Construction progress management method and system based on improved BiLSTM and PERT three-point estimation

Abstract

The invention belongs to the technical field of construction progress management, and particularly relates to a construction progress management method and system based on improved BiLSTM and PERT three-point estimation, and the method comprises the steps: firstly obtaining a historical construction task progress data set, inputting the historical construction task progress data set into an improved BiLSTM model, and carrying out the dynamic prediction of the future construction task progress; then based on the predicted future construction task progress prediction result data set, the completion probability that the construction task is completed before the target time point is calculated through a PERT three-point estimation method, progress risk quantification is achieved, and finally construction progress management is achieved according to the completion probability that the construction task is completed before the target time point. It can be ensured that high-risk tasks obtain sufficient resource support, and then it is ensured that construction tasks can be completed on time. Dynamic prediction of future construction task progress and quantitative evaluation of progress risks are integrated, the efficiency and the intelligent degree of construction progress management are improved, and the method is suitable for progress management of large-scale construction projects.

Description

Technical Field

[0001] This invention belongs to the field of construction progress management technology, specifically relating to a construction progress management method and system based on improved BiLSTM and PERT three-point estimation. Background Technology

[0002] Construction schedule management is a crucial aspect of building construction. Traditional methods often face challenges such as missing data, inaccurate forecasts, and difficulties in adjusting task priorities. With the development of Building Information Modeling (BIM) and machine learning technologies, intelligent prediction and risk assessment of construction schedules have become new research directions. However, existing methods still have many shortcomings in practical applications, such as missing data, excessively long decision generation times, and low accuracy in schedule risk assessment in traditional construction schedule management.

[0003] The existing literature, "Research and Application of BIM-Based Subway Construction Progress Management Scheme," authored by Ye Mingzhu and published in *Modern Urban Rail Transit*, 2023, Issue 3, proposes a scheme for progress management applicable to urban rail transit general contracting projects, addressing the issues of poor timeliness and low visualization in subway construction general contracting projects. While this method achieves good accuracy, BIM-based progress management primarily relies on static models and plans, making it difficult to adapt to complex changes during construction in real time. Furthermore, this method lacks functions such as schedule prediction and risk assessment in progress management, resulting in a low level of intelligence in construction progress management. Summary of the Invention

[0004] The purpose of this invention is to address the aforementioned problems in the existing technology by providing a construction schedule management method and system based on improved BiLSTM and PERT three-point estimation, which integrates dynamic prediction of future construction task progress and quantitative assessment of schedule risks to improve the efficiency and intelligence of construction schedule management.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention provides a construction schedule management method based on improved BiLSTM and PERT three-point estimation, the construction schedule management method comprising:

[0007] The historical construction task progress time series data is obtained and preprocessed. Short-term and long-term features are extracted from the preprocessed historical construction task progress time series data to obtain the historical construction task progress dataset.

[0008] The historical construction task progress dataset is input into the improved BiLSTM model to predict the future construction task progress, and the future construction task progress prediction result dataset is obtained; the improved BiLSTM model refers to the BiLSTM model that uses the WMA algorithm to optimize the parameters.

[0009] Based on the dataset of future construction task progress prediction results, the probability of completing a construction task before the target time point is calculated using the PERT three-point estimation method. Construction progress management is then implemented based on the probability of completing a construction task before the target time point.

[0010] The WMA algorithm is an improved WMA algorithm that utilizes the leader-follower strategy and the adaptive migration strategy. The steps for optimizing the parameters of the BiLSTM model include:

[0011] S1. Humpback whale population initialization;

[0012] S2. Calculate the fitness value of all humpback whale individuals in the humpback whale population, sort all humpback whale individuals in the humpback whale population in descending order of fitness value, and take the humpback whale individual ranked first as the current best humpback whale.

[0013] S3. Select the top-ranking humpback whale population. One humpback whale individual is designated as the current leader, and the remaining humpback whale individuals are designated as the current followers. The average position of the current leader is calculated, and the positions of the leader and followers in the humpback whale population are updated based on the average position of the current leader. The formula for updating the position of the followers is as follows:

[0014] ;

[0015] In the above formula, This represents the updated position of the i-th humpback whale that is acting as a follower. The average position of current leaders; A vector representing the generation of random numbers; This represents the position of the i-th humpback whale that was a follower before the update. The position of the (i-1)th humpback whale that is a follower; This is the current optimal position for the humpback whale; It is a factor for adjusting the amplitude of variation; The product of Hadamard;

[0016] The formula for updating the leader's position is:

[0017] ;

[0018] In the above formula, The updated position of the j-th humpback whale that is the leader; The position of the j-th humpback whale that was the leader before the update; The position of the (j-1)th humpback whale that is the leader; , All 3D random variable; , These are the starting position vector and the ending position vector, respectively.

[0019] S4. Return to step S3 for iterative calculation. Stop iterating when the maximum number of iterations is reached, and output the optimal solution.

[0020] The calculation of the probability of completion of the construction task before the target time point using the PERT three-point estimation method includes:

[0021] Based on the dataset of future construction task progress predictions, the predicted completion time of the construction tasks is obtained. The expected completion time and standard deviation of the construction tasks are then calculated using the following formula:

[0022] ;

[0023] ;

[0024] In the above formula, The expected completion time for the construction task; An optimistic estimate of the time required to complete the construction task; The predicted completion time for the construction task; A pessimistic estimate of the time required to complete the construction task; The standard deviation of the construction period for the construction task;

[0025] The probability of completing a construction task before the target time point is calculated based on the expected completion time and standard deviation of the task. The formula for calculating the probability of completing a construction task before the target time point is as follows:

[0026] ;

[0027] In the above formula, This indicates the probability that the construction task will be completed before the target time point; The cumulative distribution function represents the standard normal distribution; This indicates the preset target time point.

[0028] Construction progress management is implemented based on the probability of completing construction tasks before the target time, including:

[0029] Calculate the priority factor for construction tasks based on the probability of completion before the target time point:

[0030] ;

[0031] In the above formula, This represents the priority factor for the t-th construction task;

[0032] Calculate the resource allocation ratio for construction tasks based on their priority factors:

[0033] ;

[0034] In the above formula, Let be the resource allocation ratio for the t-th construction task; This represents the priority factor for the i-th construction task; This represents the sum of priority factors for all construction tasks;

[0035] Construction progress management is based on the resource allocation ratio of construction tasks.

[0036] The preprocessing includes first using the DBSCAN algorithm to detect outliers in the construction progress data, then manually correcting the outliers and using the ARIMAX model to fill in the missing values, and finally using a generative adversarial network (GAN) to enhance the construction progress data after filling in the missing values.

[0037] Secondly, the present invention provides a construction schedule management system based on improved BiLSTM and PERT three-point estimation, the construction schedule management system comprising:

[0038] The data processing module is used to acquire historical construction task progress time series data and preprocess it. It performs short-term feature extraction and long-term feature extraction on the preprocessed historical construction task progress time series data to obtain historical construction task progress dataset.

[0039] The construction progress prediction module is used to input the historical construction task progress dataset into the improved BiLSTM model to predict the future construction task progress and obtain the future construction task progress prediction result dataset; the improved BiLSTM model refers to the BiLSTM model that uses the WMA algorithm to optimize the parameters.

[0040] The construction progress management module is used to calculate the probability of a construction task being completed before the target time point based on the dataset of future construction task progress prediction results using the PERT three-point estimation method, and to implement construction progress management based on the probability of the construction task being completed before the target time point.

[0041] The WMA algorithm is an improved WMA algorithm that utilizes the leader-follower strategy and the adaptive migration strategy. The construction progress prediction module optimizes the parameters of the BiLSTM model according to the following steps:

[0042] S1. Humpback whale population initialization;

[0043] S2. Calculate the fitness value of all humpback whale individuals in the humpback whale population, sort all humpback whale individuals in the humpback whale population in descending order of fitness value, and take the humpback whale individual ranked first as the current best humpback whale.

[0044] S3. Select the top-ranking humpback whale population. One humpback whale individual is designated as the current leader, and the remaining humpback whale individuals are designated as the current followers. The average position of the current leader is calculated, and the positions of the leader and followers in the humpback whale population are updated based on the average position of the current leader. The formula for updating the position of the followers is as follows:

[0045] ;

[0046] In the above formula, This represents the updated position of the i-th humpback whale that is acting as a follower. The average position of current leaders; A vector representing the generation of random numbers; This represents the position of the i-th humpback whale that was a follower before the update. The position of the (i-1)th humpback whale that is a follower; This is the current optimal position for the humpback whale; It is a factor for adjusting the amplitude of variation; The product of Hadamard;

[0047] The formula for updating the leader's position is:

[0048] ;

[0049] In the above formula, The updated position of the j-th humpback whale that is the leader; The position of the j-th humpback whale that was the leader before the update; The position of the (j-1)th humpback whale that is the leader; , All 3D random variable; , These are the starting position vector and the ending position vector, respectively.

[0050] S4. Return to step S3 for iterative calculation. Stop iterating when the maximum number of iterations is reached, and output the optimal solution.

[0051] The construction progress management module calculates the probability of completing the construction task before the target time point based on the following steps:

[0052] Based on the dataset of future construction task progress predictions, the predicted completion time of the construction tasks is obtained. The expected completion time and standard deviation of the construction tasks are then calculated using the following formula:

[0053] ;

[0054] ;

[0055] In the above formula, The expected completion time for the construction task; An optimistic estimate of the time required to complete the construction task; The predicted completion time for the construction task; A pessimistic estimate of the time required to complete the construction task; The standard deviation of the construction period for the construction task;

[0056] The probability of completing a construction task before the target time point is calculated based on the expected completion time and standard deviation of the task. The formula for calculating the probability of completing a construction task before the target time point is as follows:

[0057] ;

[0058] In the above formula, This indicates the probability that the construction task will be completed before the target time point; The cumulative distribution function represents the standard normal distribution; This indicates the preset target time point.

[0059] The construction progress management module implements construction progress management according to the following steps:

[0060] Calculate the priority factor for construction tasks based on the probability of completion before the target time point:

[0061] ;

[0062] In the above formula, This represents the priority factor for the t-th construction task;

[0063] Calculate the resource allocation ratio for construction tasks based on their priority factors:

[0064] ;

[0065] In the above formula, Let be the resource allocation ratio for the t-th construction task; This represents the priority factor for the i-th construction task; This represents the sum of priority factors for all construction tasks;

[0066] Construction progress management is based on the resource allocation ratio of construction tasks.

[0067] The data processing module performs preprocessing according to the following steps: first, it uses the DBSCAN algorithm to detect outliers in the construction progress data; then, it manually corrects the outliers and uses the ARIMAX model to fill in the missing values; finally, it uses a generative adversarial network (GAN) to enhance the construction progress data after filling in the missing values.

[0068] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0069] 1. The construction schedule management method based on improved BiLSTM and PERT three-point estimation described in this invention first uses an improved BiLSTM model to dynamically predict the progress of future construction tasks, obtaining a dataset of future construction task progress prediction results. Then, based on this dataset, the probability of completing a construction task before the target time point is calculated using the PERT three-point estimation method, thus quantifying schedule risk and improving the accuracy and controllability of construction schedule management. Finally, construction schedule management is achieved based on the probability of completing a construction task before the target time point, ensuring that high-risk tasks receive sufficient resource support and guaranteeing timely completion. This method integrates dynamic prediction of future construction task progress and quantitative assessment of schedule risk, improving the efficiency and intelligence of construction schedule management, and is suitable for schedule management of large-scale construction projects.

[0070] 2. The construction progress management method based on improved BiLSTM and PERT three-point estimation described in this invention uses the WMA algorithm to optimize the parameters of the BiLSTM model. The WMA algorithm is an improved WMA algorithm using the leader-follower strategy and the adaptive migration strategy. By balancing the search diversity and convergence speed through the leader-follower strategy and the adaptive migration strategy, it accelerates the convergence of the global optimum while avoiding getting trapped in local optima, thereby improving the prediction accuracy of the improved BiLSTM.

[0071] 3. The construction progress management method based on improved BiLSTM and PERT three-point estimation described in this invention preprocesses the acquired historical construction task progress time series data. The preprocessing involves first using the DBSCAN algorithm to detect outliers in the construction progress data, then manually correcting the outliers, and finally using the ARIMAX model to fill in missing values. Finally, a generative adversarial network (GAN) is used to augment the construction progress data after missing value filling. Through outlier correction, missing value filling, and data augmentation, high-quality data input can be provided for the subsequent improved BiLSTM, thereby further improving the prediction accuracy of the improved BiLSTM for future construction task progress. Attached Figure Description

[0072] Figure 1 This is a flowchart of the construction progress management method described in this invention.

[0073] Figure 2 This is a structural block diagram of the construction progress management system described in this invention.

[0074] Figure 3 This is a comparison chart of the construction progress before and after optimization using the construction progress management method described in this invention. Detailed Implementation

[0075] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0076] Example 1:

[0077] See Figure 1 A construction schedule management method based on improved BiLSTM and PERT three-point estimation is proposed, which proceeds in the following steps:

[0078] Step 1: Obtain historical construction task progress time series data and preprocess it. Then, extract short-term and long-term features from the preprocessed historical construction task progress time series data to obtain the historical construction task progress dataset.

[0079] Specifically, historical construction task progress time series data not only includes historical construction progress data at each time point, but also considers short-term fluctuation data, long-term variation data, and exogenous variable data. This data collectively provides rich input for the BiLSTM model, helping the model understand the dynamic changes in construction progress and improve the accuracy of construction progress prediction. Short-term fluctuation data captures short-term changes (daily, weekly, or monthly) within the construction project itself, driven by project manager decisions or detailed project schedule arrangements, such as short-term work arrangements, labor allocation, weather changes, task priorities, and equipment deployment. Long-term variation data captures periodic changes in the construction project, helping the model identify long-term fluctuations or trends in construction progress. This is typically related to the project's seasonal impacts and annual plans, such as seasonal variations, annual target progress, and long-term trends. Seasonal variations refer to data that fluctuates regularly with the seasons, usually manifesting as similar changes occurring within the same time period each year. For construction projects, seasonal variations mainly include the following three aspects: Weather impacts: For example, some regions experience heavy rainfall in winter and high temperatures in summer, and construction progress may be affected by these seasonal weather changes. Winter's cold weather may slow construction progress, while summer's high temperatures or heavy rains may affect construction. Operating conditions vary by season; for example, cold winters may be unsuitable for some outdoor operations, while summer may affect concrete setting time. Resource availability is also a factor; seasonal changes can alter resource availability, such as differences in the difficulty and cost of transporting certain raw materials in different seasons. Long-term trends refer to the continuous changes in construction progress or overall project performance over a longer time span (e.g., several years or more). These trends typically reflect the long-term development and evolution of the project. Long-term trends mainly include the following three points: Project target schedule: As the project progresses, there may be phased schedule targets, which are usually long-term goals and will gradually change over time. Technological advancements: With the application of new technologies and processes, construction progress and efficiency may continuously improve over the long term. The impact of economic factors: Long-term economic development and increases or decreases in capital investment may have a long-term impact on construction progress, such as budget increases or decreases and changes in equipment procurement cycles. Exogenous variables refer to external factors that affect construction progress but do not originate directly from within the construction project. These variables have a significant impact on construction progress and are usually uncontrollable, such as weather changes, labor allocation, equipment deployment, and holidays. Although some exogenous variables (such as weather and labor) may appear as short-term fluctuation data, their emphasis differs in the model, and therefore they can be interpreted through different modeling methods in the model design.

[0080] Since short-term fluctuation data and long-term change data are mainly used to help capture the short-term and long-term dynamic changes in construction progress, while exogenous variable data are used to supplement external factors affecting construction progress, short-term fluctuation data and long-term change data are mainly used in the BiLSTM model for feature extraction and dynamic prediction, while exogenous variable data are mainly used in the ARIMAX model to capture more external influencing factors, thereby improving prediction accuracy.

[0081] Specifically, in construction progress management, data preprocessing is fundamental to ensuring the accuracy of subsequent analysis and modeling. The preprocessing includes: first, using the DBSCAN algorithm to detect outliers in the construction progress data, treating them as anomalies, and manually correcting them; then, performing a differencing operation on the manually corrected construction progress data to convert non-stationary sequences into stationary sequences to ensure data stationarity; finally, using the ARIMAX model to impute missing values ​​in the stationary construction progress data, thereby ensuring data quality. The ARIMAX model, by combining AR and MA components and exogenous variables, establishes a dynamic time series model to predict and impute missing values ​​in the construction progress data. By introducing external variables, the ARIMAX model can capture more factors affecting the time series, thereby improving prediction accuracy. The expression for the ARIMAX model is:

[0082] ;

[0083] In the above formula, is the target variable, i.e., the construction progress data at time t; p is the autoregression order; q is the moving average order; for Construction progress data at any given time; These are the autoregressive coefficients; for Random errors at time points are used to capture the portion that the model fails to explain; The moving average coefficient; This represents the data for the k-th exogenous variable at time t. This represents the coefficient corresponding to the k-th exogenous variable, used to quantify the degree of influence of that exogenous variable on the target variable; The error term at time t; For constant terms;

[0084] Finally, a Generative Adversarial Network (GAN) is used to augment the construction progress data after missing value imputation, thereby expanding the data scale. The GAN generates virtual samples through a generator G and distinguishes virtual samples from real samples through a discriminator D. Through adversarial training between the generator G and the discriminator D, the model generates increasingly realistic samples. The scale of the augmented construction progress data is [data missing]. ,in The scale of the historical construction task progress time series data before data augmentation. To increase the proportion.

[0085] Specifically, after extracting short-term and long-term features from the preprocessed historical construction task progress time series data, the data is fused to obtain a historical construction task progress dataset. The fusion formula is as follows:

[0086] ;

[0087] In the above formula, To integrate the obtained historical construction task progress dataset; This is preprocessed historical construction task progress time series data; The short-term features obtained by extracting short-term features from the preprocessed historical construction task progress time series data are based on the rolling window method. The long-term features are obtained by extracting long-term features from the preprocessed historical construction task progress time series data. The long-term feature extraction is based on seasonal decomposition and Fourier transform.

[0088] Step 2: Input the historical construction task progress dataset into the improved BiLSTM model to predict the future construction task progress and obtain the future construction task progress prediction result dataset.

[0089] Specifically, the improved BiLSTM model includes parameter optimization using the WMA algorithm. The WMA algorithm is an improved version of the WMA algorithm that utilizes a leader-follower strategy and an adaptive migration strategy. The parameter optimization steps for the BiLSTM model include:

[0090] S1. Humpback whale population initialization: randomly generated. For each humpback whale individual, the search space is determined according to the following formula. Internal initialization:

[0091] ;

[0092] In the above formula, This represents the initial position of the i-th humpback whale individual in the humpback whale population; , These are the starting position vector and the ending position vector, respectively.

[0093] S2. Calculate the fitness value of all humpback whale individuals in the humpback whale population. Sort all humpback whale individuals in the population in descending order of fitness value, and take the humpback whale with the highest fitness value as the current best humpback whale. The formula for calculating the fitness value of a humpback whale individual is as follows:

[0094] ;

[0095] In the above formula, This represents the fitness value of the current humpback whale individual; This represents the loss function value of the BiLSTM model based on the model parameters corresponding to the current humpback whale individual; the smaller the loss function value, the larger the fitness value, thus allowing a better humpback whale individual to be selected as the leader or the best individual; the loss function value can be either mean squared error or cross-entropy.

[0096] S3. Select the top-ranking humpback whale population. One humpback whale individual is designated as the current leader, and the remaining humpback whale individuals are designated as the current followers. The average position of the current leader is calculated, and the positions of the leader and followers in the humpback whale population are updated based on the average position of the current leader. The formula for updating the position of the followers is as follows:

[0097] ;

[0098] In the above formula, This represents the updated position of the i-th humpback whale that is acting as a follower. The average position of current leaders; A vector representing the generation of random numbers; This represents the position of the i-th humpback whale that was a follower before the update. The position of the (i-1)th humpback whale that is a follower; This is the current optimal position for the humpback whale; It is a factor for adjusting the amplitude of variation; The product of Hadamard;

[0099] In the follower's position update formula This represents the average position of the leader group. This direction represents the average position of experienced individuals in the humpback whale population. It is used to guide each individual to move according to the overall trend of the population, thereby helping the algorithm converge to a better solution and enhancing the stability and convergence of the search process. The term represents the inertia term, meaning that each individual will change its position based on the position of the previous individual. and current location The individual adjusts its position by considering the difference between the previous position and the current position, so that the individual continues to move in the current direction. By considering the difference between the previous position and the current position, the individual's movement is not too random, which is conducive to the stable convergence of the algorithm. The term represents the globally optimal guiding direction for an individual based on the current globally optimal solution. and the average position of current leaders By adjusting its position according to the differences between individuals, and by guiding individuals toward the global optimal position and the group optimal position, the process of finding the global optimal solution is accelerated, which helps the algorithm converge more quickly. The term represents the Gaussian perturbation term, which, by introducing a normal probability density function, simulates a stochastic guiding force, causing individuals to react according to the current optimal solution. With current location The distance between them is subject to a certain perturbation. By adding a Gaussian perturbation term, individuals can escape local optima during global search, helping the algorithm avoid premature convergence and increasing its exploration capability. This can be achieved by adjusting the mutation amplitude adjustment factor. It can control the magnitude of the perturbation, thereby affecting the search range and the degree of exploration;

[0100] The formula for updating the leader's position is:

[0101] ;

[0102] In the above formula, The updated position of the j-th humpback whale that is the leader; The position of the j-th humpback whale that was the leader before the update; The position of the (j-1)th humpback whale that is the leader; , All 3D random variable;

[0103] In the leader's position update formula The term indicates the position of the previous individual. and current location The differences between them are used to simulate the random perturbation of individuals in the search space, so as to guide individuals to adjust according to their historical positions, increase the randomness and diversity of the search, and avoid premature convergence to local optima. The term represents the adjustment made by an individual along the direction of its starting position, the degree of which is determined by a random variable. control, The term represents the individual's direction of migration. Adjustments are made, and the degree of adjustment is determined by the random variable. With control These two terms ensure that individuals are affected by randomness when updating their positions, changing the pace of their movement, thereby increasing the diversity of the search space and avoiding premature convergence.

[0104] S4. Return to step S3 for iterative calculation. Stop iterating when the maximum number of iterations is reached, and output the optimal solution.

[0105] To improve the generalization ability of BiLSTM models and prevent overfitting, a Dropout layer is incorporated into the BiLSTM model. Combining the Dropout layer with the BiLSTM model allows a portion of neurons to be randomly dropped during each training cycle. This prevents the neural network from over-relying on the features of certain neurons, thereby reducing overfitting and improving the model's predictive ability on unseen data. The Dropout layer is applied not only between the input and output of each layer of the BiLSTM model but also during the state update process of LSTM units. Specifically, it includes:

[0106] (1) Input gate: The input gate is used to control how the current input affects the current cell state. Its expression is:

[0107] ;

[0108] In the above formula, The output of the input gate represents the degree of influence of the current input on the cell state; It is the weight matrix of the input gate, which controls the input data. With hidden state Relationship, input data It is the input data at the current time step (after standardization) ); It is the input gate and the hidden state of the previous time step. The weight matrix; It is the bias term of the input gate; It is the sigmoid activation function, which controls the activation value of the input data, ranging from [0, 1], and determines the degree to which the input information is preserved; It is the mask of the Dropout layer in the input gate, used to randomly discard a portion of the neuron's output to reduce model overfitting;

[0109] (2) Forget gate: The forget gate is used to determine the cell state at the previous time step. Which information will be forgotten, expressed as:

[0110] ;

[0111] In the above formula, It is the output of the forget gate, representing the proportion of information that has been forgotten in the current cell state; It is the weight matrix of the forget gate; It is the forgetting gate and the hidden state of the previous time step. The weight matrix; It is the bias term of the forget gate; It is the mask of the Dropout layer in the forget gate, which controls the random dropping of the output of the forget gate;

[0112] (3) Candidate cell state: The candidate cell state determines the combination of the current input and the previous cell state, and its expression is:

[0113] ;

[0114] In the above formula, It represents the candidate cell state, indicating the candidate value of the cell state at the current moment; It is the weight matrix of cell states; It is the cell state and the hidden state of the previous time step. The weight matrix; It is a bias term for cell state; It is the mask of the Dropout layer in the candidate cell state, which controls the random dropping of the candidate cell state output;

[0115] (4) Cell state: The current cell state is the result of the forget gate determining which historical information is retained and the input gate determining which new information is written into the cell state. Its expression is:

[0116] ;

[0117] In the above formula, It is the cell state at the current time step, which contains long-term memory information at the current moment; The output of the forget gate determines how much information is forgotten from the previous time step; It represents the cell state at the previous time step; The output of the input gate determines the degree of influence of the current input. This is the current state of the candidate cells;

[0118] (5) Output gate: The output gate determines the influence of the current cell state on the current output, and its expression is:

[0119] ;

[0120] In the above formula, It is the output of the output gate, representing the output of the cell state at the current moment; It is the weight matrix of the output gate; It is the output gate and the hidden state of the previous time step. The weight matrix; It is the bias term of the output gate; It is the mask of the dropout layer in the D output gate, which controls the random dropping of the output gate's output;

[0121] (6) Final output: Final output The current cell state The nonlinear transformation of is expressed as:

[0122] ;

[0123] In the above formula, It is the hidden state of the current time step, and it is the output of the current time step.

[0124] Finally, the trained Bi-LSTM model will output a dataset of predictions for future construction task schedules. :

[0125] ;

[0126] In the above formula, This represents the predicted progress of the construction task at the i-th time step.

[0127] The dataset of future construction task progress prediction results It not only filled in the missing parts of the original data, but also predicted the progress of future construction tasks, providing high-quality input data for subsequent completion probability calculations.

[0128] Step 3: Based on the dataset of future construction task progress prediction results, calculate the probability of completion of the construction task before the target time point using the PERT three-point estimation method, and realize construction progress management based on the probability of completion of the construction task before the target time point.

[0129] Specifically, the calculation of the probability of completion of the construction task before the target time point using the PERT three-point estimation method includes the following steps:

[0130] First, based on the dataset of future construction task progress predictions, the predicted completion time of the construction tasks is obtained. Then, the expected completion time and standard deviation of the construction tasks are calculated using the following formula:

[0131] ;

[0132] ;

[0133] In the above formula, The expected completion time for the construction task; In construction schedule management, the optimistic estimated time for completing a construction task refers to the shortest time to complete the task under optimal conditions. It is usually assessed by experts based on project experience and the actual situation of the project, taking into account favorable factors such as weather conditions, resource supply, and team efficiency. The predicted completion time for the construction task; The pessimistic estimated time for completing a construction task is the longest time to complete the task under the most unfavorable conditions in construction schedule management. It is also assessed by experts based on project experience and the actual situation of the project. The assessment takes into account delay factors such as material delays, labor shortages, equipment failures, bad weather, resource constraints, policy changes, and supply chain problems. The standard deviation of the construction period is used to reflect the fluctuation in the completion time of the task.

[0134] Then, based on the expected completion time and standard deviation of the construction task, the probability of completing the construction task before the target time point is calculated. The probability of completing the construction task before the target time point is calculated using the cumulative distribution function of the normal distribution, and the calculation formula is as follows:

[0135] ;

[0136] In the above formula, This indicates the probability that the construction task will be completed before the target time point; The cumulative distribution function represents the standard normal distribution; This indicates the preset target time point.

[0137] Assessing the probability of completing construction tasks before the target deadline helps evaluate the schedule risk of each task in the project, thus providing project managers with a quantitative basis for whether tasks can be completed on schedule.

[0138] Specifically, construction progress management is implemented based on the probability of completing construction tasks before the target time, including:

[0139] First, calculate the priority factor of the construction task based on the probability of completion before the target time point:

[0140] ;

[0141] In the above formula, This represents the priority factor for the t-th construction task;

[0142] Then, the resource allocation ratio for the construction tasks is calculated based on the priority factor of the construction tasks:

[0143] ;

[0144] In the above formula, Let be the resource allocation ratio for the t-th construction task; This represents the priority factor for the i-th construction task; This represents the sum of priority factors for all construction tasks;

[0145] Finally, construction progress management is carried out based on the resource allocation ratio of construction tasks, and a construction progress management plan is output. Priority factors are used to adjust task priorities and optimize resource allocation to ensure that high-risk tasks receive sufficient resource support, thereby improving the reliability of the overall project progress.

[0146] Performance verification:

[0147] The construction progress management method described in this invention was applied to the entire lifecycle of a 110kV substation construction project for simulation calculations. The optimized construction progress management scheme was compared with the unoptimized one, and the results are as follows: Figure 3 As shown, in the comparison of construction progress over ten months, the construction progress after optimization by the method described in this invention was 8.2% higher than the construction progress before optimization, which proves the effectiveness of the method described in this invention.

[0148] Example 2:

[0149] See Figure 2 A construction progress management system based on improved BiLSTM and PERT three-point estimation is disclosed. The system includes a data processing module, a construction progress prediction module, and a construction progress management module. The data processing module acquires and preprocesses historical construction task progress time series data, performing short-term and long-term feature extraction on the preprocessed data to obtain a historical construction task progress dataset. Specifically, the data processing module performs preprocessing according to the following steps: first, it uses the DBSCAN algorithm to detect outliers in the construction progress data; after manual correction, it uses the ARIMAX model to fill in missing values; finally, it uses a Generative Adversarial Network (GAN) to enhance the data after missing value filling. The construction progress prediction module inputs the historical construction task progress dataset into the improved BiLSTM model to predict future construction task progress, obtaining a future construction task progress prediction result dataset. The improved BiLSTM model refers to the BiLSTM model with parameter optimization using the WMA algorithm. Specifically, the WMA algorithm is an improved version using a leader-follower strategy and an adaptive migration strategy. The construction progress prediction module optimizes the parameters of the BiLSTM model according to the following steps:

[0150] S1. Humpback whale population initialization;

[0151] S2. Calculate the fitness value of all humpback whale individuals in the humpback whale population, sort all humpback whale individuals in the humpback whale population in descending order of fitness value, and take the humpback whale individual ranked first as the current best humpback whale.

[0152] S3. Select the top-ranking humpback whale population. One humpback whale individual is designated as the current leader, and the remaining humpback whale individuals are designated as the current followers. The average position of the current leader is calculated, and the positions of the leader and followers in the humpback whale population are updated based on the average position of the current leader. The formula for updating the position of the followers is as follows:

[0153] ;

[0154] In the above formula, This represents the updated position of the i-th humpback whale that is acting as a follower. The average position of current leaders; A vector representing the generation of random numbers; This represents the position of the i-th humpback whale that was a follower before the update. The position of the (i-1)th humpback whale that is a follower; This is the current optimal position for the humpback whale; It is a factor for adjusting the amplitude of variation; The product of Hadamard;

[0155] The formula for updating the leader's position is:

[0156] ;

[0157] In the above formula, The updated position of the j-th humpback whale that is the leader; The position of the j-th humpback whale that was the leader before the update; The position of the (j-1)th humpback whale that is the leader; , All 3D random variable; , These are the starting position vector and the ending position vector, respectively.

[0158] S4. Return to step S3 for iterative calculation. Stop iterating when the maximum number of iterations is reached, and output the optimal solution.

[0159] The construction progress management module calculates the probability of completing the construction task before the target time point based on the following steps:

[0160] Based on the dataset of future construction task progress predictions, the predicted completion time of the construction tasks is obtained. The expected completion time and standard deviation of the construction tasks are then calculated using the following formula:

[0161] ;

[0162] ;

[0163] In the above formula, The expected completion time for the construction task; An optimistic estimate of the time required to complete the construction task; The predicted completion time for the construction task; A pessimistic estimate of the time required to complete the construction task; The standard deviation of the construction period for the construction task;

[0164] The probability of completing a construction task before the target time point is calculated based on the expected completion time and standard deviation of the task. The formula for calculating the probability of completing a construction task before the target time point is as follows:

[0165] ;

[0166] In the above formula, This indicates the probability that the construction task will be completed before the target time point; The cumulative distribution function represents the standard normal distribution; Indicates the preset target time point;

[0167] The construction progress management module is used to calculate the probability of completing a construction task before the target time point using the PERT three-point estimation method based on the dataset of future construction task progress prediction results. Construction progress management is then implemented based on this probability. Specifically, the construction progress management module implements construction progress management according to the following steps:

[0168] Calculate the priority factor for construction tasks based on the probability of completion before the target time point:

[0169] ;

[0170] In the above formula, This represents the priority factor for the t-th construction task;

[0171] Calculate the resource allocation ratio for construction tasks based on their priority factors:

[0172] ;

[0173] In the above formula, Let be the resource allocation ratio for the t-th construction task; This represents the priority factor for the i-th construction task; This represents the sum of priority factors for all construction tasks;

[0174] Construction progress management is based on the resource allocation ratio of construction tasks.

[0175] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0176] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0177] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0178] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A construction schedule management method based on improved BiLSTM and PERT three-point estimation, characterized by: The construction progress management method includes: The historical construction task progress time series data is obtained and preprocessed. Short-term and long-term features are extracted from the preprocessed historical construction task progress time series data to obtain the historical construction task progress dataset. The historical construction task progress dataset is input into the improved BiLSTM model to predict the future construction task progress, and the future construction task progress prediction result dataset is obtained; the improved BiLSTM model refers to the BiLSTM model that uses the WMA algorithm to optimize the parameters. Based on the dataset of future construction task progress prediction results, the probability of completing a construction task before the target time point is calculated using the PERT three-point estimation method. Construction progress management is then implemented based on the probability of completing a construction task before the target time point.

2. The construction progress management method based on improved BiLSTM and PERT three-point estimation according to claim 1, characterized in that: The WMA algorithm is an improved WMA algorithm that utilizes the leader-follower strategy and the adaptive migration strategy. The steps for optimizing the parameters of the BiLSTM model include: S1. Humpback whale population initialization; S2. Calculate the fitness value of all humpback whale individuals in the humpback whale population, sort all humpback whale individuals in the humpback whale population in descending order of fitness value, and take the humpback whale individual ranked first as the current best humpback whale. S3. Select the top-ranking humpback whale population. One humpback whale individual is designated as the current leader, and the remaining humpback whale individuals are designated as the current followers. The average position of the current leader is calculated, and the positions of the leader and followers in the humpback whale population are updated based on the average position of the current leader. The formula for updating the position of the followers is as follows: ； In the above formula, This represents the updated position of the i-th humpback whale that is acting as a follower. The average position of current leaders; A vector representing the generation of random numbers; This represents the position of the i-th humpback whale that was a follower before the update. The position of the (i-1)th humpback whale that is a follower; This is the current optimal position for the humpback whale; It is a factor for adjusting the amplitude of variation; The product of Hadamard; The formula for updating the leader's position is: ； In the above formula, The updated position of the j-th humpback whale that is the leader; The position of the j-th humpback whale that was the leader before the update; The position of the (j-1)th humpback whale that is the leader; , All 3D random variable; , These are the starting position vector and the ending position vector, respectively. S4. Return to S3 for iterative calculation. Stop iterating when the maximum number of iterations is reached, and output the optimal solution.

3. The construction schedule management method based on improved BiLSTM and PERT three-point estimation according to claim 1 or 2, characterized in that: The calculation of the probability of completion of the construction task before the target time point using the PERT three-point estimation method includes: Based on the dataset of future construction task progress predictions, the predicted completion time of the construction tasks is obtained. The expected completion time and standard deviation of the construction tasks are then calculated using the following formula: ； ； In the above formula, The expected completion time for the construction task; An optimistic estimate of the time required to complete the construction task; The predicted completion time for the construction task; A pessimistic estimate of the time required to complete the construction task; The standard deviation of the construction period for the construction task; The probability of completing a construction task before the target time point is calculated based on the expected completion time and standard deviation of the task. The formula for calculating the probability of completing a construction task before the target time point is as follows: ； In the above formula, This indicates the probability that the construction task will be completed before the target time point; The cumulative distribution function represents the standard normal distribution; This indicates the preset target time point.

4. The construction progress management method based on improved BiLSTM and PERT three-point estimation according to claim 3, characterized in that: Construction progress management is implemented based on the probability of completing construction tasks before the target time, including: Calculate the priority factor for construction tasks based on the probability of completion before the target time point: ； In the above formula, This represents the priority factor for the t-th construction task; Calculate the resource allocation ratio for construction tasks based on their priority factors: ； In the above formula, Let be the resource allocation ratio for the t-th construction task; This represents the priority factor for the i-th construction task; This represents the sum of priority factors for all construction tasks; Construction progress management is based on the resource allocation ratio of construction tasks.

5. The construction schedule management method based on improved BiLSTM and PERT three-point estimation according to claim 1 or 2, characterized in that: The preprocessing includes first using the DBSCAN algorithm to detect outliers in the construction progress data, then manually correcting the outliers and using the ARIMAX model to fill in the missing values, and finally using a generative adversarial network (GAN) to enhance the construction progress data after filling in the missing values.

6. A construction schedule management system based on improved BiLSTM and PERT three-point estimation, characterized in that: The construction progress management system includes: The data processing module is used to acquire historical construction task progress time series data and preprocess it. It performs short-term feature extraction and long-term feature extraction on the preprocessed historical construction task progress time series data to obtain historical construction task progress dataset. The construction progress prediction module is used to input the historical construction task progress dataset into the improved BiLSTM model to predict the future construction task progress and obtain the future construction task progress prediction result dataset; the improved BiLSTM model refers to the BiLSTM model that uses the WMA algorithm to optimize the parameters. The construction progress management module is used to calculate the probability of a construction task being completed before the target time point based on the dataset of future construction task progress prediction results using the PERT three-point estimation method, and to implement construction progress management based on the probability of the construction task being completed before the target time point.

7. The construction progress management system based on improved BiLSTM and PERT three-point estimation according to claim 6, characterized in that: The WMA algorithm is an improved WMA algorithm that utilizes the leader-follower strategy and the adaptive migration strategy. The construction progress prediction module optimizes the parameters of the BiLSTM model according to the following steps: S1. Humpback whale population initialization; S2. Calculate the fitness value of all humpback whale individuals in the humpback whale population, sort all humpback whale individuals in the humpback whale population in descending order of fitness value, and take the humpback whale individual ranked first as the current best humpback whale. S3. Select the top-ranking humpback whale population. One humpback whale individual is designated as the current leader, and the remaining humpback whale individuals are designated as the current followers. The average position of the current leader is calculated, and the positions of the leader and followers in the humpback whale population are updated based on the average position of the current leader. The formula for updating the position of the followers is as follows: ； In the above formula, This represents the updated position of the i-th humpback whale that is acting as a follower. The average position of current leaders; A vector representing the generation of random numbers; This represents the position of the i-th humpback whale that was a follower before the update. The position of the (i-1)th humpback whale that is a follower; This is the current optimal position for the humpback whale; It is a factor for adjusting the amplitude of variation; The product of Hadamard; The formula for updating the leader's position is: ； In the above formula, The updated position of the j-th humpback whale that is the leader; The position of the j-th humpback whale that was the leader before the update; The position of the (j-1)th humpback whale that is the leader; , All 3D random variable; , These are the starting position vector and the ending position vector, respectively. S4. Return to S3 for iterative calculation. Stop iterating when the maximum number of iterations is reached, and output the optimal solution.

8. The construction progress management system based on improved BiLSTM and PERT three-point estimation according to claim 6 or 7, characterized in that: The construction progress management module calculates the probability of completing the construction task before the target time point based on the following steps: Based on the dataset of future construction task progress predictions, the predicted completion time of the construction tasks is obtained. The expected completion time and standard deviation of the construction tasks are then calculated using the following formula: ； ； In the above formula, The expected completion time for the construction task; An optimistic estimate of the time required to complete the construction task; The predicted completion time for the construction task; A pessimistic estimate of the time required to complete the construction task; The standard deviation of the construction period for the construction task; The probability of completing a construction task before the target time point is calculated based on the expected completion time and standard deviation of the task. The formula for calculating the probability of completing a construction task before the target time point is as follows: ； In the above formula, This indicates the probability that the construction task will be completed before the target time point; The cumulative distribution function represents the standard normal distribution; This indicates the preset target time point.

9. The construction progress management system based on improved BiLSTM and PERT three-point estimation according to claim 8, characterized in that: The construction progress management module implements construction progress management according to the following steps: Calculate the priority factor for construction tasks based on the probability of completion before the target time point: ； In the above formula, This represents the priority factor for the t-th construction task; Calculate the resource allocation ratio for construction tasks based on their priority factors: ； In the above formula, Let be the resource allocation ratio for the t-th construction task; This represents the priority factor for the i-th construction task; This represents the sum of priority factors for all construction tasks; Construction progress management is based on the resource allocation ratio of construction tasks.

10. The construction progress management system based on improved BiLSTM and PERT three-point estimation according to claim 6 or 7, characterized in that: The data processing module performs preprocessing according to the following steps: first, it uses the DBSCAN algorithm to detect outliers in the construction progress data; then, it manually corrects the outliers and uses the ARIMAX model to fill in the missing values; finally, it uses a generative adversarial network (GAN) to enhance the construction progress data after filling in the missing values.

Images