Bridge engineering full-period cost hyperbranched early warning and bidding and tendering collaborative management and control method and device

Abstract

The invention discloses a bridge engineering full-period cost hyperbranched early warning and bidding and tendering collaborative management and control method and device. Comprising the following steps: S1, full-period multi-dimensional data acquisition; s2, data preprocessing and feature engineering; s3, constructing a collaborative control and early warning model; s4, performing collaborative management and control in a bidding and tendering stage; and S5, performing dynamic early warning and management and control adjustment in the construction and operation and maintenance stages. By constructing a distributed multi-dimensional data acquisition network and a standardized preprocessing process, static and dynamic data of the whole stage of bridge engineering are integrated, data barriers among the stages are broken, a comprehensive and real-time data basis is provided for cost early warning and collaborative management and control, the problem of one-sided data support in the prior art is solved, and the method is suitable for popularization and application. And the decision making scientificity is improved.

Description

Technical Field

[0001] This invention relates to the field of information technology for bridge engineering, specifically to a method and equipment for early warning of cost overruns and collaborative management of bidding and tendering in bridge engineering throughout its entire lifecycle. Background Technology

[0002] Bridge engineering, as a core component of transportation infrastructure, is characterized by long construction cycles, numerous stages, high technical complexity, large capital investment, and dynamically changing influencing factors. The efficiency of cost control throughout the entire project lifecycle and the coordination in the bidding process directly determines the project's economic benefits and construction quality. Currently, cost control and bidding management techniques in the bridge engineering field generally suffer from stage-specific fragmentation and insufficient coordination, specifically manifested in the following aspects: 1. Insufficient data coverage and integration in the bridge engineering industry: Existing technologies mostly focus on a single stage (such as cost monitoring during the construction phase or bidding process management), lacking the ability to integrate data throughout the entire lifecycle of design, bidding, construction, and operation and maintenance. Moreover, data collection is mostly limited to local dimensions and does not include key dynamic data such as dynamic prices in the building materials market, credit ratings of bidding entities, and policy and regulatory adjustments. This results in a lack of foresight in cost early warning and a one-sided basis for control decisions.

[0003] 2. Lack of coordination between bidding and cost control in the bridge engineering industry: Core parameters in bidding documents, such as the maximum bid limit, performance bond ratio, and payment milestones, are mostly statically set based on experience and fail to be linked and matched with the risk of cost overruns throughout the entire cycle. This makes it difficult for the bidding process to play a role in preventing risks in advance, and it is easy for subsequent problems such as insufficient performance capacity of the winning bidder and weak cost control capabilities to occur.

[0004] 3. Insufficient accuracy and adaptability of early warning models: Existing cost overrun early warning technologies mostly use a single algorithm (such as simple LSTM or traditional statistical models), which is weak in capturing non-periodic and sudden influencing factors such as sudden changes in building material prices and policy adjustments. There are problems such as delayed early warning, false early warning or missed early warning, which cannot provide reliable support for management and control decisions.

[0005] 4. Rigid and poorly coordinated control measures: When faced with the risk of cost overruns, existing control measures are mostly post-event remedies (such as additional investment and schedule adjustments), lacking a dynamic adjustment mechanism that is deeply linked to the bidding and contract performance terms. Furthermore, the "one-size-fits-all" review and control model is adopted for bidders with different credit ratings and risk levels, resulting in low control efficiency, untimely risk handling, and difficulty in effectively reducing overrun losses.

[0006] The aforementioned technical deficiencies result in a persistently high risk of cost overruns throughout the entire lifecycle of bridge engineering projects, making it difficult to guarantee the quality of bidding and tendering. This seriously affects the stability and economy of project construction, and there is an urgent need for a technical solution that can achieve full-cycle data connectivity, deep collaboration between bidding and cost early warning, accurate early warning, and dynamic management and control. Summary of the Invention

[0007] Therefore, to address the shortcomings in the aforementioned industry, this invention provides a method and equipment for early warning of cost overruns and collaborative management of bidding in bridge engineering throughout its entire lifecycle. This technology integrates static and dynamic data from all stages of bridge engineering by constructing a distributed, multi-dimensional data acquisition network and a standardized preprocessing process, breaking down data barriers between stages and providing a comprehensive and real-time data foundation for cost early warning and collaborative management, avoiding the one-sided data support of existing technologies. Simultaneously, by establishing a linkage mechanism between early warning results and bidding parameter optimization, as well as differentiated qualification review standards, it achieves dynamic adjustment of bidding document parameters and precise screening of bidding entities. Furthermore, through a deep linkage mechanism between tiered early warning and bidding performance clauses, it realizes a shift from "post-event remediation" to "pre-event prediction and dynamic adjustment during the event," and the differentiated management model balances management accuracy and efficiency, taking targeted measures for different risk levels, significantly improving the timeliness and effectiveness of risk disposal, significantly reducing cost overrun losses throughout the entire lifecycle of bridge engineering, and ensuring the economic efficiency and stability of project construction.

[0008] This invention is implemented by constructing a method for early warning of cost overruns and collaborative management of bidding processes throughout the entire lifecycle of bridge engineering projects, comprising the following steps: S1. Full-cycle multi-dimensional data collection: Construct a distributed data collection network covering the design, bidding, construction and operation and maintenance phases of bridge engineering, and collect design parameter data, dynamic price data of building materials market, credit data of bidding entities, construction progress data, operation and maintenance loss data and policy and regulation adjustment data. Among them, the credit data of bidding entities includes historical performance records, cost control capability rating and violation penalty records. S2. Data Preprocessing and Feature Engineering: The data collected in step S1 is processed by outlier removal, missing value completion and format standardization. Based on the bridge engineering cost influencing factor system, core feature variables are extracted, including the economic indicators of the design scheme, the fluctuation coefficient of building material prices, the deviation of bidding price and the waste rate of construction resources. S3. Construct a collaborative management and early warning model: Based on the improved LSTM-random forest fusion algorithm, input the core feature variables processed in step S2, and construct a cost overrun early warning sub-model and a bidding parameter optimization sub-model. The early warning sub-model determines the overrun threshold range through training with historical overrun cases, and the optimization sub-model establishes the mapping relationship between bidding parameters and the full-cycle cost. S4. Collaborative Management and Control Execution in the Bidding and Tendering Stage: The pre-overspending risk level of the early warning sub-model is used as the input parameter of the optimization sub-model. The maximum bid limit, performance bond ratio and payment node terms in the bidding documents are dynamically adjusted to generate differentiated bidding qualification review standards and strengthen qualification review for high-risk bidders. S5. Dynamic Early Warning and Control Adjustment during Construction and Operation and Maintenance Phases: Input actual cost data during the construction phase and loss data during the operation and maintenance phase into the early warning sub-model in real time. When the cost deviates from the over-cost threshold, a tiered early warning signal is triggered, and the performance control clauses during the bidding phase are linked to initiate adjustments to project payment, accountability for breach of contract, or supplementary bidding processes.

[0009] According to the method described in this invention, in step S1, the distributed data acquisition network adopts a collaborative architecture of edge computing nodes and cloud servers. The edge computing nodes are deployed at construction sites, building materials trading markets, and bidding platforms to realize real-time data acquisition and preliminary filtering, while the cloud servers are responsible for centralized data storage and synchronous updates.

[0010] According to the method of the present invention, in step S3, the improved LSTM-random forest fusion algorithm optimizes the feature extraction process of LSTM through an attention mechanism, enhances the ability to capture non-periodic factors such as sudden changes in building material prices and policy adjustments, and uses the random forest algorithm to correct the output results of LSTM, thereby reducing the prediction bias of a single algorithm.

[0011] According to the method of the present invention, in step S4, the differentiated bidding qualification review standard is formulated based on the matching result of the credit rating of the bidding entity and the risk level output by the early warning sub-model. The review process is simplified for bidding entities with a credit rating of A and a low risk level, while the technical solution defense and financial strength verification are added for bidding entities with a credit rating of C or below or a high risk level.

[0012] According to the method described in this invention, in step S5, the graded early warning signals are divided into Level 1, Level 2, and Level 3 early warnings, and the corresponding control measures triggered are as follows: Level 1 early warning initiates the submission mechanism for the cost deviation analysis report; Level 2 early warning adjusts the payment ratio of the project funds; and Level 3 early warning initiates the deduction of the performance bond and the supplementary bidding process.

[0013] A bridge engineering full-cycle cost overrun early warning and bidding collaborative management device, comprising: Data acquisition module: It adopts a distributed sensor array and interface adapter unit to collect design parameter data, dynamic price data of building materials market, credit data of bidding entities, construction progress data, operation and maintenance loss data and policy and regulation adjustment data throughout the entire life cycle of bridge engineering. Data preprocessing module: Communicates with the data acquisition module to perform outlier removal, missing value completion, and format standardization on the acquired data, and extracts core feature variables; Model building and computation module: It has a built-in improved LSTM-random forest fusion algorithm, which communicates with the data preprocessing module to build a cost overrun early warning sub-model and a bidding parameter optimization sub-model, and to perform early warning calculation and parameter optimization calculation. Bidding and tendering collaborative management module: It communicates with the model building and calculation module and is used to dynamically adjust the parameters of the bidding documents, generate differentiated bidding qualification review standards, and perform qualification review of bidding entities based on the early warning results; Early warning and adjustment module: It communicates with the model building and calculation module and the bidding and tendering collaborative management module to monitor the cost data of the whole cycle in real time, trigger the graded early warning signal, and link the execution of project payment adjustment, breach of contract accountability or supplementary bidding process; Storage and Interaction Module: Communicates with the above modules to store data, model parameters, and control records, and provides a human-machine interface for users to view early warning information and control instructions.

[0014] According to the device of the present invention, the interface adaptation unit of the data acquisition module supports protocol docking with public service platforms for bidding and tendering, e-commerce platforms for building materials, construction progress management systems and operation and maintenance monitoring systems, and is compatible with multiple data formats such as XML, JSON and CSV.

[0015] According to the device of the present invention, the model building and calculation module further includes a model updating unit, which is used to periodically update the parameters of the early warning sub-model and the optimization sub-model based on the newly added full-cycle cost data and bidding case data, so as to improve the adaptability and prediction accuracy of the model.

[0016] According to the device of the present invention, the early warning and adjustment module further includes an early warning push unit, which pushes early warning information and control instructions to project managers, bidding supervision departments and bidding entities through three methods: SMS, platform messages and email.

[0017] The technical problem solved by this invention, in contrast to existing technologies in the industry, is as follows: To address the issues of incomplete full-cycle data coverage and weak connectivity in existing technologies, this paper proposes a solution for building a multi-dimensional, distributed data collection and integration system. This system enables real-time collection, standardized processing, and deep integration of key data across all stages of design, bidding, construction, and operation and maintenance, providing comprehensive data support for collaborative management and early warning.

[0018] To address the lack of coordination between bidding and cost control, this paper proposes a solution to establish a dynamic mapping relationship between bidding parameters and the risk of cost overruns throughout the entire project lifecycle. This would enable precise optimization of core parameters in bidding documents and differentiated qualification review for bidders, making the bidding process a key node for proactive cost risk control.

[0019] To address the issues of existing early warning models' insufficient capture of non-periodic factors and low accuracy, this paper proposes optimizing the algorithm architecture to improve the model's responsiveness to dynamic factors such as sudden changes in building material prices and policy adjustments, reduce prediction bias, and achieve accurate and timely early warning of the risk of cost overruns.

[0020] To address the issues of rigid and poorly coordinated control measures, this paper proposes a dynamic adjustment mechanism that deeply integrates tiered early warning systems with bidding and tendering performance terms. This mechanism would seamlessly connect risk warning signals with control measures such as adjustments to project payments, deductions of performance bonds, and supplementary bidding, thereby enhancing the targetedness and effectiveness of risk management.

[0021] The present invention has the following advantages: 1. Enables seamless and precise data support throughout the entire lifecycle: By constructing a distributed, multi-dimensional data acquisition network and a standardized preprocessing process, it integrates static and dynamic data from all stages of bridge engineering, breaks down data barriers between stages, provides a comprehensive and real-time data foundation for cost early warning and collaborative management, solves the problem of one-sided data support in existing technologies, and improves the scientific nature of decision-making.

[0022] 2. Enhances the synergistic efficiency of bidding and cost control: By establishing a linkage mechanism between early warning results and bidding parameter optimization, as well as differentiated qualification review standards, dynamic adjustment of bidding document parameters and accurate screening of bidding entities are achieved. This enables the bidding process to avoid high-risk entities and potential cost overruns in advance, solving the problem of lack of synergy between the two and significantly improving the quality of bidding and the proactive nature of cost control.

[0023] 3. Significantly improves the accuracy of cost overrun warnings: The LSTM-random forest fusion algorithm, optimized based on the attention mechanism, enhances the ability to capture non-periodic and sudden influencing factors. At the same time, the algorithm complementarity reduces the prediction bias of a single model. Compared with existing single algorithm models, the warning accuracy is improved by more than 30%, effectively solving problems such as false warnings and missed warnings, and providing a reliable basis for risk management.

[0024] 4. Enables dynamic and precise risk management: Through a deep linkage mechanism between tiered early warning and bidding performance clauses, the risk management model has shifted from "post-event remediation" to "pre-event prediction and dynamic adjustment during the event." The differentiated management model balances the accuracy and efficiency of management, and takes targeted measures for different risk levels, which greatly improves the timeliness and effectiveness of risk disposal, significantly reduces the cost overrun losses of bridge projects throughout the entire cycle, and ensures the economy and stability of project construction. Attached Figure Description

[0025] Figures 1-3 This is a flowchart illustrating the method described in the invention; Figure 4 This is a schematic diagram of the device described in this invention. Detailed Implementation

[0026] The following will be combined with the appendix Figures 1-4 This invention will be described in detail, and the technical solutions in the embodiments of this invention will be clearly and completely described. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0027] Example 1; This invention provides a method for early warning of cost overruns and collaborative management of bidding processes throughout the entire lifecycle of bridge engineering projects, such as... Figures 1-3 As shown, it includes the following steps: S1. Full-cycle multi-dimensional data collection: Construct a distributed data collection network covering the design stage, bidding stage, construction stage and operation and maintenance stage of bridge engineering, and collect design parameter data, dynamic price data of building materials market, credit data of bidding entities, construction progress data, operation and maintenance loss data and policy and regulation adjustment data. Among them, the credit data of bidding entities includes historical performance records, cost control capability rating and violation penalty records. S2. Data Preprocessing and Feature Engineering: The data collected in step S1 is processed by outlier removal, missing value completion and format standardization. Based on the bridge engineering cost influencing factor system, core feature variables are extracted, including the economic indicators of the design scheme, the fluctuation coefficient of building material prices, the deviation of bidding price and the waste rate of construction resources. S3. Construct a collaborative management and early warning model: Based on the improved LSTM-random forest fusion algorithm, input the core feature variables processed in step S2, and construct a cost overrun early warning sub-model and a bidding parameter optimization sub-model. The early warning sub-model determines the overrun threshold range through training with historical overrun cases, and the optimization sub-model establishes the mapping relationship between bidding parameters and the full-cycle cost. S4. Collaborative Management and Control Execution in the Bidding and Tendering Stage: The pre-overspending risk level of the early warning sub-model is used as the input parameter of the optimization sub-model. The maximum bid limit, performance bond ratio and payment node terms in the bidding documents are dynamically adjusted to generate differentiated bidding qualification review standards and strengthen qualification review for high-risk bidders. S5. Dynamic Early Warning and Control Adjustment during Construction and Operation and Maintenance Phases: Input actual cost data during the construction phase and loss data during the operation and maintenance phase into the early warning sub-model in real time. When the cost deviates from the over-cost threshold, a tiered early warning signal is triggered, and the performance control clauses during the bidding phase are linked to initiate adjustments to project payment, accountability for breach of contract, or supplementary bidding processes.

[0028] In step S1 of the method described in this embodiment, the distributed data acquisition network adopts a collaborative architecture of edge computing nodes and cloud servers. The edge computing nodes are deployed at construction sites, building materials trading markets, and bidding platforms to realize real-time data acquisition and preliminary filtering, while the cloud servers are responsible for centralized data storage and synchronous updates.

[0029] In step S3 of the method described in this embodiment, the improved LSTM-random forest fusion algorithm optimizes the feature extraction process of LSTM through an attention mechanism, enhances the ability to capture non-periodic factors such as sudden changes in building material prices and policy adjustments, and uses the random forest algorithm to correct the output results of LSTM, thereby reducing the prediction bias of a single algorithm.

[0030] In step S4 of the method described in this embodiment, the differentiated bidding qualification review standard is formulated based on the matching result of the credit rating of the bidding entity and the risk level output by the early warning sub-model. The review process is simplified for bidding entities with a credit rating of A and a low risk level, while the technical solution defense and financial strength verification are added for bidding entities with a credit rating of C or below or a high risk level.

[0031] In step S5 of the method described in this embodiment, the graded early warning signal is divided into a first-level early warning, a second-level early warning, and a third-level early warning. The corresponding control measures triggered are as follows: a first-level early warning initiates the submission mechanism for a cost deviation analysis report; a second-level early warning adjusts the payment ratio of the project funds; and a third-level early warning initiates the deduction of the performance bond and a supplementary bidding process.

[0032] Example 2; A device for early warning of cost overruns and collaborative management of bidding for bridge engineering throughout its entire lifecycle, such as... Figure 4 As shown, the device comprises: Data acquisition module: It adopts a distributed sensor array and interface adapter unit to collect design parameter data, dynamic price data of building materials market, credit data of bidding entities, construction progress data, operation and maintenance loss data and policy and regulation adjustment data throughout the entire life cycle of bridge engineering. Data preprocessing module: Communicates with the data acquisition module to perform outlier removal, missing value completion, and format standardization on the acquired data, and extracts core feature variables; Model building and computation module: It has a built-in improved LSTM-random forest fusion algorithm, which communicates with the data preprocessing module to build a cost overrun early warning sub-model and a bidding parameter optimization sub-model, and to perform early warning calculation and parameter optimization calculation. Bidding and tendering collaborative management module: It communicates with the model building and calculation module and is used to dynamically adjust the parameters of the bidding documents, generate differentiated bidding qualification review standards, and perform qualification review of bidding entities based on the early warning results; Early warning and adjustment module: It communicates with the model building and calculation module and the bidding and tendering collaborative management module to monitor the cost data of the whole cycle in real time, trigger the graded early warning signal, and link the execution of project payment adjustment, breach of contract accountability or supplementary bidding process; Storage and Interaction Module: Communicates with the above modules to store data, model parameters, and control records, and provides a human-machine interface for users to view early warning information and control instructions.

[0033] In the device described in this embodiment, the interface adaptation unit of the data acquisition module supports protocol interfacing with public service platforms for bidding and tendering, e-commerce platforms for building materials, construction progress management systems, and operation and maintenance monitoring systems, and is compatible with multiple data formats such as XML, JSON, and CSV.

[0034] In the device described in this embodiment, the model building and calculation module further includes a model update unit, which is used to periodically update the parameters of the early warning sub-model and the optimization sub-model based on the newly added full-cycle cost data and bidding case data, so as to improve the adaptability and prediction accuracy of the model.

[0035] In the device described in this embodiment, the early warning and adjustment module further includes an early warning push unit, which pushes early warning information and control instructions to project managers, bidding supervision departments and bidding entities through three methods: SMS, platform messages and email.

[0036] The following is a principle explanation of the innovative part of this application: I. In terms of existing technological pain points Currently, there is a significant disconnect between cost control and bidding management in the field of bridge engineering, and existing technologies generally have the following shortcomings: Data coverage limitations: Most technologies only focus on cost monitoring during the construction phase or process management in a single bidding process, failing to achieve data integration across the entire lifecycle of design, bidding, construction, and operation and maintenance, resulting in delayed early warnings and a lack of forward-looking control. Insufficient coordination: The setting of bidding parameters is disconnected from the subsequent cost overrun risk. The price limit and security deposit clauses in the bidding documents are mostly based on experience and cannot be dynamically adjusted according to the cost risk throughout the entire cycle. Low accuracy of early warning models: Existing early warning technologies mostly use a single algorithm, which is weak in capturing non-cyclical factors such as sudden changes in building material prices and policy adjustments, and is prone to false or missed warnings. Rigid control measures: When faced with the risk of cost overruns, there is a lack of dynamic adjustment mechanisms linked to bidding and contract performance terms, and the control measures are simplistic and have low execution efficiency.

[0037] II. Core Creative Aspects of this Plan (I) Innovation in Full-Lifecycle Data Fusion and Collaborative Architecture: This application breaks through the limitations of existing technologies that separate different stages, constructing a distributed data acquisition network covering the entire lifecycle of bridge engineering. For the first time, it deeply integrates multi-dimensional data such as design parameters, building material prices, bidding credit, construction progress, and operation and maintenance losses. Simultaneously, it establishes a closed-loop collaborative architecture of "data acquisition - model calculation - bidding control - dynamic adjustment." This architecture achieves precise matching between parameter settings during the bidding stage and subsequent full-lifecycle cost risks, solving the core problem of "two separate systems for bidding and cost control" in existing technologies, and realizing an upgrade from "post-event remediation" to "pre-event prediction and in-event control."

[0038] (II) Technological Breakthrough of the Improved Fusion Algorithm: This application adopts an improved LSTM-Random Forest fusion algorithm. It optimizes the feature extraction capability of LSTM through an attention mechanism, enhancing the capture of non-periodic influencing factors. Simultaneously, it utilizes the random forest algorithm to correct the prediction bias of the single LSTM model. Compared to existing single-algorithm models, the early warning accuracy is improved by more than 30%. This algorithm innovatively and specifically addresses the problem of inaccurate early warnings caused by the complexity and dynamic nature of factors influencing bridge engineering costs, providing reliable technical support for collaborative management and control.

[0039] (III) Innovative Design of Differentiated Bidding and Tendering Control Mechanism: This application proposes for the first time a differentiated control mechanism based on the risk level of cost overruns and the credit rating of bidding entities. By dynamically adjusting core parameters such as the price limit, bid bond ratio, and payment schedule in the tender documents, and by implementing tiered qualification reviews, it achieves precise and personalized bidding and tendering control. This mechanism avoids the inefficiency of the traditional "one-size-fits-all" review model and mitigates the performance risks of high-risk entities in advance through risk-linked clauses, significantly improving the quality of bidding and tendering and the effectiveness of full-cycle cost control.

[0040] (iv) Technological Innovation in Full-Cycle Dynamic Early Warning and Linked Adjustment: This application establishes a tiered early warning and deeply linked adjustment mechanism with bidding and tendering performance terms. When a risk of cost overrun is detected, it can not only trigger multi-level early warnings but also directly initiate control measures such as adjusting project payments and deducting performance bonds, achieving seamless integration of risk response and control execution. Compared with the problem of disconnect between early warning and control in existing technologies, this mechanism significantly improves the timeliness and effectiveness of risk handling and effectively reduces losses from cost overruns.

[0041] Figures 1-3 This is the execution flowchart corresponding to this application; the execution flowchart uses method steps as a framework to intuitively present the entire closed loop of "data collection - model calculation - collaborative management and control - dynamic early warning", and clearly defines key judgment nodes and differentiated execution paths.

[0042] Figure 4 This is the system structure diagram of this application: The system structure diagram takes the corresponding device as the core, clearly presents the hierarchical relationship of each functional module, the data flow, and external interaction objects, highlighting the core architecture of "distributed acquisition - centralized processing - collaborative control - dynamic feedback". It clarifies the composition of each module (such as the interface adaptation unit of the data acquisition module and the update unit of the model building module), the communication relationships between modules, and the external interaction objects, intuitively demonstrating the innovative architecture of "distributed acquisition - centralized processing - collaborative control".

[0043] The specific implementation process and engineering examples of this application are described below; I. Specific Implementation Process: The implementation process of this application's solution is based on the core logic of "data-driven - model calculation - collaborative management and control - dynamic feedback," combining hardware deployment and software algorithm implementation to complete the full-cycle cost overrun early warning and collaborative management and control of bidding in stages. The specific steps are as follows: (I) Preliminary preparations and equipment deployment Hardware Deployment: A distributed data acquisition network is established, deploying edge computing nodes (such as industrial-grade edge gateways EG900) and sensor arrays (including building material price acquisition terminals, construction progress GPS positioning sensors, and maintenance structural stress sensors) at bridge engineering design unit terminals, bidding and tendering public service platform interfaces, and construction sites. Simultaneously, a cloud server (adopting a dual-machine hot standby architecture, configured with an Intel Xeon Gold 6330 processor and 2TB SSD storage) is deployed at the project management center. The interface adapter unit of the data acquisition module is integrated with external systems (bidding and tendering platforms, building material e-commerce platforms, etc.) to ensure compatibility with XML, JSON, and CSV data formats.

[0044] Software Setup: Deploy core functional modules on a cloud server, including a data preprocessing module (integrating Python's Pandas library for data cleaning), a model building and computation module (developing and improving an LSTM-random forest fusion algorithm engine based on the TensorFlow framework, setting the number of LSTM hidden layers to 3 and the number of random forest decision trees to 100), a bidding and tendering collaborative management module, and an early warning and adjustment module. At the same time, build the database (using a MySQL cluster) and human-computer interaction interface (B / S architecture, supporting multi-terminal access) for the storage and interaction module.

[0045] Data and Model Initialization: Collect full-cycle historical data (design schemes, cost records, bidding cases, operation and maintenance reports, etc.) of similar bridge projects over the past 10 years, as well as the current project's basic design parameters, regional building material market benchmark prices, policy and regulatory documents, etc., and complete data initialization and storage. Use historical cost overrun case data to train a cost overrun early warning sub-model, and determine the overrun threshold ranges for Level 1, Level 2, and Level 3 early warnings as [5%, 10%), [10%, 20%), and ≥20%, respectively; through correlation analysis between bidding parameters and historical costs, establish the mapping relationship of the bidding parameter optimization sub-model.

[0046] (II) Implement the operation in stages throughout the entire cycle 1. Implementation during the design and bidding phase Data Acquisition and Preprocessing: The system collects design data such as structural parameters and load standards from design units in real time via a distributed network; real-time price data of core building materials such as steel bars and concrete from the building materials market; and credit ratings of bidding entities (obtained from third-party credit agency databases) and historical performance records. The data preprocessing module automatically removes abnormal data (such as building material price data exceeding three standard deviations), uses linear interpolation to complete missing progress data, and extracts core characteristic variables such as economic indicators of the design scheme (e.g., cost per unit area), building material price fluctuation coefficients, and bid price deviation after standardization.

[0047] Model computation and parameter optimization: Core feature variables are input into the early warning sub-model to predict the cost overrun risk level under the current design scheme and potential bid price combination. If the predicted risk level is high, the optimization sub-model automatically adjusts the bidding document parameters: the maximum bid limit is reduced by 3%, the performance bond ratio is increased from 5% to 8%, and the payment schedule is changed from monthly to payment according to project progress milestones (completion of main structure, completion of bridge deck paving, etc.). Simultaneously, differentiated qualification review standards are generated. For bidding companies with a credit rating of A and a low risk level, the review process is simplified (only basic qualification documents are required), while for companies with a credit rating of C and below, a technical solution defense and verification of project fund flows over the past three years are added.

[0048] Bidding and tendering control execution: The bidding and tendering collaborative control module pushes the revised bidding documents to the bidding and tendering platform, automatically reviews the materials submitted by bidding entities for eligibility, and screens out qualified bidders. During the bid evaluation process, an early warning model is simultaneously invoked to conduct a secondary assessment of the potential cost overrun risk of each bid, serving as a supplementary basis for bid evaluation.

[0049] 2. Implementation during the construction phase Real-time data monitoring: Construction progress data (such as daily concrete pouring volume) and actual cost expenditure data (material procurement costs, labor costs, etc.) are collected in real time through edge computing nodes at the construction site and uploaded to the cloud server. The data preprocessing module cleans and extracts features from the collected data daily, updating variables such as building material price fluctuation coefficients and construction resource waste rates.

[0050] Dynamic early warning and control adjustment: The early warning sub-model receives updated feature variables in real time. If a cost deviation from the over-budget threshold is detected, the corresponding level of early warning is triggered. Level 1 Warning (5%-10% Overspending): The warning push unit notifies project management personnel via SMS and platform messages, activates the cost deviation analysis report submission mechanism, and requires the construction unit to submit the cause of the deviation and rectification plan within 3 days. Level 2 warning (10%-20% overspending): The payment ratio for the project will be automatically adjusted from the original 80% to 60%, and the construction unit will be required to submit supplementary performance guarantees. Level 3 warning (overspending ≥ 20%): Initiate the performance bond deduction process (deducting 50% of the overspending amount) and trigger a supplementary bidding process to re-bid for the sub-projects with severe overspending.

[0051] 3. Implementation during the operation and maintenance phase Operation and maintenance data collection and analysis: The operation and maintenance monitoring system collects bridge structure loss data (such as the degree of steel corrosion and the damaged area of ​​the bridge deck) and operation and maintenance cost data (repair material costs, labor costs, etc.), and inputs them into the early warning sub-model to continuously monitor whether the total cost throughout the entire cycle exceeds the expected total cost.

[0052] Model Iteration and Subsequent Management: The model update unit incorporates newly added operation and maintenance data and management records into the training set, and iteratively optimizes the parameters of the early warning sub-model and the optimization sub-model regularly (quarterly). If cost overruns occur during the operation and maintenance phase, the performance clauses of the construction phase will be linked to initiate a breach of contract accountability process against the responsible party, and relevant records will be synchronized to the bidding entity's credit database as a basis for reviewing the eligibility of bidding for similar projects in the future.

[0053] (III) Final Stage Operations After project completion and the end of the operation and maintenance warranty period, the storage and interaction module generates a full-cycle cost control report, covering data records at each stage, early warning triggering situations, and the effectiveness of control measures. Simultaneously, the project data is archived to the database to provide case support for the management of similar bridge projects in the future. The system then enters standby mode, awaiting the start of the next project.

[0054] II. Application Example: Taking a cross-river bridge project in a certain city as an application case, the total budget for this project is 5 billion yuan, the construction period is 5 years, and the operation and maintenance warranty period is 10 years. The implementation effect after applying this patented solution is as follows: (I) Basic information about the project: The main bridge of the cross-river bridge is 2.8 kilometers long and adopts a cable-stayed bridge structure. It involves complex processes such as large steel structure processing and underwater foundation construction. The price of building materials is greatly affected by the fluctuation of the steel market. In addition, there are many potential bidders with significant differences in qualifications. Under the traditional management model, there are problems such as high risk of cost overrun and low efficiency of bidding and screening.

[0055] (II) Application process and effects of the solution: During the bidding stage: Data acquisition modules were used to obtain steel market price fluctuation data for the past six months (fluctuation range reaching 12%) and credit data of 15 bidding companies. Among them, three companies had a credit rating of C and an average historical project cost overrun rate of 18%. The early warning sub-model predicted a high risk of cost overrun if traditional bidding document parameters were used. The optimization sub-model then adjusted its parameters: the maximum bid price was lowered from 5 billion yuan to 4.85 billion yuan, the performance bond ratio was increased to 8%, and technical presentations and financial strength verification were added for the three high-risk companies. Ultimately, six high-quality bidding companies were selected, with a winning bid of 4.78 billion yuan, a 1.4% reduction from the previous price limit, effectively mitigating the performance risk of high-risk entities.

[0056] Construction Phase: In the second year of the project, a sudden surge in steel prices (monthly increase of 15%) triggered an 8% cost overrun by the early warning sub-model, establishing a Level 1 warning. The system immediately issued a warning, and the analysis report submitted by the construction unit indicated that increased steel procurement costs were the primary cause. Based on the warning recommendations, project management adjusted the steel procurement plan (changing from spot purchases to long-term agreements) and initiated a second round of supplier price comparisons. Within one month, the cost deviation was controlled to within 3%, preventing further cost overruns. During subsequent construction, heavy rains caused delays, resulting in a 2% cost increase. While the system did not trigger a warning, the construction plan was continuously monitored and adjusted to ensure the total cost did not exceed the bid price.

[0057] Operation and Maintenance Phase: In the third year of the warranty period, the bridge deck damage area exceeded expectations by 20%, and the operation and maintenance costs exceeded the budget by 6%, triggering a Level 1 warning in the system. By linking the performance clauses from the construction phase, it was discovered that the bridge deck paving process by the construction company had quality defects. The breach of contract accountability process was immediately initiated, with 30 million yuan of the construction company's performance bond deducted for bridge deck repairs. The company's quality breach of contract record was also synchronized to the credit database. Simultaneously, the model update unit optimized the warning parameters based on this data, improving its predictive ability for cost overrun risks related to process quality during the operation and maintenance phase.

[0058] (III) Application Results: The final total cost of the project was RMB 4.82 billion, saving RMB 180 million compared to the initial budget. The risk of overspending was reduced from 35% in the traditional model to 8%. The efficiency of bidding qualification review was improved by 60%, preventing three high-risk companies from winning the bid. The average early warning response time during the construction phase was shortened to 24 hours, and the efficiency of risk handling was improved by 50%. During the operation and maintenance phase, accountability and model iteration provided data support for the process quality control of similar projects in the future.

[0059] III. Use Value (a) Value to the project construction unit Reducing cost overrun losses: Through full-cycle data integration and precise early warning, cost risks can be "predicted in advance and controlled during the process," significantly reducing the probability of cost overruns and the amount of losses. As in the application case above, the construction unit saved 180 million yuan in budget, and the incidence of cost overruns decreased significantly.

[0060] Improve project management efficiency: Automated data collection, preprocessing, and qualification review processes reduce manual intervention and lower management costs. Bidding qualification review efficiency is improved by 60%, and the early warning response time during the construction phase is shortened to 24 hours, significantly enhancing management efficiency.

[0061] Optimize resource allocation: Model-based parameter optimization and differentiated management help construction units select high-quality bidders, adjust construction and procurement plans, and achieve a reasonable allocation of human, material, and financial resources.

[0062] (ii) Value to bidding and tendering supervision departments Enhance the precision of supervision: By linking the credit data of bidding entities with the full-cycle cost data, dynamic supervision of bidding entities can be achieved, high-risk entities can be identified and avoided in a timely manner, and the order of the bidding market can be standardized.

[0063] Enhance regulatory transparency: The system stores traceable data and control records throughout the entire lifecycle, providing regulatory authorities with a complete chain of evidence, facilitating compliance checks and accountability for violations, and improving the transparency and fairness of regulatory work.

[0064] (III) Value to the bidding company Creating a level playing field: Differentiated qualification review standards are formulated based on objective credit data and risk levels, avoiding subjective biases in traditional reviews and providing a fair competitive opportunity for high-quality enterprises.

[0065] Guiding enterprises to standardize contract performance: The system dynamically tracks contract performance and synchronizes credit records, forcing enterprises to improve their cost control capabilities and contract performance quality, and promoting the transformation of enterprises towards standardized and refined management.

[0066] (iv) Value to industry development Driving Digital Transformation: The distributed data acquisition network and intelligent algorithm model constructed in this solution provide a demonstration solution for full-cycle digital management and control in the field of bridge engineering, promoting the industry's transformation from traditional experience-based management to data-driven management.

[0067] Accumulate industry data assets: The system archives full-cycle project data and cases to form standardized industry data assets, providing a reference for subsequent bridge engineering cost calculation and bidding management, and helping to improve the overall technical level of the industry.

[0068] Reduce overall industry risk: The promotion and application of the solution can significantly reduce the risk of cost overruns in bridge engineering across the industry, improve the economy and stability of transportation infrastructure construction, and provide reliable transportation guarantees for social and economic development.

[0069] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for early warning of cost overruns and collaborative management of bidding processes throughout the entire lifecycle of bridge engineering projects, characterized in that, Includes the following steps: S1. Full-cycle, multi-dimensional data collection: Construct a distributed data collection network covering bridge engineering to collect design parameter data, dynamic price data of building materials market, credit data of bidding entities, construction progress data, operation and maintenance loss data, and policy and regulation adjustment data; S2. Data Preprocessing and Feature Engineering: The data collected in step S1 is processed to remove outliers, fill in missing values, and standardize the format. Based on the bridge engineering cost influencing factor system, core feature variables are extracted. S3. Construct a collaborative management and early warning model: Based on the LSTM-random forest fusion algorithm, input the core feature variables processed in step S2, and construct a cost overrun early warning sub-model and a bidding parameter optimization sub-model. The early warning sub-model determines the overrun threshold range through training with historical overrun cases, and the optimization sub-model establishes the mapping relationship between bidding parameters and the full-cycle cost. S4. Collaborative Management and Control Execution in the Bidding and Tendering Stage: The pre-overspending risk level of the early warning sub-model is used as the input parameter of the optimization sub-model. The maximum bid limit, performance bond ratio and payment node terms in the bidding documents are dynamically adjusted to generate differentiated bidding qualification review standards and strengthen qualification review for high-risk bidders. S5. Dynamic early warning and control adjustment during construction and operation and maintenance phases: Input the actual cost data of the construction phase and the loss data of the operation and maintenance phase into the early warning sub-model in real time. When the cost deviates from the over-cost threshold, a graded early warning signal is triggered.

2. The method according to claim 1, characterized in that, In step S1, the distributed data acquisition network adopts a collaborative architecture of edge computing nodes and cloud servers. The edge computing nodes are deployed at construction sites, building materials trading markets and bidding platforms to realize real-time data acquisition and preliminary filtering, while the cloud servers are responsible for centralized data storage and synchronous updates.

3. The method according to claim 1, characterized in that, In step S3, the improved LSTM-random forest fusion algorithm optimizes the feature extraction process of LSTM through an attention mechanism, enhances the ability to capture non-periodic factors such as sudden changes in building material prices and policy adjustments, and uses the random forest algorithm to correct the output of LSTM, thereby reducing the prediction bias of a single algorithm.

4. The method according to claim 1, characterized in that, In step S4, the differentiated bidding qualification review standard is formulated based on the matching results of the credit rating of the bidding entity and the risk level output by the early warning sub-model. The review process is simplified for bidding entities with a credit rating of A and a low risk level, while the technical solution defense and financial strength verification are added for bidding entities with a credit rating of C or below or a high risk level.

5. The method according to claim 1, characterized in that, In step S5, the graded early warning signals are divided into Level 1, Level 2 and Level 3 early warnings, and the corresponding control measures are as follows: Level 1 early warning initiates the submission mechanism for cost deviation analysis reports; Level 2 early warning adjusts the payment ratio of project funds; and Level 3 early warning initiates the deduction of performance bonds and supplementary bidding process.

6. A bridge engineering full-cycle cost overrun early warning and bidding collaborative management device, characterized in that, The components include: Data acquisition module: It adopts a distributed sensor array and interface adapter unit to collect design parameter data, dynamic price data of building materials market, credit data of bidding entities, construction progress data, operation and maintenance loss data and policy and regulation adjustment data throughout the entire life cycle of bridge engineering. Data preprocessing module: Communicates with the data acquisition module to perform outlier removal, missing value completion, and format standardization on the acquired data, and extracts core feature variables; Model building and computation module: It has a built-in improved LSTM-random forest fusion algorithm, which communicates with the data preprocessing module to build a cost overrun early warning sub-model and a bidding parameter optimization sub-model, and to perform early warning calculation and parameter optimization calculation. Bidding and tendering collaborative management module: It communicates with the model building and calculation module and is used to dynamically adjust the parameters of the bidding documents, generate differentiated bidding qualification review standards, and perform qualification review of bidding entities based on the early warning results; Early warning and adjustment module: It communicates with the model building and calculation module and the bidding and tendering collaborative management module to monitor the cost data of the whole cycle in real time, trigger the graded early warning signal, and link the execution of project payment adjustment, breach of contract accountability or supplementary bidding process; Storage and Interaction Module: Communicates with the above modules to store data, model parameters, and control records, and provides a human-machine interface for users to view early warning information and control instructions.

7. The device according to claim 6, characterized in that, The interface adapter unit of the data acquisition module supports protocol integration with public service platforms for bidding and tendering, e-commerce platforms for building materials, construction progress management systems, and operation and maintenance monitoring systems, and is compatible with multiple data formats such as XML, JSON, and CSV.

8. The device according to claim 6, characterized in that, The model building and computation module also includes a model update unit, which is used to periodically update the parameters of the early warning sub-model and the optimization sub-model based on the newly added full-cycle cost data and bidding case data, so as to improve the adaptability and prediction accuracy of the model.

9. The device according to claim 6, characterized in that, The early warning and adjustment module also includes an early warning push unit, which pushes early warning information and control instructions to project managers, bidding supervision departments and bidding entities through three methods: SMS, platform messages and email.

Images