period deviation, generates a construction period anomaly report, and dynamically adjusts the adaptive weight parameters of the expected construction period.

[0022] Example 2

[0023] This embodiment is implemented based on Embodiment 1, and details the basic structure and core workflow of the artificial intelligence-based project progress management system, such as... Figure 1 As shown, the specific content is as follows: The AI-based engineering progress management system includes a multi-dimensional matching terminal, a construction difficulty assessment terminal, a material supply control terminal, a construction cycle prediction terminal, and a closed-loop correction terminal. The multi-dimensional matching terminal is communicatively connected to the construction difficulty assessment terminal and the material supply control terminal, respectively. The outputs of the construction difficulty assessment terminal and the material supply control terminal are both connected to the construction cycle prediction terminal. The construction cycle prediction terminal and the closed-loop correction terminal form a feedback loop.

[0024] The multidimensional matching end locks the multidimensional feature parameters of the target project, constructs the multidimensional fusion features associated with the target project, generates a comprehensive feature vector, and locks the comprehensive matching degree between each historical project and the target project from the historical project database based on the Euclidean distance between the comprehensive feature vectors, thereby determining the reference project set.

[0025] Specifically, the multidimensional matching end extracts five multidimensional feature parameters from the BIM model of the target project: engineering structure type, construction process complexity level, total building area, contract period, and construction region climate classification. Each parameter is numerically and standardized to form a multidimensional fusion feature vector. Then, the weighted vector is generated by weighting the parameters based on the contribution of each feature parameter to the construction cycle in the historical project database.

[0026] The historical project database is a relational database deployed on the server, storing the aforementioned feature parameters and construction records of completed projects. The multi-dimensional matching end traverses the records in the historical project database that have the same structural type as the target project, calculates the Euclidean distance of the corresponding comprehensive feature vector, filters out the reference project set based on the preset Euclidean distance threshold and percentage, and persists the set to system memory.

[0027] The construction difficulty assessment end obtains the first-class difficulty index of the current node task of the target project based on the construction design drawings, and performs weighted fusion with the second-class difficulty index of each reference project to generate the standardized construction difficulty coefficient of the current node task. The construction difficulty assessment end parses the node task sequence from the construction design drawings according to the construction schedule plan. For the current execution node, it extracts three first-class difficulty indicators: the number of trades per unit area, the number of procedures per unit area, and the proportion of high-risk operations. The weight of each indicator is determined by the analytic hierarchy process and the weighted synthesis is used to form the basic value of construction complexity. Meanwhile, the average delay rate of the corresponding node tasks in the reference project set is extracted as a second-class difficulty index. The basic value of construction complexity is integrated with the second-class difficulty index to obtain the standardized construction difficulty coefficient.

[0028] The material supply control unit determines the theoretical material quantity for the current node task, calculates the expected material quantity by combining the material loss rate of the reference project during the same node task, and locks in the timely material supply rate by combining the actual material procurement quantity of the target project for the current node task.

[0029] The material supply control unit reads the theoretical material quantity from the construction design drawing quantity calculation module, queries the material loss rate records of the same node task in the reference project set, calculates the typical loss rate by arithmetic average, and then calculates the expected material quantity including loss. It then obtains the actual purchased material quantity that has been signed for from the ERP system, and calculates the material supply timeliness rate by taking the smaller value between the ratio of the actual purchased material quantity and the expected material quantity and 1.

[0030] The construction cycle prediction end extracts the construction cycle of the reference project at the corresponding node task, and combines it with the standardized construction difficulty coefficient output by the construction difficulty assessment end, the material supply timeliness rate output by the material supply control end, and the adaptive weight parameters stored in the system parameter table to calculate the expected construction cycle of the target project at the current node task.

[0031] The construction cycle prediction terminal first calculates the arithmetic mean of the actual construction cycles of the same node task in the reference project set. Then, based on this average value, it multiplies by (1 + standardized construction difficulty coefficient) and (2 - timely material supply rate), and multiplies by the current adaptive weight parameter to obtain the expected construction cycle. The calculation result is then written into the progress prediction log table.

[0032] The closed-loop correction end determines the actual construction period after the completion of the current node task of the target project in real time, calculates the period deviation from the expected construction period, generates a construction period anomaly report, and dynamically adjusts the adaptive weight parameters.

[0033] After the node task is accepted, the closed-loop correction terminal extracts the actual construction period from the construction log system, calculates the deviation amount and deviation rate, and sends an anomaly report to the project management personnel terminal when the deviation rate exceeds the preset tolerance threshold. At the same time, it iteratively updates the adaptive weight parameters according to the deviation rate and the preset learning rate, and writes the updated weight parameters back to the system parameter table for use in predicting the construction period of the next node task.

[0034] The system presented in this embodiment locks down a reference set from historical projects through multi-dimensional feature matching, overcoming the subjectivity of single empirical analogy, and enabling the construction cycle prediction to simultaneously quantify and consider three major factors: difficulty, material timeliness, and reference construction period. This allows for a more dynamic and objective progress forecast than traditional Gantt chart estimation.

[0035] Example 3

[0036] This embodiment is based on Embodiment 2, focusing on the implementation of the multi-dimensional matching end, including the complete implementation of engineering feature parameter standardization, differential weight calculation, and reference engineering set screening, specifically including the following: In this embodiment, the target project Q is a residential project in a certain city.

[0037] The multidimensional matching end extracts multidimensional feature parameters from the construction design drawing QM of the target project Q. The project structure type is obtained by reading the ObjectType attribute of the IfcBuilding entity in the BIM model, such as frame-shear wall structure. Then, it is encoded into numerical values ​​according to the preset project structure type mapping table. The mapping table structure is as follows: The code for brick-concrete structure is 1.0, the code for frame structure is 2.0, the code for frame-shear wall structure is 3.0, the code for shear wall structure is 4.0, and the code for steel structure is 5.0. Therefore, f1 = 3.0.

[0038] The complexity level of the construction process is assessed by the technical manager based on factors such as the number of new processes involved in the project and the degree of limitation of construction space. It is rated from level 1 to 5 and linearly mapped to the interval [0,1]: level 1 is mapped to 0.2, level 2 to 0.4, level 3 to 0.6, level 4 to 0.8, and level 5 to 1.0.

[0039] The current project is rated as level 4, therefore f2 = 0.8.

[0040] The total building area of ​​the project is S = 45,200 square meters. Using the natural logarithm transformation, we get ln(S)≈10.718. Then, based on the mean μ_ln=10.2 and the standard deviation σ_ln=0.85 of the logarithmic values ​​of the total building area of ​​all projects in the historical project database, we perform Z-score standardization and get f3=(10.718-10.2) / 0.85≈0.609.

[0041] The contract duration ratio f4 is calculated by dividing the target project's contract duration of 360 days by the average duration of similar benchmark projects of 380 days. The similar benchmark projects consist of 5 typical similar projects marked by the technical manager from the system, and their average duration is 380 days. Therefore, f4 = 360 / 380 ≈ 0.947.

[0042] The climate classification of the construction area is based on the "Standard for Building Climate Zoning". The project is located in a cold region and is coded as 2 in the preset climate zoning mapping table. The mapping table is as follows: severe cold region = 1, cold region = 2, hot summer and cold winter region = 3, hot summer and warm winter region = 4, temperate region = 5. Therefore, f5 = 2.0.

[0043] Thus, a multidimensional fusion feature vector FQ=(3.0,0.8,0.609,0.947,2.0) is constructed.

[0044] The multidimensional matching end further generates a comprehensive feature vector. For the above five feature parameters, the entropy weight method is used to analyze the contribution of each item to the dispersion of the construction cycle from the historical project database to determine the differential weight coefficient.

[0045] The entropy weight method calculation process is as follows: Extract all n historical records from the historical project database, construct an n×5 feature matrix, normalize the data in each column, and calculate the information entropy. The greater the information utility, the higher the weight.

[0046] In this embodiment, w1=0.25, w2=0.30, w3=0.10, w4=0.15, w5=0.20 were calculated, and the sum of the weights was 1. The comprehensive feature vector FQ'=(0.75,0.24,0.0609,0.14205,0.40) was obtained by weighting.

[0047] When filtering the reference project set, first extract the engineering structure type code value QA=3.0 of the target project Q. Use engineering structure type=3.0 as the search condition to perform a precise query in the historical project database and sort them in descending order by completion and acceptance time to obtain the candidate historical project sequence Q1, Q2,...,Qj, where j=86.

[0048] Extract the comprehensive feature vectors FQ1', FQ2', ..., FQ86' of each of the 86 projects, and calculate the Euclidean distance between each historical comprehensive feature vector and the comprehensive feature vector FQ' of the target project. For example, the distance to a certain historical project is: D = √[(0.75-0.68)]. 2 +(0.24-0.22) 2 +(0.0609-0.055) 2 +(0.14205-0.150) 2 +(0.40-0.38) 2], thus obtaining the Euclidean distance sequence.

[0049] The operator pre-sets the Euclidean distance threshold to 1.2 and the percentage α% to 20% in the system configuration interface. The system removes items greater than 1.2 from 86 distance values, keeping 42. These distances are then sorted in ascending order, and the first 20%, i.e., the first 9 (42 × 0.2 = 8.4 rounded up to 9) distances, correspond to the historical projects, forming the reference project set CQ = {CQ1, CQ2, ..., CQ9}, where m = 9.

[0050] This embodiment uses multi-dimensional feature standardization and differentiated weighting based on objective entropy weighting to ensure that the comprehensive feature vector can accurately reflect the similarity between projects. At the same time, it adopts a strategy of first filtering by structural type and then adaptive percentage screening by Euclidean distance to ensure that the reference project set has both structural similarity and similarity in construction period characteristics.

[0051] Example 4

[0052] This embodiment is implemented based on Embodiment 3, and details the collaborative working method of the construction difficulty assessment end and the material supply control end. It covers the integrated calculation of the first and second difficulty indicators, the determination of the material supply timeliness rate, and the material early warning logic. The implementation of the remaining ends in this embodiment is the same as in Embodiment 3. The results obtained in Embodiment 3 are used as reference engineering sets, as follows: The construction difficulty assessment first extracts the node task sequence R1, R2, ..., Ro from the construction schedule associated with the construction design drawing QM of the target project Q according to the construction time sequence, where o=14, corresponding to earthwork excavation, foundation pad, raft foundation, basement structure, main structure floors 1-5, main structure floors 6-10, etc. The current node task is R3=raft foundation, u=3.

[0053] One method for obtaining difficulty indicators is as follows: Read the area A=1200 square meters of the raft foundation construction area in the BIM model. From the construction process simulation data, count the number of trade types involved in this node (C_trade=5, for example: steelworker, formwork worker, concrete worker, waterproofing worker, electrician), and the number of process steps (C_proc=18). Identify the number of process steps involving high-altitude operations or large-scale hoisting (H=3), and obtain: Number of job types per unit area = C_trade / (A / 100) = 5 / 12 = 0.417 per 100 square meters; Number of processes per unit area = C_proc / (A / 100) = 18 / 12 = 1.5 processes / 100 square meters; The percentage of high-risk operations = H / C_proc = 3 / 18 ≈ 0.1667.

[0054] The weights of these three indicators were determined using the analytic hierarchy process (AHP): a pairwise comparison judgment matrix was constructed, and experts determined the importance of the number of job types per unit area relative to the number of work processes per unit area to be 3, and the importance relative to the proportion of high-risk operations to be 5; the importance relative to the proportion of high-risk operations to be 2.

[0055] The calculated weight vector is (0.648, 0.230, 0.122), and the consistency ratio CR = 0.008 < 0.1, thus passing the consistency test.

[0056] The weighted fusion generates the basic value of construction complexity, Su = 0.648 × 0.417 + 0.230 × 1.5 + 0.122 × 0.1667 ≈ 0.270 + 0.345 + 0.020 ≈ 0.635.

[0057] The second type of difficulty index is the average delay rate of the 9 reference projects in the reference project set CQ at the same node task "raft foundation". The actual construction period and planned period of the raft foundation of these 9 projects were extracted from the historical project database, and the delay rates were calculated to be 0.12, 0.05, 0.18, 0.07, 0.10, 0.03, 0.15, 0.08, and 0.11, respectively, with an average delay rate of 0.1044.

[0058] The construction difficulty assessment module integrates the two types of difficulty indicators with the basic value of construction complexity. It calculates the standardized construction difficulty coefficient η_3 = 0.635 × (1 + 0.1044) ≈ 0.701 according to η_u = Su × (1 + average delay rate), and stores this standardized construction difficulty coefficient in the current node task context.

[0059] Then, based on the bill of quantities in the construction design drawing QM, the material supply control end extracts the theoretical concrete usage Vu=860 cubic meters for the current node task R3. The material loss rate records of 9 projects in the reference project set at the raft foundation node task are queried: ρ1=0.04, ρ2=0.06, ρ3=0.05, ρ4=0.03, ρ5=0.07, ρ6=0.04, ρ7=0.05, ρ8=0.06, ρ9=0.05. The average material loss rate ρ=0.05 is calculated.

[0060] Based on YVu=Vu×(1+ρ), the expected material quantity is calculated as YVu=860×1.05=903 cubic meters.

[0061] The material supply control system obtains the actual amount of purchased materials GVu that has been signed for under the current node task of the target project Q through a data interface established with the enterprise's ERP system. The interface adopts RESTful API, returns data in JSON format, and is called once per hour.

[0062] In this study, GVu = 870 cubic meters was obtained, and the timely material supply rate was calculated as γ_u = min(1, GVu / YVu) = min(1, 870 / 903) ≈ 0.963.

[0063] Since γ_3=0.963<1, the material early warning logic is triggered, and the material shortage ΔVu=YVu-GVu=903-870=33 cubic meters is calculated.

[0064] The system reads the preset redundant material quantity ΔX, which is configured by the project material administrator in the system. Here it is set to 5 cubic meters. A material early warning report is generated, which includes the node task name, theoretical usage, expected usage, actual usage, shortage, and suggested replenishment quantity of 33+5=38 cubic meters. The report is then pushed to the mobile terminal and desktop client of the material management personnel via the message queue telemetry transmission protocol to remind them to replenish the material.

[0065] This embodiment combines the micro-difficulty index extracted from its own design drawings with the macro-delay rate of the reference project to derive a standardized construction difficulty coefficient. This makes the difficulty evaluation reflect both the specific process complexity of the project and the actual delay level in history. The calculation of the material timeliness rate incorporates the loss rate of the reference project and sets up an early warning compensation mechanism for redundancy. This helps to avoid work stoppages due to untimely material arrival or insufficient loss estimation, thereby improving resource security capabilities.

[0066] Example 5

[0067] This embodiment builds upon the previous embodiment, explaining the coordinated implementation of the construction cycle prediction and closed-loop correction ends. It covers the calculation of the expected construction cycle, the determination of cycle deviation, and the dynamic adjustment of adaptive weight parameters, as detailed below: This embodiment is implemented based on the standardized construction difficulty coefficient η_3=0.701 and material supply timeliness rate γ_3=0.963 obtained in Embodiment 4; The actual construction periods Ti (in days) of the 9 reference projects in the reference project set CQ at the node task R3 raft foundation are extracted from the construction period prediction end: T1=22, T2=25, T3=28, T4=20, T5=24, T6=23, T7=27, T8=21, T9=26. Calculate the average construction period: T_avg=(22+25+28+20+24+23+27+21+26) / 9=216 / 9=24 days.

[0068] The system reads the initial value λ_0 of the adaptive weight parameter λ from the parameter configuration table. In this project, λ_0 is set to 1.00. It should be noted that the physical meaning of the adaptive weight parameter λ is the global schedule correction factor, which defaults to 1, indicating that no additional correction is applied.

[0069] The expected construction period of node R3 is calculated using the formula Tu=T_avg×(1+η_u)×(2-γ_u)×λ. Tu = 24 × (1 + 0.701) × (2 - 0.963) × 1.00 = 24 × 1.701 × 1.037 ≈ 42.3 days, rounded up to 43 days.

[0070] When node R3 is actually completed, the closed-loop calibration terminal obtains the actual construction period Tu_re of the node from the construction log system through the integration interface. The construction log system pushes the node completion event to the closed-loop calibration terminal using the WebSocket protocol. The data message contains the node ID, actual start time, and actual end time. The actual construction period Tu_re is parsed to be 49 days. The period deviation ΔTu is calculated as Tu_re - Tu = 49 - 43 = 6 days, and the period deviation rate δ_u = ΔTu / Tu = 6 / 43 ≈ 0.1395. The system reads the preset periodic deviation rate tolerance threshold δ_th, which is set by the project chief engineer in the configuration interface; here, δ_th = 0.10. Since δ_3 = 0.1395 > 0.10, an anomaly detection is triggered.

[0071] The closed-loop correction terminal generates an anomaly report for the construction cycle. The report data structure includes: project number, node task name, expected construction period, actual construction period, deviation rate, exceedance threshold, and brief analysis guidelines for possible causes.

[0072] Anomaly reports are sent to the project manager and planning engineer accounts via email and system in-system messages. At the same time, the actual construction period and deviation information are written to the historical deviation database, which is stored in a MongoDB document database for traceability.

[0073] The logic for dynamically adjusting the adaptive weight parameters is executed immediately to obtain the current preset learning rate θ. The learning rate θ is set by the system administrator based on historical parameter tuning experience, and a value in the range of 0.2 to 0.5 can ensure stable convergence.

[0074] The adjustment method is λ_new=λ_old×(1+θ×δ_u)=1.00×(1+0.3×0.1229)=1.00×1.03687≈1.037. The closed-loop correction terminal writes the updated λ_new=1.037 into the system parameter table, overwriting the original value. When predicting the construction period of the basement structure in the next node task R4, the construction period prediction terminal will read λ=1.037 and participate in the calculation, thereby making the subsequent prediction automatic compensation system tend to have a positive bias.

[0075] If δ_u≤δ_th occurs, no abnormal report is generated. Instead, ΔTu is recorded in the corresponding record of that node in the historical period deviation database, and the adaptive weight parameter iterative adjustment is still performed, so that the system can continue to fine-tune even in non-abnormal situations.

[0076] This embodiment establishes the construction cycle prediction based on four factors: reference average construction period, increased difficulty, compressed material supply, and adaptive weights. Through real-time comparison and automatic adjustment of weight parameters by the closed-loop correction terminal, the prediction model has self-learning capabilities. As the project progresses, the prediction accuracy gradually improves, and the anomaly reporting mechanism ensures that major deviations are promptly addressed, enhancing the initiative and reliability of progress control.

[0077] Example 6

[0078] This embodiment is implemented based on the above embodiments, fully presenting the collaborative operation and adaptive iteration effects of the aforementioned embodiments, and further illustrating the system's performance under continuous multi-node tasks, as detailed below: Taking the target project Q as a residential project in a certain city as an example, its node task sequence R1 to R14 has been given in Example 4. When the system is initialized, the multi-dimensional matching end determines the reference project set CQ in the manner of Example 3 and stores it in the shared memory area. The construction difficulty assessment end, material supply control end, construction cycle prediction end and closed-loop correction end process each node task in turn.

[0079] During the earthwork excavation at node R1, a type of difficulty index was extracted from the construction design drawings and weighted by AHP to obtain Su=0.320. The average delay rate of the same node in the reference project was 0.08, so η_1=0.345. The theoretical earthwork volume is calculated to be Vu=1500 cubic meters, the reference loss rate is 0.02, the expected usage is YVu=1530, the actual purchase volume is GVu=1540 (due to sufficient previous purchases), and γ_1=1.0. The average construction period of the reference project is T_avg=8 days. The current λ=1.00. The expected construction period is calculated as Tu=8×(1+0.345)×(2-1.0)×1.00≈10.76 days. The actual completion time is 10 days. The deviation rate is δ_1≈-0.07. |δ_1|≤0.10. Only the deviation is recorded. λ is updated to 1.00×(1+0.3×(-0.07))=0.979.

[0080] In the foundation padding of node R2, Su=0.280, the average delay rate is 0.05, and η_2=0.294; Theoretical concrete quantity Vu=210, loss rate 0.03, expected YVu=216.3, actual purchased GVu=215, γ_2≈0.994; reference average construction period T_avg=6 days, predicted Tu=6×(1+0.294)×(2-0.994)×0.979≈7.69 days, actual completion time 7.5 days, deviation rate -0.025, λ updated to 0.979×(1+0.3×0.025)≈0.986.

[0081] The raft foundation at node R3 is operated as described in Examples 4 and 5, and λ has been updated to 1.037. At node R4, the basement structure is affected by the weight amplification of the previous node, and the predicted construction period is appropriately extended. However, if the actual construction is shortened due to management improvements, the deviation rate will be negative, and λ will be adjusted downward.

[0082] After 14 consecutive iterations of the task, λ eventually converges to around 1.08, giving the system a better prior basis for subsequent similar projects.

[0083] In multiple node tasks, the material supply timeliness rate γ_u was less than 1 on several occasions. The system completed the material replenishment scheduling under the early warning mechanism described in Example 4. The material gaps were filled in time, and there were no work stoppages due to insufficient materials. The construction cycle abnormality report was triggered in nodes R3 and R11. Based on this, the project manager re-examined the resource allocation plan of the corresponding nodes.

[0084] After the completion of the entire project, the historical project database was expanded to include all process data of the target project Q, including the final λ value and the actual construction period of each node, thus providing a richer reference benchmark for future target projects.

[0085] This embodiment demonstrates that, in the continuous operation mode throughout the entire project cycle, the system not only ensures the efficient advancement of single-node tasks through a closed-loop feedback mechanism, but also enables the prediction parameters of the entire system to adaptively optimize as the project progresses. It exhibits good robustness and dynamic tracking capabilities in complex construction environments, reducing the cumulative deviation between the schedule plan and the actual progress.

[0086] All data in the formulas described above have been calculated with dimensions removed. Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0087] The above description is merely an example and illustration of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0088] It should be stated that all user data collected in this application was collected with the user's consent and authorization. Furthermore, the uses of user data are legal and compliant, and the use and processing of user data comply with the relevant laws, regulations, and standards of the relevant regions.

Claims

1. An AI-based project progress management system, characterized in that, The system includes: The multi-dimensional matching end locks in the multi-dimensional feature parameters of the target project, constructs the multi-dimensional fusion features associated with the target project, generates a comprehensive feature vector, and locks in the comprehensive matching degree between each historical project and the target project from the historical project database based on the Euclidean distance between the comprehensive feature vectors, thus determining the reference project set. On the construction difficulty assessment side, a first-class difficulty index of the current node task of the target project is obtained based on the construction design drawings, and a weighted fusion is performed by combining the second-class difficulty index of each reference project to generate a standardized construction difficulty coefficient of the current node task. On the material supply control side, the theoretical material quantity of the current node task is determined, the expected material quantity is calculated by combining the material loss rate of the reference project during the same node task, and the material supply timeliness rate is locked in by combining the actual material procurement quantity of the target project for the current node task. On the construction cycle prediction end, the construction cycle of the reference project at the corresponding node task is extracted, and the standardized construction difficulty coefficient, material supply timeliness rate and adaptive weight parameters of the target project are combined to calculate the expected construction cycle of the current node task of the target project. The closed-loop correction end determines the actual construction period after the completion of the current node task of the target project in real time, locks the period deviation, generates a construction period anomaly report, and dynamically adjusts the adaptive weight parameters of the expected construction period.

2. The system according to claim 1, characterized in that, In the multidimensional matching process, the specific method for locking the multidimensional feature parameters of the target project and constructing the multidimensional fusion features associated with the target project is as follows: Determine the target project Q and its construction design drawings QM; Multidimensional feature parameters of the target project Q are extracted from QM, including project structure type, construction process complexity level, total building area, contract period, and climate classification of construction region. Among them, the construction process complexity level is obtained by technical personnel assessment. Standardize the multidimensional feature parameters to construct a multidimensional fusion feature vector FQ=(f1,f2,f3,f4,f5) for the target project Q, where f1 is the project structure type encoding value, which is obtained by converting the project structure type based on a preset project structure type mapping table; f2 is the normalized value of the construction process complexity level, which is obtained by mapping the construction process complexity level assessed by technical personnel to the interval [0,1]. f3 is the logarithmically standardized value of the total building area of ​​the project, obtained by performing Z-score standardization after natural logarithmic transformation; f4 is the relative ratio of contract duration, which is expressed as the ratio of the contract duration of the target project to the average duration of similar benchmark projects. The similar benchmark projects are projects pre-marked by technical personnel. f5 is the climate classification code value for the construction area, which is obtained by converting the climate classification of the construction area based on the preset climate zoning mapping table.

3. The system according to claim 2, characterized in that, In the multidimensional matching process, the specific method for generating the comprehensive feature vector is as follows: Differential weight coefficients are assigned to the feature parameters of each dimension in the multidimensional fusion feature vector FQ, namely w1, w2, w3, w4, w5. The differential weight coefficients are determined based on the entropy weight method analysis of the contribution of each dimension feature parameter in the historical project database to the construction cycle, and the sum of the differential weight coefficients is 1. By weighting, we obtain FQ'=(w1f1,w2f2,w3f3,w4f4,w5f5), which is labeled as the comprehensive feature vector FQ' of the target project Q.

4. The system according to claim 3, characterized in that, In the multi-dimensional matching process, based on Euclidean distance, the comprehensive matching degree between each historical project and the target project is determined from the historical project database. The specific method for determining the reference project set is as follows: Extract the engineering structure type of the target project Q, and denote it as QA; Using QA as the search criteria, a precise search is performed from the historical project database to extract candidate historical projects, which are then sorted in reverse chronological order to obtain the candidate historical project sequence Q1, Q2, ..., Qj, where j is the total number of candidate historical projects; Extract the comprehensive feature vector of each candidate historical project, and align the candidate historical project sequence to form the historical comprehensive feature vector FQ1', FQ2', ..., FQj'; Calculate the Euclidean distance between each historical comprehensive feature vector and the comprehensive feature vector FQ' of the target project Q in sequence to obtain the Euclidean distance sequence D1',D2',...,Dj'; Extract the operator's preset Euclidean distance threshold, filter j Euclidean distances, remove Euclidean distances greater than the threshold, and then extract the historical projects corresponding to the smallest α% Euclidean distances from the remaining Euclidean distances, denoted as the reference project set CQ={CQ1,CQ2,...,CQm}, where m is the total number of reference projects and α% is the preset percentage.

5. The system according to claim 4, characterized in that, The construction difficulty assessment module also includes the construction design drawing QM based on the target project Q, and the node tasks of the target project Q are determined according to the construction time sequence, denoted as the node task sequence R1, R2, ..., Ro, where o is the total number of node tasks; Determine the task Ru of the current node, 1 ≤ u ≤ o; Extract a type of difficulty index for the current node task Ru, including the number of job types per unit area, the number of processes per unit area, and the proportion of high-risk operations; Next, obtain the two types of difficulty indicators for m reference projects in the reference project set CQ={CQ1,CQ2,...,CQm} when they are on the same node task, specifically the average delay rate; The Analytic Hierarchy Process (AHP) is used to compare each difficulty index in a class of difficulty indices pairwise to determine the index weights, and then weighted and fused to generate the basic value of construction complexity, Su. The two types of difficulty indicators are integrated with the basic value of construction complexity Su to obtain the standardized construction difficulty coefficient η_u of the current node task Ru of the target project Q, where η_u = Su × (1 + average delay rate).

6. The system according to claim 5, characterized in that, In the material supply control aspect, the specific method for locking the timely supply rate of materials based on the actual procurement quantity of materials for the current node task of the joint target project is as follows: Based on the construction design drawing QM of the target project Q, extract the theoretical material quantity Vu required for the target project Q at the current node task Ru; Extract the material loss rate ρ of m reference projects in the reference project set CQ={CQ1,CQ2,...,CQm} when they are at the same node task; The expected material quantity YVu of the target project Q at the current node task Ru is calculated by YVu=Vu×(1+ρ); Obtain the actual material purchase quantity GVu of the target project Q at the current node task Ru, and calculate the material supply timeliness rate γ_u=min(1,GVu / YVu).

7. The system according to claim 6, characterized in that, In the material supply control mechanism, a material early warning is triggered when the material supply timeliness rate γ_u < 1. The material shortage amount ΔVu is calculated using ΔVu=YVu-GVu, and a corresponding material early warning report is generated to notify the operator to perform material replenishment. The replenishment amount is ΔVu+ΔX, where ΔX is the preset redundant material amount.

8. The system according to claim 6, characterized in that, In the construction cycle prediction module, the specific method for calculating the expected construction cycle of the current node task of the target project is as follows: Extract the actual construction period Ti of each reference project in the reference project set CQ={CQ_1,CQ_2,...,CQ_m} at the corresponding node task R_u, 1≤i≤m, and calculate the average construction period T_avg=(1 / m)×ΣT_i; Obtain the standardized construction difficulty coefficient η_u of the current node task R_u of the target project Q, and obtain the timely material supply rate γ_u; Extract the adaptive weight parameter λ, and the initial value of the adaptive weight parameter λ is a preset constant λ_0; The expected construction period Tu of the current node task R_u of the target project Q is calculated by Tu=T_avg×(1+η_u)×(2-γ_u)×λ.

9. The system according to claim 8, characterized in that, In the closed-loop correction stage, the specific method for locking the cycle deviation and generating a construction cycle anomaly report is as follows: Extract the actual construction period Tu_re after the completion of the current node task R_u of the target project Q; Extract the expected construction period Tu of the current node task R_u of the target project Q, calculate the period deviation ΔTu=Tu_re-Tu, and calculate the period deviation rate δ_u=ΔTu / Tu; Compare the preset period deviation rate tolerance threshold δ_th with the period deviation rate δ_u: If δ_u≤δ_th, record ΔTu in the record of the current node task R_u of the target project Q in the historical cycle deviation database; If δ_u > δ_th, a construction cycle anomaly report is generated, reminding operators that the actual construction cycle of the current node task R_u is abnormal.

10. The system according to claim 9, characterized in that, In the closed-loop correction stage, the specific methods for dynamically adjusting the adaptive weight parameters of the expected construction period include: The closed-loop correction end performs iterative adjustments based on the actual periodic deviation after each node task is completed. The adjustment method is λ_new=λ_old×(1+θ×deviation rate), where θ is the preset learning rate, λ_old is the adaptive weight parameter before iterative adjustment, and λ_new is the adaptive weight parameter after iterative adjustment.

Images