Engineering cost multi-dimensional intelligent analysis processing system and method based on big data

Abstract

The invention relates to the technical field of engineering cost data processing, and discloses an engineering cost multi-dimensional intelligent analysis processing system and method based on big data. A projection correction module; a topology construction module; a process derivation module; and a verification module. The method comprises the steps that a project list and environment big data are obtained and mapped into initial multi-dimensional feature vectors; sparse decomposition and orthogonal projection are carried out by using the over-complete dictionary matrix, and a corrected feature vector after noise filtering is generated; constructing an adjacent matrix reflecting process logic according to the environment component; executing a transitive closure operation to derive a hidden process set under a specific working condition; and evaluating the cost logic integrity based on the actual unit price vector and the residual error of the implicit process comprehensive consumption module length. According to the method, noise is filtered through multi-source fusion, the hidden process is accurately identified according to the dynamic working condition, and the logic integrity and accuracy of cost analysis are ensured.

Description

Technical Field

[0001] This invention relates to the field of engineering cost data processing technology, specifically to a multi-dimensional intelligent analysis and processing system and method for engineering cost based on big data. Background Technology

[0002] With the digital transformation of the construction industry, the field of engineering cost has accumulated massive big data resources covering bills of quantities, construction logs, and environmental parameters. Engineering cost management, as the core of project investment control, involves in-depth interaction of multi-dimensional information such as material attributes, construction techniques, and on-site conditions. How to efficiently integrate and intelligently analyze this multi-source heterogeneous information using big data technology to achieve accurate cost data mining and logical verification has become an important direction for the refined management of engineering costs.

[0003] Current engineering cost analysis and processing technologies mainly rely on standardized pricing software and static quota databases. These technologies are typically based on pre-built rule bases, using keyword matching or manual selection to unidirectionally link items in the bill of quantities with price database data, thereby generating cost documents. Although some systems have introduced data analysis functions, most are limited to statistical comparisons of historical cost indicators or simple price fluctuation calculations, focusing on numerical summarization and verification of bill of quantities items.

[0004] However, existing analytical methods have significant shortcomings in handling logical deduction under complex working conditions. Because they mostly employ static rule-matching mechanisms, current technologies struggle to dynamically establish deep logical connections between geological, meteorological, and other environmental data and construction procedures. This leads to the omission of auxiliary, implicit procedures necessary for specific geological conditions due to a lack of logical transmission paths. Furthermore, existing technologies lack effective methods for filtering noise from multi-source big data and logical closed-loop verification mechanisms based on process consumption and price, making it difficult to detect cost distortions caused by logical breaks. Consequently, they fail to meet the requirements of logical integrity and adaptability for multi-dimensional intelligent analysis of engineering costs in a big data environment.

[0005] Therefore, this invention proposes a multi-dimensional intelligent analysis and processing system and method for engineering cost based on big data to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a multi-dimensional intelligent analysis and processing system and method for engineering cost based on big data. It solves the technical problems of existing engineering cost analysis relying on static rule matching, making it difficult to intelligently deduce implicit processes based on dynamic environmental conditions, and lacking noise filtering and logical integrity closed-loop verification of multi-source big data, which leads to the risk of omissions and logical distortion in the cost analysis results.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a multi-dimensional intelligent analysis and processing system for engineering cost based on big data, comprising: The feature mapping module is used to obtain the bill of quantities, construction logs and environmental data of the project to be analyzed, and to map the bill of quantities items into an initial multidimensional feature vector. The projection correction module is used to perform sparse representation and orthogonal projection on the initial multidimensional feature vector using a preset overcomplete dictionary matrix to generate a corrected feature vector after filtering out noise components. The topology construction module is used to calculate the connection weights between list items based on the environmental components in the modified feature vector, and to construct an adjacency matrix that reflects the logical relationship of the process. The process derivation module is used to perform transitive closure operation based on the adjacency matrix to derive the set of implicit processes that exist for the list item under the current environmental conditions; The verification module is used to calculate the actual unit price vector magnitude of the listed sub-items and the comprehensive consumption magnitude of the implicit process set, and to evaluate the logical integrity of the project cost based on the residual between the actual unit price vector magnitude and the comprehensive consumption magnitude.

[0008] Preferably, the step of mapping the items in the bill of quantities to an initial multidimensional feature vector in the feature mapping module includes: Establish the index of the sub-items in the list, and generate the initial multidimensional feature vector composed of material attribute components, process attribute components, environmental condition components, and time dimension components.

[0009] Preferably, the step in the projection correction module of performing sparse representation and orthogonal projection on the initial multidimensional feature vector using a preset overcomplete dictionary matrix to generate a corrected feature vector after filtering out noise components includes: Construct and solve the least squares optimization problem with L1 norm regularization to obtain the sparse coefficient vector of the initial multidimensional feature vector under the overcomplete dictionary matrix; Based on the non-zero elements in the sparse coefficient vector, a set of activation atoms in the overcomplete dictionary matrix is ​​determined, and the activation atoms are used to span a local standard feature subspace; The initial multidimensional feature vector is orthogonally projected into the local standard feature subspace to obtain the modified feature vector.

[0010] Preferably, the step of calculating the connection weights between list items based on the environmental components in the modified feature vector in the topology construction module includes: Extract the environmental component from the modified feature vector corresponding to any two list sub-item nodes; The cosine similarity algorithm is used to calculate the environmental coupling coefficient between the two environmental components. Call the preset logic parameter library to obtain the environmental sensitivity coefficient and inherent logic bias between the two list sub-item nodes; The connection weights are calculated using a nonlinear activation function based on the environmental sensitivity coefficient, the environmental coupling coefficient, and the inherent logic bias.

[0011] Preferably, the step of constructing an adjacency matrix reflecting the logical relationships of processes in the topology construction module includes: Set a connectivity threshold, traverse all possible list item node pairs, and compare the calculated connection weight with the connectivity threshold; When the connection weight is greater than the connectivity discrimination threshold, it is determined that there is a valid logical dependency path between the two list sub-item nodes, and the corresponding element in the adjacency matrix is ​​marked as connected. When the connection weight is less than or equal to the connectivity discrimination threshold, it is determined that there is no direct logical connection between the two list sub-item nodes under the current working condition, and the corresponding element in the adjacency matrix is ​​marked as disconnected. The adjacency matrix is ​​generated based on the traversal and judgment results.

[0012] Preferably, the step of performing transitive closure operation based on the adjacency matrix in the process derivation module includes: Receive the adjacency matrix and iteratively calculate the adjacency matrix using the Warshall algorithm; A reachability matrix is ​​generated according to a Boolean logic iterative formula. The elements in the reachability matrix reveal whether there are direct or indirect logical pathways between the listed items under the current environmental conditions.

[0013] Preferably, the step in the process derivation module to derive the set of implicit processes existing in the list item under the current environmental conditions includes: For node pairs in the reachability matrix that indicate the existence of logical paths, retrieve the set of standard logical paths between the two list item nodes in the preset standard process logic database; Calculate the difference between the standard logical path set and the current bill of quantities index set, mark the nodes in the difference set as candidate implicit processes, and generate a candidate implicit process set; Retrieve the environmental activation threshold vector corresponding to each of the candidate hidden processes, and compare the environmental activation threshold vector with the environmental components of the current project; The candidate hidden processes that satisfy the physical existence conditions are retained, and a set of verified hidden processes is generated.

[0014] Preferably, the step in the verification module to calculate the actual unit price vector magnitude of the item in the inventory and the comprehensive consumption magnitude of the implicit process set includes: Extract the unit prices of labor costs, material costs, machinery costs, and management fees for the sub-items in the bill of quantities, generate an actual unit price vector, and calculate the Euclidean norm of the actual unit price vector to obtain the modulus of the actual unit price vector. Filter out the subset of implicit processes that have logical dependencies on the list items, call the preset consumption quota database to retrieve the standard resource consumption of each implicit process in the subset of implicit processes and generate a standard consumption vector; The total consumption vector is obtained by superimposing the standard consumption vectors of all implicit processes in the implicit process subset, and the magnitude of the total consumption vector is calculated to obtain the comprehensive consumption magnitude.

[0015] Preferably, the step in the verification module to assess the logical integrity of the project cost based on the residual between the actual unit price vector magnitude and the comprehensive consumption magnitude includes: The comprehensive consumption modulus is corrected using the efficiency conversion coefficient, and the difference between the actual unit price vector modulus and the corrected comprehensive consumption modulus is calculated to obtain the residual. Set an integrity judgment threshold range that includes the lower limit deviation threshold and the upper limit deviation threshold of the cost logic; When the residual is within the integrity judgment threshold range, the logic integrity is deemed qualified; when the residual is less than the lower limit deviation threshold of the cost logic, there is a risk of omission or logical break. When the residual exceeds the upper limit deviation threshold of the cost logic, it is determined that there is excessive pricing or insufficient derivation of hidden processes.

[0016] This invention also provides a multi-dimensional intelligent analysis and processing method for engineering cost based on big data, including the following steps: The feature mapping module is used to obtain the bill of quantities, construction logs and environmental data of the project to be analyzed, and the bill of quantities items are mapped to initial multidimensional feature vectors. The projection correction module calls a preset overcomplete dictionary matrix to perform sparse decomposition and orthogonal projection reconstruction on the initial multidimensional feature vector, generating a corrected feature vector after filtering out noise components. The environmental component in the corrected feature vector is extracted using the topology construction module. The connection weights between the list items are calculated based on the environmental components. An adjacency matrix reflecting the logical relationship of the process is constructed based on the connection weights. The process derivation module performs a transitive closure operation based on the adjacency matrix to deduce the set of implicit processes existing in the list item under the current environmental conditions; The logic integrity of evaluating project cost is achieved by using a verification module to calculate the actual unit price vector magnitude of the listed sub-items and the comprehensive consumption magnitude of the set of implicit processes, and by evaluating the residual between the actual unit price vector magnitude and the comprehensive consumption magnitude.

[0017] This invention provides a multi-dimensional intelligent analysis and processing system and method for engineering cost based on big data. It has the following beneficial effects: 1. This invention achieves deep fusion of multi-source heterogeneous big data, including bill of quantities, construction logs, and environmental data, through a feature mapping module and a projection correction module. By utilizing an overcomplete dictionary matrix to perform sparse representation and orthogonal projection on the initial features, noise interference in the original big data is effectively filtered out, constructing a corrected feature vector that accurately reflects materials, processes, and operating conditions. This provides a high-fidelity data foundation for multi-dimensional intelligent analysis and processing of engineering costs, significantly improving the system's feature extraction accuracy and robustness in complex engineering scenarios.

[0018] 2. This invention constructs an adjacency matrix reflecting the logical relationships between work processes and dynamically calculates the connection weights between items in the bill of quantities using environmental condition components. Unlike traditional static rule bases, this method can intelligently identify and activate implicit work process logical dependencies based on specific geological and meteorological environmental conditions. By deriving the set of implicit work processes through transitive closure operations, it accurately reconstructs auxiliary engineering content that is essential under specific harsh working conditions but easily overlooked during bill of quantities compilation, effectively solving the problem of omissions caused by environmental changes in engineering cost analysis.

[0019] 3. This invention establishes a logical integrity assessment mechanism based on vector magnitude through a verification module, quantitatively calculating the residual between the actual unit price vector and the implicit process comprehensive consumption magnitude. This method elevates engineering cost analysis from simple numerical verification to a logically consistent level, effectively identifying logical breaks caused by omissions or data anomalies caused by over-quotation. While ensuring the logical rigor of engineering cost, it achieves closed-loop verification of multi-dimensional intelligent analysis results, ensuring the objectivity and rationality of the final cost assessment result. Attached Figure Description

[0020] Figure 1 This is an architecture diagram of the big data-based multi-dimensional intelligent analysis and processing system for engineering cost in this invention. Figure 2 This is a schematic diagram of the process derivation module of the present invention; Figure 3 This is a flowchart of the big data-based multi-dimensional intelligent analysis and processing method for engineering cost of the present invention.

[0021] Among them, 110 is the feature mapping module; 120 is the projection correction module; 130 is the topology construction module; 140 is the process derivation module; and 150 is the verification module. Detailed Implementation

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see the appendix Figure 1 This invention provides a multi-dimensional intelligent analysis and processing system for engineering cost based on big data. The system runs on a general-purpose computer device or server cluster with computing capabilities. The system is logically divided into multiple sub-modules, including: a feature mapping module 110, a projection correction module 120, a topology construction module 130, a process derivation module 140, and a verification module 150.

[0024] The feature mapping module 110 is used to obtain the bill of quantities, construction logs and environmental data of the project to be analyzed, and to map the bill of quantities items into an initial multidimensional feature vector. The projection correction module 120 is used to perform sparse representation and orthogonal projection on the initial multidimensional feature vector using a preset overcomplete dictionary matrix, and generate a corrected feature vector after filtering out noise components. The topology building module 130 is used to calculate the connection weights between list items based on the environmental components in the modified feature vector, and to build an adjacency matrix that reflects the logical relationship of the process. The process derivation module 140 is used to perform transitive closure operations based on the adjacency matrix to derive the set of implicit processes that exist for the list items under the current environmental conditions. The verification module 150 is used to calculate the actual unit price vector magnitude of the item in the bill of quantities and the comprehensive consumption magnitude of the implicit process set, and to evaluate the logical integrity of the project cost based on the residual between the two.

[0025] See attached document Figure 3 This invention provides a multi-dimensional intelligent analysis and processing method for engineering cost based on big data, including the following steps: S100, use the feature mapping module 110 to obtain the bill of quantities, construction log and environmental data of the project to be analyzed, and map the bill of quantities items into an initial multidimensional feature vector; S200: The projection correction module 120 calls the preset overcomplete dictionary matrix to perform sparse decomposition and orthogonal projection reconstruction on the initial multidimensional feature vector, generating a corrected feature vector after filtering out noise components. S300: The topology building module 130 is used to extract the environmental component in the modified feature vector, calculate the connection weight between the list items based on the environmental component, and construct an adjacency matrix that reflects the logical relationship of the process based on the connection weight. S400, using the process derivation module 140 to perform transitive closure operation based on the adjacency matrix, derives the set of implicit processes that exist for the list items under the current environmental conditions; S500 uses the verification module 150 to calculate the actual unit price vector magnitude of the item in the bill of quantities and the comprehensive consumption magnitude of the implicit process set, and evaluates the logical integrity of the project cost based on the residual of the two.

[0026] The above system and method will be described in detail below.

[0027] See attached document Figure 1 and Figure 3 The feature mapping module 110 is configured to perform data initialization and spatial mapping tasks for the project to be analyzed. The logic of the feature mapping module 110 is to transform heterogeneous engineering entities in the physical world into mathematical objects in a high-dimensional vector space, thereby enabling the computer to perform geometric calculations and analyses on the logical rationality of engineering costs.

[0028] The feature mapping module 110 first connects to the database of the project to be analyzed through a data interface to obtain three types of core source data: bill of quantities, construction logs, and environmental data. Among them: The bill of quantities for the project to be analyzed is a structured data file prepared in accordance with pricing specifications, and its smallest unit is the bill of quantities item (e.g., "C30 cast-in-place reinforced concrete wall" or "site leveling").

[0029] The construction log of the project to be analyzed refers to unstructured text data that records the daily construction site conditions, including work content, machinery allocation, and descriptions of emergencies.

[0030] The environmental data of the project to be analyzed refers to the spatiotemporal parameters corresponding to the project's construction period and geographical location, including the geotechnical parameters in the geological survey report and the daily temperature, humidity and rainfall recorded by the meteorological station.

[0031] The feature mapping module 110 treats each item in the bill of quantities as an independent object to be processed and establishes an index. To comprehensively characterize the cost composition logic of this item under specific working conditions, the feature mapping module 110 maps it into an initial multidimensional feature vector. The initial multidimensional feature vector is composed of four concatenated components: material properties, process properties, environmental conditions, and time dimension. Its mathematical definition is as follows: ; in, Indicates the material property components. Indicates process attribute components, Indicates environmental operating condition components. Represents the time dimension component. This indicates the transpose operation.

[0032] In practice, the generation and processing of each component is as follows: Material property components The numerical encoding is performed. The feature mapping module 110 parses the item feature description fields in the list sub-items. For discrete features in the description (such as material type "HRB400 steel bar" or "silicate cement"), one-hot encoding or hash mapping is used to convert them into binary vectors. For continuous numerical features in the description (such as concrete strength grade, steel diameter, slab thickness), the max-min normalization method is used to map them to the interval [0, 1].

[0033] The discrete feature vectors mentioned above are concatenated with the continuous feature values ​​to form the material property components. This process ensures that material properties with different physical dimensions have computable distances in vector space.

[0034] Process attribute components Semantic extraction and mapping are performed. Since construction logs are unstructured text and have no direct key-value correspondence with the list items, the feature mapping module 110 first determines the operation time window of the list items in the construction schedule and extracts construction log fragments within that time window. Then, it uses a pre-built engineering domain keyword library (containing terms such as "secondary handling," "high formwork," "deep foundation pit," and "nighttime construction") to scan and match the log text. The feature mapping module 110 maintains a process difficulty coefficient table, converting the matched keywords into corresponding difficulty values. For example, the keyword "high groundwater level" corresponds to a difficulty increment of 0.2, and "requires backhoe excavator relay" corresponds to a difficulty increment of 0.3.

[0035] Process attribute components The calculation is a weighted sum of the difficulty coefficients of all process keywords associated with the sub-item in the list. For common text segmentation and keyword retrieval techniques, those skilled in the art can use existing TF-IDF or regular expression matching algorithms, which will not be elaborated upon here.

[0036] Environmental operating condition components The spatiotemporal fusion calculation is performed. This environmental condition component is used to quantify the objective constraints of the external physical environment on construction efficiency. The feature mapping module 110 extracts geological parameters (soil bearing capacity). ) and meteorological parameters (rainfall) Temperature To eliminate dimensional differences (such as kPa and...). (Cannot be directly calculated), the feature mapping module 110 uses the Z-Score normalization method to process the above data, and combines the processed values ​​into a vector form. The calculation formula is as follows: ; in, and These are the mean and standard deviation of historical geological data for the project location; and These are the highest and lowest rainfall amounts recorded for the same period in the region's history; and This represents the statistical characteristic value of historical temperatures for the same period. Using this formula, environmental data is transformed into dimensionless values ​​that follow a standard normal or normalized distribution, objectively reflecting the degree of deviation of current operating conditions from standard operating conditions.

[0037] Time dimension components The dynamic calculation incorporates both market and construction period factors into this time dimension component. The feature mapping module 110 obtains the material price index published on the cost information website for the current construction period. Compared with the base period price index Calculate price volatility Simultaneously calculate the urgency of the construction period. Set up a list of sub-items. The planned construction period is The standard construction period is ,but .when When the time frame is compressed, it indicates that additional costs are needed to expedite the work. The time dimension component is constructed as follows: .

[0038] Finally, the feature mapping module 110 will use the above calculations to... , , , Concatenate them in a predetermined order to generate initial multidimensional feature vectors. This initial multidimensional feature vector unifies the static attributes, environmental constraints, process difficulty, and time factors of the list items into a single vector space, providing a comprehensive input basis for the projection correction module 120 to perform noise removal and data correction.

[0039] See attached document Figure 1 and Figure 3 The projection correction module 120 is connected to the feature mapping module 110 and is used to receive the initial multidimensional feature vector and perform numerical denoising and normalization processing on it using sparse representation theory. The projection correction module 120 introduces a preset overcomplete dictionary matrix and projects the engineering data to be analyzed into a feature subspace that conforms to standard construction logic, thereby generating a corrected feature vector with logical consistency.

[0040] The projection correction module 120 internally stores a preset, overcomplete dictionary matrix. An overcomplete dictionary matrix is ​​a mathematical basis set built upon a large amount of audited historical compliance engineering data. During the training phase before system deployment, a dictionary learning algorithm is executed on a high-quality historical sample set to extract atomic vectors that characterize standard processes, material properties, and environmental response patterns. Defined as ,in The dimension of the feature vector. This represents the number of atoms in a supercomplete dictionary matrix, and This ensures that the overcomplete dictionary matrix possesses overcompleteness, enabling it to flexibly represent any logically consistent engineering features in the form of a sparse linear combination. For the specific implementation of dictionary learning, those skilled in the art can use the K-SVD algorithm or online dictionary learning algorithms, which will not be elaborated upon here.

[0041] The projection correction module 120 initializes the multidimensional feature vector for each input. Perform sparse representation computation. This step is based on an engineering assumption: any compliant cost feature vector can be accurately reconstructed by a linear combination of a few atoms in an overcomplete dictionary matrix, while noise components containing recording errors or logical paradoxes cannot satisfy this sparsity constraint. The projection correction module 120 constructs and solves an L1-norm regularized least squares optimization problem to obtain the corresponding sparse coefficient vector. This optimization process aims to find a set of the sparsest set of weight coefficients such that their corresponding atomic combinations approximate the initial multidimensional feature vector as closely as possible. The specific mathematical model is as follows: ; in, Represents the sparse coefficient vector to be solved; This represents the L1 norm, used to constrain the sparsity of the coefficient vector; Represents the L2 norm; This represents the preset reconstruction error threshold (geometrically defined as the allowable deviation radius centered on the standard feature subspace, used to delineate the boundary between valid data fluctuations and invalid noise interference. In this embodiment, the value is typically set to 10). -2 Up to 5×10-2 Within this range, the value range ensures that while filtering out random sampling noise, it retains subtle feature changes that reflect engineering specificity. For solving the above optimization problem, those skilled in the art can use the Orthogonal Matching Pursuit (OMP) algorithm or the Basis Pursuit algorithm.

[0042] To obtain the optimal sparse coefficient vector Subsequently, the projection correction module 120 performs subspace projection and reconstruction operations. The specific execution mechanism of subspace projection is as follows: sparse coefficient vector The non-zero elements in the matrix determine the overcomplete dictionary matrix. A set of activated atoms in the model spans a local standard feature subspace. To ensure that the final corrected feature vector falls geometrically accurately within the local standard feature subspace, the projection correction module 120 adjusts the initial multidimensional feature vector... The vector is orthogonally projected onto this local standard feature subspace. This process ensures that the reconstructed vector is geometrically located exactly within the initial multidimensional feature vector. At the foot of the perpendicular in the local standard feature subspace, the projection correction module 120 calculates the corrected feature vector according to the following formula. : ; Based on the above calculations, the initial multidimensional feature vector is obtained. It is orthogonally decomposed into two parts: the modified eigenvectors falling within the local eigenspace. and the residual vector perpendicular to the local feature subspace. The residual vector corresponds to the noise component in the original data that cannot be explained by standard process logic. The projection correction module 120 outputs a corrected feature vector after filtering out the noise component. This corrected feature vector retains the valid information in the original data that conforms to standard process logic, ensuring that the data processed by the subsequent topology construction module 130 is based on real physical conditions and compliant construction logic, rather than random errors or recording distortions in the data acquisition process.

[0043] See attached document Figure 1 and Figure 3 The topology construction module 130 is connected to the projection correction module 120 and receives the corrected feature vector output by it. The main function of the topology construction module 130 is to use the physical environment information carried in the data to quantify the logical dependency strength between the list items under specific working conditions, thereby generating an adjacency matrix. The adjacency matrix is ​​a data structure that reflects the logical relationship between processes, and the numerical state of its internal elements is dynamically adjusted with changes in environmental parameters.

[0044] The topology building module 130 parses each element in the modified feature vector set. For any two list item nodes... and (in Not equal to The topology building module 130 first starts from its corresponding modified feature vector. and Extract the environmental components. Record the list of sub-items. The environmental component is List of items The environmental component is The environmental component is the numerical component corresponding to the environmental condition component in the initial multidimensional feature vector defined in the feature mapping module 110 after being corrected by the projection correction module 120. This environmental component occupies a specific dimensional index in the vector space and clearly corresponds to the numerical information of geological and meteorological parameters.

[0045] Topology building module 130 calculates the environmental coupling coefficient between these two environmental components. The environmental coupling coefficient measures the consistency or interaction strength of two process nodes under the current physical environment. In this embodiment, the cosine similarity algorithm is used to calculate the environmental coupling coefficient to reflect the geometric similarity of their response patterns to the environment. The calculation formula is as follows: ; in, This represents the vector dot product operation; This represents the L2 norm of a vector. The numerical range is [-1, 1].

[0046] After obtaining the environmental coupling coefficient, the topology building module 130 calculates the list items in conjunction with the pre-set logical parameter library. To list sub-items Connection weights The pre-defined logical parameter library is a structured data collection stored in the system memory. It contains pairwise association attribute parameters between standard work process categories defined in the engineering quota specifications. This logical parameter library stores the environmental sensitivity coefficient corresponding to each pair of work process category combinations in key-value pairs. and inherent logical bias The connection weight represents the weight of the list item nodes under the current environmental conditions. The existence of the construction activities represented leads to the item nodes in the bill of quantities. The conditional probability that the construction activity will inevitably occur is represented. The calculation process introduces a nonlinear activation function to integrate environmental influences with inherent technological logic. Connection weights... The calculation formula is defined as follows: ; in, Defined as an environmental sensitivity coefficient, it is obtained by calling a pre-defined logical parameter library. This environmental sensitivity coefficient is a scalar used to quantify the relationship between a specific inventory sub-item node and the target data. The degree to which the logical relationships between items are affected by environmental factors. For example, the item node representing "excavation of the foundation pit" in the bill of quantities. Points to the item node in the inventory representing "excavation pit dewatering". The logical relationship, its Larger values ​​indicate that the dependency is highly influenced by rainfall and groundwater level data (i.e., environmental components); while for the list item node representing "masonry"... Points to the list item node representing "plastering" The logical relationship, its The value is relatively small.

[0047] Defined as an inherent logical bias, it is also obtained by calling a pre-defined logical parameter library, representing the list item node under standard environmental conditions. Point to list sub-item node The strength of basic dependencies. If the list item node... It is a list of sub-item nodes For mandatory prerequisite nodes (such as formwork for casting), then If the two are large positive values, then... It is a relatively large negative value.

[0048] is the base of the natural logarithm.

[0049] Topology building module 130 calculates the connection weights. Construct an adjacency matrix. This topology construction module 130 is configured with a connectivity threshold. (For example The topology building module 130 iterates through all possible list item node pairs. The calculated connection weights Connectivity discrimination threshold Compare them.

[0050] when At that time, determine the sub-item node of the list. With list sub-item nodes If there is a valid logical dependency path, the system will mark the corresponding element in the adjacency matrix as connected (i.e., there is a path from the list of sub-item nodes). Point to list sub-item node (Logical pointer).

[0051] when If the system determines that the two have no direct logical connection under the current operating conditions, it will mark the corresponding element in the adjacency matrix as disconnected.

[0052] Through the above traversal and judgment process, the topology construction module 130 generates an adjacency matrix. Adjacency matrix elements The assignment rules are as follows: ; The adjacency matrix The adjacency matrix fully maps the process logic relationships of the project under specific geological and meteorological conditions. By quantifying the connection states, it eliminates theoretically possible but unnecessary redundant paths under actual working conditions, while activating implicit paths essential for specific harsh conditions. This provides accurate matrix data input for the transitive closure operation of the subsequent process derivation module 140. For the specific algorithm implementation of adjacency matrix storage and traversal, those skilled in the art can employ sparse matrix compressed storage format (CSR) and depth-first search (DFS) or breadth-first search (BFS) strategies, which will not be elaborated upon here.

[0053] See attached document Figures 1-3 The process derivation module 140 is connected to the topology construction module 130 and is used to receive the adjacency matrix. Furthermore, based on the matrix transitive closure algorithm, it uncovers hidden processes not listed in the bill of quantities but inherently present according to the current construction logic. The technical problem addressed by this process derivation module 140 is that traditional cost audits only focus on point-to-point checks of bill of quantities sub-items, making it difficult to identify omissions caused by breaks in the logical chain of processes. The process derivation module 140 uses mathematical calculations to extend local direct dependencies into global path reachability relationships, thereby discovering missing nodes hidden deep within the logical path.

[0054] Process derivation module 140 is based on adjacency matrix Perform the transitive closure operation. This is based on the input adjacency matrix. This only represents the direct logical dependencies (i.e., one-step reachability) between the list items. However, in real-world complex engineering networks, the constraints between processes often span multiple levels of nodes. The process derivation module 140 uses the Warshall algorithm to analyze the adjacency matrix. Perform iterative calculations to generate the reachability matrix. .set up This represents the total number of sub-item nodes in the list. This represents the reachability matrix when only intermediate nodes with numbers less than or equal to are allowed to pass through. Initial state equal to adjacency matrix The process derivation module 140 calculates according to the following Boolean logic iterative formula. : ; in, Represents the reachability matrix The Middle Line 1 The element values ​​of the column can be either 0 or 1; Represents the Boolean logic "OR" operation; This represents the Boolean logic AND operation; The value is traversed from 1 to... .go through After several iterations, the final reachability matrix is ​​obtained. The reachability matrix elements in This reveals the sub-item nodes of the inventory under the current specific environmental conditions. It is a list of sub-item nodes The Warshall algorithm consists of direct or indirect preceding steps, and there is a necessary logical path between them. For the specific programming implementation of the Warshall algorithm, those skilled in the art can use a triple loop structure, which will not be elaborated here.

[0055] The process derivation module 140 uses a pre-built standard process logic database to derive the set of implicit processes existing under the current environmental conditions for the bill of quantities items. The standard process logic database is a set of structured data stored in the system memory. It is constructed based on the national or industry-issued building engineering consumption quotas and construction specifications. It records the full element path relationships between each process node in the standard construction process in the form of a directed graph or adjacency list, especially clarifying the intermediate node information that must be included in the complex construction chain that spans multiple processes.

[0056] Process derivation module 140 targets the reachability matrix The text indicates that there are node pairs with logical paths (i.e., satisfying...). The list sub-nodes corresponding to the non-diagonal elements With list sub-item nodes ), retrieve from the standard process logic database the list of sub-item nodes To list sub-item node Standard logical path set Standard logical path set Defined in industry standards, this involves completing the process from the list of sub-items. To list sub-item node The collection of all intermediate process nodes that must be passed in the construction process.

[0057] The process derivation module 140 calculates the set of candidate implicit processes for the current project to be analyzed. For each of the above pairs of nodes that have a logical path... If the standard logical path set It contains the current bill of quantities index set. If a node does not exist in the process, it is marked as a candidate hidden process. The process derivation module 140 calculates the initial set of missing items according to the following formula: ; in, This represents the set of candidate implicit processes, which is the set of potential missing processes that have been initially screened based on logical accessibility but have not yet undergone environmental adaptability verification. The union operation represents the set operation; The difference operation represents the set difference; This represents the set of all listed item nodes in the bill of quantities for the project to be analyzed. This calculation mathematically identifies process nodes that are covered by the contextual logic pathways but do not appear in the actual bill of quantities.

[0058] To ensure that the derivation results strictly match the current physical conditions, the process derivation module has 140 pairs of... Each candidate node in the process undergoes an environment adaptability check. The process derivation module 140 retrieves each implicit process node. Corresponding environmental activation threshold vector (Environmental activation threshold vector) It is a set of multidimensional numerical limits used to define implicit process nodes. Critical environmental conditions that are physically necessary. For example, for implicit process nodes representing "strong drainage of the foundation pit". Its corresponding environmental activation threshold vector Trigger thresholds were set for "groundwater level > base elevation" and "recent rainfall > 50mm". The environmental activation threshold vector was then compared with the environmental components of the current project. A comparison is performed. A node is retained only if the current environmental component meets the physical existence condition of the process (i.e., the environmental value falls within the range defined by the activation threshold vector). The process derivation module 140 ultimately outputs a verified set of implicit processes. This process transforms static rule checking into dynamic environment-driven deduction, enabling the automatic discovery of auxiliary engineering content that is triggered by specific severe operating conditions but is easily overlooked when compiling the inventory list.

[0059] See attached document Figure 1 and Figure 3The verification module 150 is connected to the process derivation module 140 to verify the rationality of the derived implicit process set at the cost level. The verification module 150 constructs a multi-dimensional vector containing dimensions of labor costs, material costs, machinery costs, and management fees, mapping the price data in the bill of quantities and the resource consumption data of the physical processes to a unified metric space. It then evaluates the logical integrity of the project cost by calculating the modulus residual. This process aims to identify process combinations that, while topologically connected, contradict each other in terms of economic value logic, thereby ensuring that the final cost audit result achieves a dual closed loop of technical and economic logic.

[0060] The verification module 150 verifies each sub-item node in the list of items to be analyzed. The actual unit price vector magnitude is constructed and calculated. The verification module 150 extracts the bill of quantities sub-item nodes from the input bill of quantities data. The unit price composition of labor costs, material costs, machinery costs, and management fees is used to generate an actual unit price vector. The actual unit price vector is a The actual unit price vector is a multi-dimensional vector, where each dimension corresponds to cost components such as labor costs, material costs, machinery costs, and management fees. The verification module 150 calculates the Euclidean norm of this actual unit price vector, i.e., the magnitude of the actual unit price vector. The calculation formula is as follows: ; in, Indicates the list of sub-item nodes In the Actual quoted value across each cost dimension; This represents the total number of cost dimensions. The magnitude of the "amplitude" of the item in the price range was quantified.

[0061] Module 150 is for verification of list sub-item nodes. A set of implicit processes that are logically related The overall consumption modulus is calculated. The verification module 150 first uses the implicit process set output by the process derivation module 140. In the middle, filter out the sub-items that match the current list. A subset of implicit processes that have direct or indirect logical dependencies is denoted as The verification module 150 calls the preset consumption quota database to retrieve each hidden process node in the hidden process subset. The standard resource consumption is calculated and combined with the benchmark market price to generate a standard consumption vector. The consumption quota database is a standardized data set stored in the system memory. It is constructed based on national or local standards for construction engineering consumption quotas and market price information, and records in detail the basic data such as the number of man-days, quantities of various materials, and machine shifts required for each standard construction procedure per unit of work. The verification module 150 superimposes the standard consumption vectors of all implicit procedures in this subset to obtain the total consumption vector, and calculates its modulus, i.e., the comprehensive consumption modulus. The calculation formula is as follows: ; in, Representing implicit process nodes The corresponding standard consumption vector has the same dimension definition as the actual unit price vector. Maintain consistency; This represents the accumulation operation of vectors. Quantification was used to support the list of sub-item nodes. The total theoretical value of the implicit physical resources required to complete the construction.

[0062] After obtaining the two modulus indices mentioned above, the verification module 150 calculates the residual between them. And based on this residual, assess the logical integrity of the project cost. Residual This reflects the degree of deviation between the explicit pricing and implicit costs of each item in the inventory. The calculation formula is as follows: ; in, This is the efficiency conversion coefficient, used to correct the conversion relationship between theoretical consumption and actual price under different construction technology levels. This efficiency conversion coefficient is obtained by statistical regression based on settlement data of similar historical projects.

[0063] The verification module 150 is configured with an integrity determination threshold range. .in, Defined as the lower limit deviation threshold of the cost logic, it is used to define the critical point where the price is too low and cannot cover the hidden costs; Defined as the upper limit deviation threshold for cost logic, it is used to identify the critical point where the quotation is inflated or hidden processes are not fully identified. The verification module 150 uses the residual... The numerical values ​​are used to evaluate the logical integrity of the project cost as follows: like Located in the interval Within this range, it indicates that the actual unit price vector magnitude and the overall consumption magnitude are in a balanced state, meaning that the pricing logic of the bill of quantities item covers the necessary implicit process consumption. The system determines that the logic completeness of this part is satisfactory, and then sets up the implicit process. The corresponding process is identified as a valid supplementary process and incorporated into the final cost analysis report.

[0064] like This indicates that the actual unit price vector magnitude is significantly smaller than the overall consumption magnitude, meaning the quoted price for a bill of quantities item is insufficient to support the consumption of its associated implicit processes, posing a risk of omissions or logical breaks. The system determines that logical integrity is compromised, generates a cost inversion warning signal, and outputs suggestions to adjust the attribution of implicit processes or review the bill of quantities unit prices.

[0065] like This indicates that the actual unit price vector magnitude is significantly greater than the overall consumption magnitude, suggesting possible over-quotation or insufficient derivation of hidden processes (there may still be undiscovered consumption items). The system also determines that the logical integrity is questionable and marks it as requiring manual review.

[0066] Through the above steps, the verification module 150 completes the mapping verification from physical consumption to value dimension, ensuring that the implicit procedures output by the system are not only necessary in the construction logic, but also self-consistent in the cost composition logic.

[0067] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A big data-based multi-dimensional intelligent analysis and processing system and method for engineering cost, characterized in that, The method comprises the following steps: a feature mapping module is configured to obtain bill of quantities, construction logs and environmental data of a project to be analyzed, and map list items in the bill of quantities into initial multi-dimensional feature vectors; a projection rectification module is configured to perform sparse representation and orthogonal projection on the initial multi-dimensional feature vectors by using a preset over-complete dictionary matrix, and generate modified feature vectors after filtering noise components; a topology construction module is configured to calculate connection weights between list items according to environmental components in the modified feature vectors, and construct an adjacency matrix reflecting process logic relationships; a process derivation module is configured to perform a transitive closure operation based on the adjacency matrix, and derive a set of implicit processes in which the list items exist under current environmental conditions; a verification module is configured to calculate a vector module length of actual unit prices of the list items and a comprehensive consumption module length of the set of implicit processes, and evaluate the logical integrity of project cost based on a residual error between the vector module length of actual unit prices and the comprehensive consumption module length. 2.The big data-based multi-dimensional intelligent analysis and processing system and method for engineering cost according to claim 1, wherein, In the feature mapping module, the step of mapping list items in the bill of quantities into initial multi-dimensional feature vectors comprises: establishing an index of the list items, and generating the initial multi-dimensional feature vectors formed by cascading material attribute components, process attribute components, environmental condition components and time dimension components. 3.The big data-based multi-dimensional intelligent analysis and processing system and method for engineering cost according to claim 1, wherein, In the projection rectification module, the step of performing sparse representation and orthogonal projection on the initial multi-dimensional feature vectors by using a preset over-complete dictionary matrix, and generating modified feature vectors after filtering noise components comprises: constructing and solving an L1 norm regularized least squares optimization problem to obtain a sparse coefficient vector of the initial multi-dimensional feature vectors under the over-complete dictionary matrix; determining a set of active atoms in the over-complete dictionary matrix according to non-zero elements in the sparse coefficient vector, and using the active atoms to span a local standard feature subspace; orthogonally projecting the initial multi-dimensional feature vectors into the local standard feature subspace to obtain the modified feature vectors. 4.The big data-based multi-dimensional intelligent analysis and processing system and method for engineering cost according to claim 1, wherein, In the topology construction module, the step of calculating connection weights between list items according to environmental components in the modified feature vectors comprises: extracting the environmental components from the modified feature vectors corresponding to any two list item nodes; calculating an environmental coupling coefficient between two environmental components by using a cosine similarity algorithm; calling a preset logic parameter library to obtain an environmental sensitivity coefficient and an inherent logic bias between two list item nodes; calculating the connection weights according to the environmental sensitivity coefficient, the environmental coupling coefficient and the inherent logic bias by using a nonlinear activation function. 5.The big data-based multi-dimensional intelligent analysis and processing system and method for engineering cost according to claim 4, characterized in that, In the topology construction module, the step of constructing an adjacency matrix reflecting process logic relationships comprises: setting a connectivity discrimination threshold, traversing all possible list item node pairs, and comparing the calculated connection weights with the connectivity discrimination threshold; when the connection weight is greater than the connectivity discrimination threshold, it is determined that there is an effective logical dependency path between two list item nodes, and the corresponding element in the adjacency matrix is marked as a connected state. When the connection weight is less than or equal to the connectivity discrimination threshold, it is determined that there is no direct logical association between two list subgoal nodes under the current working condition, and the corresponding element in the adjacency matrix is marked as disconnected; The adjacency matrix is generated according to the traversal and determination results. 6.The big data-based multi-dimensional intelligent analysis and processing system and method for engineering cost according to claim 5, wherein, The step of performing a transitive closure operation based on the adjacency matrix in the procedure derivation module comprises: The adjacency matrix is received, and the Warshall algorithm is used to iteratively calculate the adjacency matrix; An accessibility matrix is generated according to a Boolean logic iterative formula, and elements in the accessibility matrix reveal whether there is a direct or indirect pre-process logical path between the list subgoals under the current environmental condition. 7.The big data-based multi-dimensional intelligent analysis and processing system and method for engineering cost according to claim 6, characterized in that, The step of deriving the implicit procedure set of the list subgoal under the current environmental condition in the procedure derivation module comprises: For the node pair in the accessibility matrix indicating that there is a logical path, the standard logic path set between the two list subgoal nodes in the pre-set standard process logic database is retrieved; The difference set between the standard logic path set and the current bill of quantities index set is calculated, the nodes in the difference set are marked as candidate implicit procedures, and a candidate implicit procedure set is generated; The environmental activation threshold vector corresponding to each candidate implicit procedure is retrieved, and the environmental activation threshold vector is compared with the environmental component of the current project; The candidate implicit procedures that meet the physical existence condition are retained, and the verified implicit procedure set is generated. 8.The big data based multi-dimensional intelligent analysis and processing system and method for engineering cost according to claim 1, wherein, The step of calculating the actual unit price vector module length of the list subgoal and the comprehensive consumption module length of the implicit procedure set in the verification module comprises: The labor cost, material cost, mechanical cost, and management cost unit prices of the list subgoal are extracted from the bill of quantities to generate an actual unit price vector, and the Euclidean norm of the actual unit price vector is calculated to obtain the actual unit price vector module length; An implicit procedure subset having a logical dependency relationship with the list subgoal is screened out, a pre-set consumption quantity quota database is called to retrieve the standard resource consumption quantity of each implicit procedure in the implicit procedure subset and generate a standard consumption vector; The standard consumption vectors of all implicit procedures in the implicit procedure subset are superimposed to obtain a total consumption vector, and the module length of the total consumption vector is calculated to obtain the comprehensive consumption module length. 9.The big data-based multi-dimensional intelligent analysis and processing system and method for engineering cost according to claim 8, wherein, The step of evaluating the logical integrity of the project cost based on the residual error between the actual unit price vector module length and the comprehensive consumption module length in the verification module comprises: The comprehensive consumption module length is corrected using an ergonomics conversion coefficient, the difference between the actual unit price vector module length and the corrected comprehensive consumption module length is calculated to obtain the residual error; An integrity determination threshold interval including a lower limit deviation threshold of cost logic and an upper limit deviation threshold of cost logic is set; When the residual error is within the integrity determination threshold interval, it is determined that the logical integrity is qualified; when the residual error is less than the lower limit deviation threshold of cost logic, it is determined that there is a risk of omission or logical breakage; When the residual error is greater than the upper limit deviation threshold of cost logic, it is determined that there is an overestimated price or an insufficient implicit procedure derivation.

10. A method for multi-dimensional intelligent analysis and processing of engineering costs based on big data, applied to the multi-dimensional intelligent analysis and processing system for engineering costs based on big data as described in any one of claims 1-9, characterized in that, The method comprises the following steps: The feature mapping module is configured to acquire bill of quantities, construction log and environmental data of a project to be analyzed, and map sub-entries in the bill of quantities into initial multi-dimensional feature vectors; The projection rectification module is configured to call a pre-set over-complete dictionary matrix to perform sparse decomposition and orthogonal projection reconstruction on the initial multi-dimensional feature vectors, and generate modified feature vectors with noise components filtered out; The topology construction module is configured to extract environmental components from the modified feature vectors, calculate connection weights between the sub-entries according to the environmental components, and construct an adjacency matrix reflecting the logical relationship between the processes based on the connection weights; The process derivation module is configured to perform a transitive closure operation based on the adjacency matrix, and derive a set of implicit processes in which the sub-entries exist under the current environmental conditions; The verification module is configured to calculate a module length of an actual unit price vector of the sub-entries and a module length of comprehensive consumption of the set of implicit processes, and evaluate the logical integrity of the project cost based on a residual error between the module length of the actual unit price vector and the module length of the comprehensive consumption.

Images