Engineering intelligent cost process control system and control method

By building an engineering intelligent cost process control system, the problem that traditional cost control relies on manual experience is solved, dynamic prediction and risk management are realized, the accuracy and targeted control of cost prediction are improved, and the controllability and economical cost is ensured. sex.

CN119027161APending Publication Date: 2024-11-26JIANGXI JIANGZI ENG CONSULTING CO LTD

Patent Information

Application Number
CN202411154258.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-22
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The traditional cost control process is highly dependent on manual experience and lacks scientific model support, which leads to inaccurate cost prediction, ineffective evaluation of cost risks in the construction stage, difficulty in comprehensively assessing the superposition effect of influencing factors, and lacks abnormal situation identification and optimization scheduling mechanisms, resulting in cost control Lag and inefficiency.

Method used

An engineering intelligent cost process control system is designed, including data acquisition and analysis module, risk and weight fitting module and real-time adjustment module. By constructing a cost prediction model, the cost deviation and risk level are determined, the cost control weight is dynamically adjusted, and the cost control weight is carried out to separate the cost prediction model. Group value detection and optimization scheduling to achieve refined cost control.

Benefits of technology

Realize dynamic cost prediction, improve the accuracy of cost prediction, introduce risk level classification management, formulate differentiated control strategies, dynamically adjust cost according to various influencing factors, identify and optimize abnormal situations, and ensure the controllability and economicality of cost. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027161A_ABST
    Figure CN119027161A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of engineering cost, and discloses an engineering intelligent cost process control system and control method. Comprising the following steps: constructing a cost prediction model for obtaining a theoretical cost value, collecting related data of an engineering project in n time periods, and determining cost deviation degrees of the n time periods according to the related data and the cost prediction model; according to the cost deviation degrees of the n time periods, cost risk levels of the n time periods are determined by adopting a clustering algorithm, and cost control weight coefficients of the n time periods are determined based on the cost risk levels; adjusting the theoretical cost value by using the cost control weight coefficient to obtain target cost values of n time periods; and outlier detection is carried out on the deviation between the actual cost of the n time periods and the target cost value, an abnormal cost time period is determined, and cost optimization scheduling is carried out on the abnormal cost time period, so that the cost risk is effectively prevented and solved, and the controllability and economy of the project cost are ensured.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of engineering cost technology, and more specifically, to an engineering intelligent cost process control system and a control method. Background Art

[0002] The patent with application publication number CN111476537A discloses a BIM-based engineering cost dynamic control system and method, including: a 5D BIM model building subsystem, a construction site information acquisition subsystem, a data processing subsystem, a 5D BIM model updating subsystem, an engineering cost monitoring subsystem, an engineering cost dynamic management subsystem and a system management subsystem; it can realize dynamic monitoring of the engineering cost of the construction project during the construction phase, so that the engineering management personnel can timely and accurately understand the project progress and cost control of the construction project, which is conducive to the engineering management personnel to adjust the construction plan in time when the engineering cost of the construction project exceeds the deviation allowed by the engineering cost plan, so that the engineering cost to be incurred is within the deviation allowed by the engineering cost plan, thereby ensuring that the entire construction project can be carried out smoothly.

[0003] However, the traditional cost control process is highly dependent on manual experience and lacks scientific theoretical model support. In actual engineering projects, cost control personnel usually rely on their own accumulated experience data and subjective judgment to estimate the project cost. This method has great uncertainty and subjectivity, and it is difficult to accurately predict and control the cost. Secondly, the existing methods cannot effectively evaluate and quantify the cost risk level of different construction stages. The entire construction process of an engineering project is usually divided into multiple stages, and the degree of cost risk in each stage is different. However, the existing methods lack quantitative analysis of this and cannot take different cost control measures in a targeted manner, resulting in lag and low efficiency in cost control. Furthermore, the influencing factors considered by the existing cost control methods are relatively single, making it difficult to comprehensively evaluate the cost situation. It is impossible to comprehensively analyze the superposition effect of various influencing factors, resulting in low accuracy of cost control. Finally, the existing methods lack an effective identification and processing mechanism for abnormal cost situations. In the actual engineering process, there may be abnormal deviations of the cost of some special stages from the expected value, but the existing methods are difficult to detect these anomalies in time, and lack corresponding optimization and scheduling methods, and cannot correct abnormal cost problems in time.

[0004] In view of this, the present invention proposes an engineering intelligent cost process control system and control method to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions: an intelligent engineering cost process control system, comprising: a data acquisition and analysis module, used to construct a cost prediction model for obtaining theoretical cost values, collect relevant data of the engineering project in n time periods, and determine the cost deviation of the n time periods according to the relevant data and the cost prediction model;

[0006] The risk and weight fitting module is used to determine the cost risk level of n time periods according to the cost deviation of n time periods, and determine the cost control weight coefficient of n time periods based on the cost risk level;

[0007] A real-time adjustment module is used to adjust the theoretical construction value using the cost control weight coefficient to obtain the target construction value for n time periods;

[0008] The analysis and scheduling module is used to perform outlier detection on the deviation between the actual cost and the target cost in n time periods, determine the abnormal cost time period, and take cost optimization scheduling for the abnormal cost time period; each module is connected by wired and / or wireless means.

[0009] Furthermore, the relevant data include actual project cost, material cost, labor cost, management cost and construction delay time.

[0010] Furthermore, the construction method of the cost prediction model includes:

[0011] Collect historical engineering project data, which includes actual cost data and project characteristics of m historical time periods;

[0012] Clean the historical project data by processing missing values ​​and outliers; obtain perfect project data; standardize or normalize the numerical data in the perfect project data; perform one-hot encoding on the categorical data to convert it into numerical data;

[0013] Based on the actual cost data of m historical time periods, the actual construction value of the corresponding time period is calculated, and according to the distribution of the calculated actual construction value, it is divided into M specific intervals and categories are defined;

[0014] In the set of training samples in the entire time period, the number of samples in each category is counted; the calculated actual construction value is used as a training sample to build a decision tree; M1 decision trees are repeatedly built and integrated to complete the construction of the cost prediction model;

[0015] Methods for obtaining the cost deviation include:

[0016] Based on the relevant data of n time periods and the constructed cost prediction model, the theoretical construction value of n time periods is predicted; the calculation formula for the cost deviation of n time periods is:

[0017] ;in, For the The cost deviation of each time period, For the The actual value of a period of time, For the Theoretical value of a time period.

[0018] Furthermore, the decision tree is constructed in the following manner:

[0019] All training samples constitute the data set , for the dataset , calculate its information entropy ;

[0020] ;in, For the dataset Middle The proportion of training samples in each category, For the The cost function value of each category is is a positive number;

[0021] ;in, is the progress impact coefficient, is a hyperparameter; is the weight coefficient of the basic cost,

[0022] For the Cost deviation value of each category; The time for which the project progress is delayed;

[0023] Schedule Impact Factor ;in, is the basic constant term, For the The weight coefficient corresponding to each item feature, For the The characteristic value of each item feature; is the hyperbolic tangent function; For the The magnification factor of each item feature; For the The steepness shift coefficient of the item characteristics; is the center shift coefficient;

[0024] Based on the dataset Information entropy , for project features , calculate its information gain ;

[0025] A decision tree is constructed based on information gain. At each node of the decision tree, the information gain of all project features is calculated, and the project feature with the largest information gain is selected as the split feature. If the information gain of all project features is less than the preset gain threshold, the current node is marked as a leaf node; the decision tree is repeatedly recursively constructed, and in the process of repeatedly recursively constructing the decision tree, the decision tree is pruned to complete the construction of the decision tree.

[0026] Furthermore, the information gain The formula is:

[0027] ;in, According to the project characteristics No. The sub-datasets after feature value division are For sub-datasets The size of For the dataset The size of For sub-datasets The information entropy of To adjust the parameters, is the exponential factor, is the feature correlation tuning parameter, For project features With The relevance of the categories; For sub-datasets The main categories that exist;

[0028] ;in, is the total number of training samples, For project features No. The eigenvalue and The frequency of co-occurrence of categories; For project features No. The eigenvalue and The expected frequency of the categories occurring simultaneously;

[0029] ;in, For project features No. The number of times a feature value appears in the entire data set. For the The number of times a category appears in the entire data set;

[0030] The pruning method includes:

[0031] Define the pruning evaluation function. The expression of the pruning evaluation function is:

[0032] ;in, is the pruning evaluation value, is the number of misclassified training samples, is the number of training samples of the current node, is the number of leaf nodes to be pruned later, is the depth of the current node, is the post-pruning adjustment parameter, is the sample adjustment parameter, is the depth adjustment parameter;

[0033] A pruning threshold is preset. If the pruning evaluation value is less than the pruning threshold, the node is retained; otherwise, it is pruned; until all training samples are correctly classified or the stopping condition is met, and the pruning evaluation values ​​are all less than the pruning threshold.

[0034] Furthermore, the method of determining the cost risk level of n time periods includes:

[0035] Standardize or normalize the cost deviations of n time periods to make their value ranges consistent; and use the standardized or normalized cost deviations as samples;

[0036] Randomly initialize a group of bat individuals, each bat individual corresponds to a clustering scheme;

[0037] Initialize parameters, including frequency, pulse emission rate, local search step, and global search step range [Amin, Amax]; where Amin is the upper limit of the local search step range, and Amax is the lower limit of the local search step range;

[0038] Define the fitness function of each bat individual ; Calculate the fitness function for each bat individual The value of

[0039] ;in, is the average distance between a sample and other samples in the same cluster, is the average distance between the sample and all samples in the nearest neighbor cluster, is the first adjustment parameter, is the exponential parameter, is the second adjustment parameter;

[0040] Defining the optimal number of clusters , using the optimal number of clusters , the individual positions of individual bats are updated iteratively; the formula for updating iteration is:

[0041] ;in, For bat individuals exist The position at the iteration, For bat individuals exist The position at the iteration, is a very small random disturbance, is the standard normal distribution function; For bat individuals exist The speed of the iteration;

[0042] ;in, is the inertia weight, For bat individuals exist The speed of the iteration; is the first acceleration constant, is the first random number, For bat individuals The best location in history, For bat individuals exist The position acceleration disturbance term at the iteration; For bat individuals frequency;

[0043] ;in, is the second acceleration constant, is the second random number; For The global optimal position at the iteration is For A good position is randomly selected at the iteration;

[0044] ;in, is the current iteration number, is the maximum number of iterations; is the minimum value of the preset frequency, is the maximum value of the preset frequency, is a random vector; is the attenuation factor;

[0045] Based on the update iteration of individual positions, a local search is performed on the current optimal bat individual. The formula for the local search is:

[0046] ;in, For the new solution, For the old solution, is the third random number, is the step length of the current local search; the final global optimal individual is the optimal clustering solution;

[0047] Analyze each cluster in the optimal clustering solution and calculate the mean value of the cost deviation of each cluster;

[0048] Based on the size of the mean value of the cost deviation of each cluster, the n time periods in each cluster are sorted in descending order and numbered as the cost risk level of the n time periods.

[0049] Furthermore, the optimal number of clusters The definitions include:

[0050] In the initial state, each sample is a cluster, and there are n clusters in total. The distance between any two clusters is calculated using the Euclidean distance, and the two clusters with the closest distance are found and merged into a new cluster. At this time, the number of clusters is reduced by 1, and the process is repeated until all samples are aggregated into one cluster, generating a bottom-up clustering tree.

[0051] Define the clustering tree function ;

[0052] ;in, , and is the weighting coefficient; For the The number of samples in a cluster; For the The number of samples in a cluster, For the Clusters and The average shortest distance of clusters; is the maximum distance between all pairs of samples; For the The shape score of the cluster, The number of clusters is Cluster distribution score at ;

[0053] ;in, is the balance parameter, For the The silhouette coefficient of the clusters, For the The convex hull area of ​​the clusters, For the The convex hull perimeter of a cluster, For the The maximum distance between any two samples in a cluster, For the The minimum distance between any two samples in a cluster;

[0054] On the clustering tree, from Start, calculate When When increasing from small to large, The value increases first and then decreases. The maximum value is obtained at the value, and the corresponding The value is the optimal number of clusters.

[0055] Furthermore, the method for determining the cost control weight coefficients of the n time periods includes:

[0056] Statistics in all clusters, The number of occurrences of the cost risk level corresponding to each time period is normalized to the interval [0, 1] to obtain the The number of risks in a time period;

[0057] The calculation formula of the cost control weight coefficient is:

[0058] ;in, and To control the weight adjustment parameters; For the The number of risks in a time period; is the maximum number of risks in all time periods within a cluster, is the minimum number of risks in all time periods within the cluster; For the The mean value of the cost deviation in a time period;

[0059] The formula for adjusting the theoretical construction value using the cost control weight coefficient is:

[0060] ;in, is the global adjustment coefficient, Adjustment parameters for construction index; For the The target value for each time period;

[0061] ;in, For the The construction difficulty coefficient of each time period; is the minimum value of the construction difficulty coefficient in all time periods, is the maximum value of the construction difficulty coefficient in all time periods,

[0062] ;in, and is the balance weight parameter; For the The weight of the construction process, For the The construction technology is in The complexity score of each time period; For the The number of construction processes involved in each time period; For the The environmental constraints are The impact score of each time period, To influence the The number of constraints in the construction environment during a period of time.

[0063] Furthermore, the method for determining the abnormal cost time period includes:

[0064] Calculate the deviation between the actual cost and the target cost and record it as the cost deviation value;

[0065] The calculation formula for the deviation is: ;in, For the The deviation between the actual construction cost and the target construction cost in a certain period of time;

[0066] Calculate the cost deviation values ​​of n time periods to obtain a deviation value set DPL; sort the cost deviation values ​​in the deviation value set DPL in ascending order to obtain an ordered set DPL'; use the cost deviation values ​​as data points;

[0067] Extract the lower quartile of the ordered set DPL' , median and upper quartile ;

[0068] Based on the upper quartile and lower quartile Calculate interquartile range ;

[0069] Based on interquartile range Determine the inner shank length and external handle length ;

[0070] ;

[0071] ;

[0072] Define a rectangular box with the lower boundary as , the upper boundary of the rectangular box is , draw a horizontal line in the rectangular box to indicate the median ,from Extend a line downwards the length of the inner handle ;

[0073] from Extend a line upwards the length of the inner handle The endpoints of these two lines are the internal handle endpoints; the data points marked outside the internal handle endpoints but within the length of the external handle are suspected outliers; the data points marked outside the length of the external handle are confirmed outliers;

[0074] The abnormal cost time period includes the suspected abnormal time period and the confirmed abnormal time period. For each time period, if its cost deviation value is within the range of the suspected outlier value, it is marked as the suspected abnormal time period; if its cost deviation value is a confirmed outlier value, it is marked as the confirmed abnormal time period.

[0075] A method for controlling engineering cost process intelligently, which is implemented based on the engineering cost process intelligent control system, comprises: S1, constructing a cost prediction model for obtaining theoretical cost value, collecting relevant data of the engineering project in n time periods, and determining the cost deviation of the n time periods according to the relevant data and the cost prediction model;

[0076] S2. Determine the cost risk level of the n time periods according to the cost deviation of the n time periods, and determine the cost control weight coefficient of the n time periods based on the cost risk level;

[0077] S3. Use the cost control weight coefficient to adjust the theoretical construction value to obtain the target construction value for n time periods;

[0078] S4. Perform outlier detection on the deviation between the actual cost and the target cost in n time periods, determine the abnormal cost time period, and take cost optimization scheduling for the abnormal cost time period.

[0079] The technical effects and advantages of an engineering intelligent cost process control system and control method of the present invention are as follows:

[0080] The present invention has the ability to dynamically predict costs, can track the actual progress of engineering projects in real time, and adjust the prediction model in time, so that the predicted cost value is closer to reality, providing accurate basic data for subsequent cost control; secondly, it introduces the concept of risk level classification management, dynamically evaluates and grades the risk level according to the cost deviation at different stages, and formulates differentiated cost control strategies, thereby realizing refined and factorized cost control and improving the pertinence and effectiveness of control; furthermore, it can dynamically adjust the target cost value according to various influencing factors such as construction difficulty, process complexity, and environmental constraints, making the cost adjustment more scientific and reasonable, avoiding the deviation caused by human subjective judgment, and improving the accuracy of cost adjustment; in addition, a recognition and optimization scheduling mechanism for abnormal cost time periods is established, and different monitoring and scheduling strategies are adopted for suspicious anomalies and confirmed anomalies, respectively, to realize real-time monitoring and closed-loop control of costs, effectively prevent and resolve cost risks, and ensure the controllability and economy of engineering costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 A schematic diagram of an engineering intelligent cost process control system of the present invention;

[0082] Figure 2 It is a schematic diagram of an engineering intelligent cost process control method of the present invention. DETAILED DESCRIPTION

[0083] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0084] Example 1

[0085] See also Figure 1 As shown, the engineering intelligent cost process control system described in this embodiment includes:

[0086] The data collection and analysis module is used to construct a cost prediction model for obtaining theoretical construction value, collect relevant data of the project in n time periods, and determine the cost deviation of n time periods based on the relevant data and the cost prediction model;

[0087] The risk and weight fitting module is used to determine the cost risk level of n time periods according to the cost deviation of n time periods, and determine the cost control weight coefficient of n time periods based on the cost risk level;

[0088] A real-time adjustment module is used to adjust the theoretical construction value using the cost control weight coefficient to obtain the target construction value for n time periods;

[0089] The analysis and scheduling module is used to detect outliers between the deviations between the actual construction costs and the target construction costs in n time periods, determine the abnormal cost time periods, and take cost optimization scheduling for the abnormal cost time periods; each module is connected by wired and / or wireless means to realize data transmission between modules.

[0090] Relevant data include actual project cost, material cost, labor cost, management cost and construction delay time.

[0091] Furthermore, the construction methods of the cost prediction model include:

[0092] Collect historical engineering project data, which include actual cost data (material costs, labor costs, management costs, etc.) and project characteristics (building area, structure type, number of floors, etc.) for m historical time periods.

[0093] Clean the historical project data by processing missing values ​​and outliers; obtain complete project data; standardize or normalize the numerical data in the complete project data to ensure that the influence of features of different magnitudes on the model is consistent; perform one-hot encoding on categorical data to convert it into numerical data.

[0094] The actual construction value of the corresponding time period is calculated based on the actual construction cost data of m historical time periods. According to the distribution of the calculated actual construction value, it is divided into M specific intervals and defined categories; for example, Category 1: the actual construction value is between 90%-100% of the planned construction value; Category 2: the actual construction value is between 100%-110% of the planned construction value; Category 3: the actual construction value is between 110%-120% of the planned construction value...

[0095] For each training sample in each time period, calculate which category interval its actual value falls into; in the set of training samples in the entire time period, count the number of samples in each category;

[0096] The calculated actual value is used as a training sample to construct a decision tree.

[0097] Specifically, all training samples constitute a data set , for the dataset , calculate its information entropy ;

[0098] ;in, For the dataset Middle The proportion of training samples in each category, For the The cost function value of each category is A small positive number used to avoid the denominator being 0.

[0099] ;in, is the progress impact coefficient, It is a hyperparameter that controls the nonlinearity of the function; is the weight coefficient of the basic cost,

[0100] For the Cost deviation value of each category; The time for project progress delay.

[0101] Schedule Impact Factor ;in, is the basic constant term, which represents the default impact coefficient of schedule delay in the absence of any project characteristic information. For the The weight coefficient corresponding to each project feature is used to adjust the weight of different features. The influence of the value, For the The characteristic value (code or number) of each item characteristic; is a hyperbolic tangent function, which performs nonlinear mapping on the eigenvalues, and the mapping range is between (-1, 1); For the The magnification factor of each project feature is used to amplify the output value of the function to a suitable value range; For the The steepness shift coefficient of each item feature controls the steepness of the nonlinear function; is the center offset coefficient, which controls the center position of the nonlinear function.

[0102] Based on the dataset Information entropy , for project features , calculate its information gain ; ;in, According to the project characteristics No. The sub-datasets after feature value division are For sub-datasets The size of For the dataset The size of For sub-datasets The information entropy of To adjust the parameters, used to control the sensitivity of information gain to sample distribution, is an exponential factor used to adjust the impact of the main categories. is the feature correlation tuning parameter, For project features With The relevance of the categories; For sub-datasets The main category that exists (the category that is the majority).

[0103] ;in, is the total number of training samples, For project features No. The eigenvalue and The frequency of co-occurrence of categories; For project features No. The eigenvalue and The expected frequency of the categories occurring simultaneously.

[0104] ;in, For project features No. The number of times a feature value appears in the entire data set. For the The number of times a category appears in the entire dataset.

[0105] A decision tree is constructed based on information gain. At each node of the decision tree, the information gain of all project features is calculated, and the project feature with the largest information gain is selected as the split feature. If the information gain of all project features is less than the preset gain threshold, the current node is marked as a leaf node; the decision tree is repeatedly recursively constructed, and the decision tree is pruned in the process of repeatedly recursively constructing the decision tree.

[0106] Ways to perform pruning include:

[0107] Define the pruning evaluation function. The expression of the pruning evaluation function is:

[0108] ;in, is the pruning evaluation value, is the number of misclassified training samples, is the number of training samples of the current node, is the number of leaf nodes to be pruned later, is the depth of the current node, is the post-pruning adjustment parameter, which adjusts the influence of the number of leaf nodes on the logarithmic term. is the sample adjustment parameter, which adjusts the influence of the number of samples. It is the depth adjustment parameter, adjust the depth value.

[0109] The smaller the pruning evaluation value, the smaller the cost of retaining the current node, and the node tends to be retained; conversely, the larger the pruning evaluation value, the smaller the pruning cost, and the node tends to be pruned.

[0110] The pruning threshold is preset. If the pruning evaluation value is less than the pruning threshold, the node is retained; otherwise, it is pruned; until all training samples are correctly classified or the stopping condition is met (the information gain is less than the preset gain threshold), and the pruning evaluation values ​​are all less than the pruning threshold; then the construction of the decision tree is completed. The constructed decision tree consists of internal nodes (representing feature tests) and leaf nodes (representing category predictions); repeatedly construct M1 decision trees and integrate them to complete the construction of the cost prediction model.

[0111] Furthermore, the method for obtaining the cost deviation includes:

[0112] Based on the relevant data of n time periods and the constructed cost prediction model, the theoretical construction value of n time periods is predicted; the calculation formula for the cost deviation of n time periods is:

[0113] ;in, For the The cost deviation of each time period, For the The actual value of a time period (based on relevant data acquisition), For the Theoretical value of a time period.

[0114] Furthermore, the method of determining the cost risk level of n time periods includes:

[0115] The cost deviations of n time periods are standardized or normalized to make their numerical ranges consistent; and the standardized or normalized cost deviations are used as samples.

[0116] A group of bat individuals (solutions) are randomly initialized, and each bat individual corresponds to a clustering scheme, which includes cluster centers and cluster allocations.

[0117] Initialize parameters, including frequency, pulse emission rate, local search step, and global search step range [Amin, Amax]; where Amin is the upper limit of the local search step range, and Amax is the lower limit of the local search step range.

[0118] Define the fitness function of each bat individual (clustering scheme) ; Calculate the fitness function for each bat individual (clustering scheme) The value of is used as the evaluation indicator of clustering quality.

[0119] ;in, is the average distance between the sample and other samples in the same cluster (cost deviation), is the average distance between the sample and all samples in the nearest neighbor cluster, is the first adjustment parameter, is the exponential parameter, is the second adjustment parameter; fitness function The value of belongs to the interval [-1, 1]. The larger the value, the more appropriate the cluster into which the sample is divided.

[0120] Defining the optimal number of clusters Specifically, in the initial state, each sample is a cluster, and there are n clusters in total; the Euclidean distance is used to calculate the distance between any two clusters, and the two closest clusters are found and merged into a new cluster; at this time, the number of clusters is reduced by 1, and the process is repeated until all samples are aggregated into one cluster, generating a bottom-up clustering tree.

[0121] Define the clustering tree function , evaluate when the number of clusters is The clustering quality when

[0122] ;in, , and is the weighting coefficient, and ; For the The number of samples in a cluster; For the The number of samples in a cluster, For the Clusters and The average shortest distance of clusters; is the maximum distance between all pairs of samples; For the The shape score of the cluster, The number of clusters is The cluster distribution score at .

[0123] ;in, is the balance parameter, For the The silhouette coefficient of the clusters, For the The area of ​​the convex hull of the clusters (the smallest convex polygon that still contains all the points), For the The convex hull perimeter of a cluster, For the The maximum distance between any two samples in a cluster, For the The minimum distance between any two samples in a cluster.

[0124] On the clustering tree, from Start, calculate When When increasing from small to large, The value increases first and then decreases. The maximum value is obtained at the value, and the corresponding The value is the optimal number of clusters.

[0125] Use the optimal number of clusters , the individual positions of the bats (clustering scheme) are updated and iterated; the formula for updating and iterating is:

[0126] ;in, For bat individuals exist The position at the iteration, For bat individuals exist The position at the iteration, is a very small random disturbance, is the standard normal distribution function; For bat individuals exist Iteration speed.

[0127] ;in, is the inertia weight, For bat individuals exist The speed of the iteration; is the first acceleration constant, is the first random number in the range [0, 1], For bat individuals The best location in history, For bat individuals exist The position acceleration disturbance term at the iteration; For bat individuals frequency.

[0128] ;in, is the second acceleration constant, is a second random number in the range of [0, 1]; For The global optimal position at the iteration is For A good position is randomly selected at the iteration.

[0129] ;in, is the current iteration number, is the maximum number of iterations; is the minimum value of the preset frequency, is the maximum value of the preset frequency, is a random vector in the range [0, 1], introducing random perturbations for each dimension; is the attenuation factor, which is a positive constant that controls the speed of attenuation.

[0130] Based on the update iteration of individual positions, a local search is performed on the current optimal bat individual. The formula for the local search is:

[0131] ;in, is the new solution (a new candidate solution position generated by a certain perturbation mechanism in the neighborhood of the old solution), is the old solution (the position of the candidate solution corresponding to a bat individual in the current iteration), is the third random number in the range of [-1, 1], is the step size of the current local search; by continuously performing such a local search process, we can gradually find better solutions and eventually converge to the (approximate) optimal solution to the problem, and finally obtain the global optimal individual, that is, the optimal clustering solution.

[0132] Analyze each cluster in the optimal clustering solution and calculate the mean value of the cost deviation of each cluster;

[0133] Based on the size of the mean cost deviation of each cluster, the n time periods in each cluster are sorted in descending order and numbered. The time period with the largest mean cost deviation is ranked first, and then ranked in descending order. After the sorting is completed, the n time periods are numbered as the cost risk level of the n time periods.

[0134] Furthermore, the method for determining the cost control weight coefficients of n time periods includes:

[0135] Statistics in all clusters, The number of occurrences of the cost risk level corresponding to each time period is normalized to the interval [0, 1] to obtain the The number of risks in a time period.

[0136] The calculation formula of the cost control weight coefficient is:

[0137] ;in, and To control the weight adjustment parameters, Greater than 0 and Belongs to the interval [0, 1]; For the The number of risks in a time period; is the maximum number of risks in all time periods within a cluster, is the minimum number of risks in all time periods within the cluster; For the The mean value of the cost deviation in a time period.

[0138] Furthermore, the formula for adjusting the theoretical construction value using the cost control weight coefficient is:

[0139] ;in, is the global adjustment coefficient, which is a constant. Adjustment parameters for construction index; For the Create value based on the goals of a time period.

[0140] ;in, For the The construction difficulty coefficient of each time period; is the minimum value of the construction difficulty coefficient in all time periods, It is the maximum value of the construction difficulty coefficient in all time periods.

[0141] ;in, and is the balance weight parameter, and ; For the The weight of the construction process, For the The construction technology is in The complexity score of each time period, with a value range of [0, 1]; For the The number of construction processes involved in each time period; For the The environmental constraints are The impact score of each time period, To influence the The number of constraints in the construction environment during a period of time.

[0142] The complexity score of the construction process and the number of constraints of the construction environment are obtained in the following ways:

[0143] Collect construction data of similar projects in the past, analyze the complexity and impact of each process and environmental factor in different time periods, and fit the and Scoring models (such as multiple linear regression, taking factors affecting process complexity and environmental impact as independent variables, and As the dependent variable, a multiple linear regression model was established to obtain

[0144] Furthermore, the method for determining the abnormal cost time period includes:

[0145] Calculate the deviation between the actual cost and the target cost and record it as the cost deviation value;

[0146] The calculation formula for the deviation is: ;in, For the The deviation between the actual construction cost and the target construction cost in a time period (cost deviation value).

[0147] Calculate the cost deviation values ​​of n time periods to obtain a deviation value set DPL; sort the cost deviation values ​​in the deviation value set DPL in ascending order to obtain an ordered set DPL'; and use the cost deviation values ​​as data points.

[0148] Extract the lower quartile of the ordered set DPL' , median and upper quartile , lower quartile , which means the data point at n / 4 in the ordered set DPL', the upper quartile Indicates the data point at 3n / 4 in the ordered set DPL'; median , represents the data point located at n / 2 in the ordered set DPL'.

[0149] Based on the upper quartile and lower quartile Calculate interquartile range ;

[0150] Based on interquartile range Determine the inner shank length and external handle length ;

[0151] ;

[0152] .

[0153] Define a rectangular box with the lower boundary as , the upper boundary of the rectangular box is , draw a horizontal line in the rectangular box to indicate the median ,from Extend a line downwards the length of the inner handle ;

[0154] from Extend a line upwards the length of the inner handle The endpoints of these two lines are the internal handle endpoints; data points marked outside the internal handle endpoints but within the length of the external handle are suspected outliers; data points marked outside the length of the external handle are confirmed outliers.

[0155] The abnormal cost time period includes the suspected abnormal time period and the confirmed abnormal time period. For each time period, if its cost deviation value is within the range of the suspected outlier value, it is marked as the suspected abnormal time period; if its cost deviation value is a confirmed outlier value, it is marked as the confirmed abnormal time period.

[0156] Furthermore, the method of optimizing the cost for the abnormal cost period includes:

[0157] Strengthen cost monitoring for time periods marked as suspicious abnormal time periods, timely grasp the deviation change trend, analyze the causes of the deviation, whether it is a temporary fluctuation or a continuous abnormality, and set an early warning threshold. When the cost deviation value exceeds the early warning threshold, the early warning is triggered, and the early warning information notifies relevant personnel to remind them to pay attention and prepare to take action; continue to track the changes in the cost deviation value of the suspicious abnormal time period; if it changes from a suspicious abnormal time period to a confirmed abnormal time period, then switch to executing the optimized scheduling strategy for the confirmed abnormality.

[0158] Optimal scheduling strategies to identify anomalies include:

[0159] Analyze the cause of the abnormality, whether it is human error or force majeure. For human errors, hold the relevant personnel accountable, and reallocate resources according to the abnormality type (positive deviation or negative deviation) of the determined abnormal time period; in case of positive deviation, resources can be deployed from other time periods to the determined abnormal time period; in case of negative deviation, some resources can be transferred from the determined abnormal time period to other time periods; and dynamically adjust the construction period, such as compressing or extending the construction period of the abnormal time period, increasing or reducing resource input, and preparing emergency plans for the determined abnormal time period, such as replacing the construction team, purchasing emergency materials, etc., to control the expansion of the abnormality.

[0160] This embodiment has the ability to dynamically predict the cost, can track the actual progress of the project in real time, and adjust the prediction model in time, so that the cost prediction value is closer to reality, providing accurate basic data for subsequent cost control; secondly, it introduces the concept of risk level classification management, dynamically evaluates and grades the risk level according to the cost deviation at different stages, and formulates differentiated cost control strategies, realizing refined and factorized cost control, and improving the pertinence and effectiveness of control; thirdly, it can dynamically adjust the target cost value according to various influencing factors such as construction difficulty, process complexity, environmental constraints, etc., making the cost adjustment more scientific and reasonable, avoiding the deviation caused by human subjective judgment, and improving the accuracy of cost adjustment; in addition, a mechanism for identifying and optimizing the scheduling of abnormal cost time periods is established, and different monitoring and scheduling strategies are adopted for suspicious anomalies and confirmed anomalies, respectively, to realize real-time monitoring and closed-loop control of costs, effectively prevent and resolve cost risks, and ensure the controllability and economy of engineering costs.

[0161] Example 2

[0162] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides an engineering intelligent cost process control method, including:

[0163] S1. Construct a cost prediction model for obtaining theoretical construction value, collect relevant data of the project in n time periods, and determine the cost deviation of n time periods based on the relevant data and the cost prediction model;

[0164] S2. Determine the cost risk level of the n time periods according to the cost deviation of the n time periods, and determine the cost control weight coefficient of the n time periods based on the cost risk level;

[0165] S3. Use the cost control weight coefficient to adjust the theoretical construction value to obtain the target construction value for n time periods;

[0166] S4. Perform outlier detection on the deviation between the actual cost and the target cost in n time periods, determine the abnormal cost time period, and take cost optimization scheduling for the abnormal cost time period.

[0167] Example 3

[0168] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the engineering intelligent cost process control method provided above is implemented.

[0169] Since the electronic device introduced in this embodiment is an electronic device used to implement a method for controlling an engineering cost process intelligently in the embodiment of this application, based on the method for controlling an engineering cost process intelligently in the embodiment of this application, a person skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not described in detail here. As long as a person skilled in the art implements the electronic device used in the method for controlling an engineering cost process intelligently in the embodiment of this application, it belongs to the scope of protection of this application.

[0170] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0171] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. An engineering intelligent cost process control system, characterized in that: include: The data collection and analysis module is used to construct a cost prediction model for obtaining theoretical construction value, collect relevant data of the project in n time periods, and determine the cost deviation of n time periods based on the relevant data and the cost prediction model; The risk and weight fitting module is used to determine the cost risk level of n time periods according to the cost deviation of n time periods, and determine the cost control weight coefficient of n time periods based on the cost risk level; A real-time adjustment module is used to adjust the theoretical construction value using the cost control weight coefficient to obtain the target construction value for n time periods; The analysis and scheduling module is used to perform outlier detection on the deviation between the actual cost and the target cost in n time periods, determine the abnormal cost time period, and take cost optimization scheduling for the abnormal cost time period; each module is connected by wired and / or wireless means.

2. The engineering intelligent cost process control system according to claim 1 is characterized in that: The relevant data include actual project cost, material cost, labor cost, management cost and construction delay time.

3. The engineering intelligent cost process control system according to claim 2 is characterized in that: The construction method of the cost prediction model includes: Collect historical engineering project data, which includes actual cost data and project characteristics of m historical time periods; Clean the historical project data by processing missing values ​​and outliers; obtain perfect project data; standardize or normalize the numerical data in the perfect project data; perform one-hot encoding on the categorical data to convert it into numerical data; Based on the actual cost data of m historical time periods, the actual construction value of the corresponding time period is calculated, and according to the distribution of the calculated actual construction value, it is divided into M specific intervals and categories are defined; In the set of training samples in the entire time period, the number of samples in each category is counted; the calculated actual construction value is used as a training sample to build a decision tree; M1 decision trees are repeatedly built and integrated to complete the construction of the cost prediction model; Methods for obtaining the cost deviation include: Based on the relevant data of n time periods and the constructed cost prediction model, the theoretical construction value of n time periods is predicted; the calculation formula for the cost deviation of n time periods is: ;in, For the The cost deviation of each time period, For the The actual value of a period of time, For the Theoretical value of a time period.

4. The engineering intelligent cost process control system according to claim 3 is characterized in that: The decision tree is constructed in the following manner: All training samples constitute the data set , for the dataset , calculate its information entropy ; ;in, For the dataset Middle The proportion of training samples in each category, For the The cost function value of each category is is a positive number; ;in, is the progress impact coefficient, is a hyperparameter; is the weight coefficient of the basic cost, For the Cost deviation value of each category; The time for which the project progress is delayed; Schedule Impact Factor ;in, is the basic constant term, For the The weight coefficient corresponding to each item feature, For the The characteristic value of each item feature; is the hyperbolic tangent function; For the The amplification factor of each item feature; For the The steepness shift coefficient of the item characteristics; is the center shift coefficient; Based on the dataset Information entropy , for project features , calculate its information gain ; A decision tree is constructed based on information gain. At each node of the decision tree, the information gain of all project features is calculated, and the project feature with the largest information gain is selected as the split feature. If the information gain of all project features is less than the preset gain threshold, the current node is marked as a leaf node; the decision tree is repeatedly recursively constructed, and in the process of repeatedly recursively constructing the decision tree, the decision tree is pruned to complete the construction of the decision tree.

5. The engineering intelligent cost process control system according to claim 4 is characterized in that: The information gain The formula is: ;in, According to the project characteristics No. The sub-datasets after feature value division are For sub-datasets The size of For the dataset The size of For sub-datasets The information entropy of To adjust the parameters, is the exponential factor, is the feature correlation tuning parameter, For project features With The relevance of the categories; For sub-datasets The main categories that exist; ;in, is the total number of training samples, For project features No. The eigenvalue and The frequency of co-occurrence of categories; For project features No. The eigenvalue and The expected frequency of the categories occurring simultaneously; ;in, For project features No. The number of times a feature value appears in the entire data set. For the The number of times a category appears in the entire data set; The pruning method includes: Define the pruning evaluation function. The expression of the pruning evaluation function is: ;in, is the pruning evaluation value, is the number of misclassified training samples, is the number of training samples of the current node, is the number of leaf nodes to be pruned later, is the depth of the current node, is the post-pruning adjustment parameter, is the sample adjustment parameter, is the depth adjustment parameter; A pruning threshold is preset. If the pruning evaluation value is less than the pruning threshold, the node is retained; otherwise, it is pruned; until all training samples are correctly classified or the stopping condition is met, and the pruning evaluation values ​​are all less than the pruning threshold.

6. The engineering intelligent cost process control system according to claim 5 is characterized in that: The method of determining the cost risk level of n time periods includes: Standardize or normalize the cost deviations of n time periods to make their value ranges consistent; and use the standardized or normalized cost deviations as samples; Randomly initialize a group of bat individuals, each bat individual corresponds to a clustering scheme; Initialize parameters, including frequency, pulse emission rate, local search step, and global search step range [Amin, Amax]; where Amin is the upper limit of the local search step range, and Amax is the lower limit of the local search step range; Define the fitness function of each bat individual ; Calculate the fitness function for each bat individual The value of ;in, is the average distance between a sample and other samples in the same cluster, is the average distance between the sample and all samples in the nearest neighbor cluster, is the first adjustment parameter, is the exponential parameter, is the second adjustment parameter; Defining the optimal number of clusters , using the optimal number of clusters , the individual positions of individual bats are updated iteratively; the formula for updating iteration is: ;in, For bat individuals exist The position at the iteration, For bat individuals exist The position at the iteration, is a very small random disturbance, is the standard normal distribution function; For bat individuals exist The speed of the iteration; ;in, is the inertia weight, For bat individuals exist The speed of the iteration; is the first acceleration constant, is the first random number, For bat individuals The best location in history, For bat individuals exist The position acceleration disturbance term at the iteration; For bat individuals frequency; ;in, is the second acceleration constant, is the second random number; For The global optimal position at the iteration is For A good position is randomly selected at the iteration; ;in, is the current iteration number, is the maximum number of iterations; is the minimum value of the preset frequency, is the maximum value of the preset frequency, is a random vector; is the attenuation factor; Based on the update iteration of individual positions, a local search is performed on the current optimal bat individual. The formula for the local search is: ;in, For the new solution, For the old solution, is the third random number, is the step length of the current local search; the final global optimal individual is the optimal clustering solution; Analyze each cluster in the optimal clustering solution and calculate the mean value of the cost deviation of each cluster; Based on the size of the mean value of the cost deviation of each cluster, the n time periods in each cluster are sorted in descending order and numbered as the cost risk level of the n time periods.

7. The engineering intelligent cost process control system according to claim 6 is characterized in that: The optimal number of clusters The definitions include: In the initial state, each sample is a cluster, and there are n clusters in total. The distance between any two clusters is calculated using the Euclidean distance, and the two clusters with the closest distance are found and merged into a new cluster. At this time, the number of clusters is reduced by 1, and the process is repeated until all samples are aggregated into one cluster, generating a bottom-up clustering tree. Define the clustering tree function ; ;in, , and is the weighting coefficient; For the The number of samples in a cluster; For the The number of samples in a cluster, For the Clusters and The average shortest distance of clusters; is the maximum distance between all pairs of samples; For the The shape score of the cluster, The number of clusters is Cluster distribution score at ; ;in, is the balance parameter, For the The silhouette coefficient of the clusters, For the The convex hull area of ​​the clusters, For the The convex hull perimeter of a cluster, For the The maximum distance between any two samples in a cluster, For the The minimum distance between any two samples in a cluster; On the clustering tree, from Start, calculate When When increasing from small to large, The value increases first and then decreases. The maximum value is obtained at the value, and the corresponding The value is the optimal number of clusters.

8. The engineering intelligent cost process control system according to claim 7 is characterized in that: The method for determining the cost control weight coefficients of the n time periods includes: Statistics in all clusters, The number of occurrences of the cost risk level corresponding to each time period is normalized to the interval [0, 1] to obtain the The number of risks in a time period; The calculation formula of the cost control weight coefficient is: ;in, and To control the weight adjustment parameters; For the The number of risks in a time period; is the maximum number of risks in all time periods within a cluster, is the minimum number of risks in all time periods within the cluster; For the The mean value of the cost deviation in a time period; The formula for adjusting the theoretical construction value using the cost control weight coefficient is: ;in, is the global adjustment coefficient, Adjustment parameters for construction index; For the The target value for each time period; ;in, For the The construction difficulty coefficient of each time period; is the minimum value of the construction difficulty coefficient in all time periods, is the maximum value of the construction difficulty coefficient in all time periods, ;in, and is the balance weight parameter; For the The weight of the construction process, For the The construction technology is in The complexity score of each time period; For the The number of construction processes involved in each time period; For the The environmental constraints are The impact score of each time period, To influence the The number of constraints in the construction environment during a period of time.

9. The engineering intelligent cost process control system according to claim 8 is characterized in that: The determination method of the abnormal cost time period includes: Calculate the deviation between the actual cost and the target cost and record it as the cost deviation value; The calculation formula for the deviation is: ;in, For the The deviation between the actual construction cost and the target construction cost in a certain period of time; Calculate the cost deviation values ​​of n time periods to obtain a deviation value set DPL; sort the cost deviation values ​​in the deviation value set DPL in ascending order to obtain an ordered set DPL'; use the cost deviation values ​​as data points; Extract the lower quartile of the ordered set DPL' , median and upper quartile ; Based on the upper quartile and lower quartile Calculate interquartile range ; Based on interquartile range Determine the inner shank length and the outer handle length ; ; ; Define a rectangular box with the lower boundary as , the upper boundary of the rectangular box is , draw a horizontal line in the rectangular box to indicate the median ,from Extend a line downwards the length of the inner handle ; from Extend a line upwards the length of the inner handle The endpoints of these two lines are the internal handle endpoints; data points marked outside the internal handle endpoints but within the length of the external handle are suspected outliers; data points marked outside the length of the external handle are confirmed outliers; The abnormal cost time period includes the suspected abnormal time period and the confirmed abnormal time period. For each time period, if its cost deviation value is within the range of the suspected outlier value, it is marked as the suspected abnormal time period; if its cost deviation value is a confirmed outlier value, it is marked as the confirmed abnormal time period.

10. A method for controlling an engineering cost process intelligently, which is implemented based on an engineering cost process intelligent control system according to claims 1 to 9, characterized in that: include: S1. Construct a cost prediction model for obtaining theoretical construction value, collect relevant data of the project in n time periods, and determine the cost deviation of n time periods based on the relevant data and the cost prediction model; S2. Determine the cost risk level of the n time periods according to the cost deviation of the n time periods, and determine the cost control weight coefficient of the n time periods based on the cost risk level; S3. Use the cost control weight coefficient to adjust the theoretical construction value to obtain the target construction value for n time periods; S4. Perform outlier detection on the deviation between the actual cost and the target cost in n time periods, determine the abnormal cost time period, and take cost optimization scheduling for the abnormal cost time period.

Citation Information

Patent Citations

  • BIM-based engineering cost dynamic control system and method

    CN111476537A

Cited By

  • Material processing cost data analysis method and system for engineering management

    CN119831174A

  • Data analysis method and system for material processing cost in engineering management

    CN119831174B

  • Method and system for identifying and replacing abnormal value of cost data of hydropower project

    CN122333298A

  • Engineering cost data abnormity intelligent identification method

    CN122636292A

  • Cost risk grading early warning and dynamic adjustment method based on deviation degree clustering analysis

    CN122779901A