Construction machinery product carbon footprint tracking method and device based on large model

Abstract

The application provides a kind of engineering machinery product carbon footprint tracking method and device based on large model, it is related to carbon accounting and product carbon footprint tracking technical field, comprising: receiving the carbon footprint tracking request of target engineering machinery product for user;Tracking request is input into engineering machinery field large model and is parsed, generates carbon footprint calculation logic chain, dispatches data layer and knowledge layer to obtain the corresponding carbon emission parameter data of target engineering machinery product, and is calculated based on carbon footprint calculation logic chain, obtains carbon footprint quantification result, generates and outputs the visual report containing carbon footprint details and optimization suggestions based on the result.The method and device provided by the application, by deeply fusing industry scene and large model technical advantages, real-time response to the carbon footprint tracking request of user, generate the visual report containing carbon footprint details and optimization suggestions, realize the accurate tracking, compliance management and low-carbon optimization of engineering machinery product whole life cycle carbon footprint.

Description

Technical Field

[0001] This invention relates to the field of carbon accounting and product carbon footprint tracking technology, and in particular to a method and apparatus for tracking the carbon footprint of engineering machinery products based on a large model. Background Technology

[0002] The construction machinery industry, encompassing earthmoving machinery, aviation equipment, and special equipment, requires carbon footprint tracking throughout its entire lifecycle, including R&D, raw materials, production, use, and recycling. However, the industry's long chain, high dynamism, and strong heterogeneity exacerbate the tracking challenges. Existing traditional technologies face significant bottlenecks: multi-source data, such as structured energy consumption, unstructured policy data, and time-series Internet of Things (IoT) data, requires manual integration, resulting in low efficiency and large errors; there is no unified carbon tracking standard for construction machinery equipment, with companies recording data based on their own production and operation activities, leading to unclear accounting boundaries, activity data, and emission factors; static rule-based statistics cannot adapt to new policies / materials in real time, resulting in long update cycles; they can only statistically analyze carbon emissions, failing to pinpoint high-carbon emission links, and lack the ability to personalize and optimize carbon reduction plans for individual companies, making it difficult for carbon footprint tracking to meet compliance requirements and the demands of low-carbon competition. While some methods using large-scale models for carbon accounting exist, they suffer from insufficient adaptation to industry characteristics and weak end-chain collaboration capabilities.

[0003] Therefore, how to accurately track, manage in compliance with regulations, and optimize for low-carbon emissions throughout the entire lifecycle of construction machinery products has become a pressing technical problem for the industry. Summary of the Invention

[0004] This invention provides a method and device for carbon footprint tracking of construction machinery products based on a large model, which addresses the shortcomings of existing technologies such as insufficient adaptation to industry characteristics of carbon accounting and weak full-chain collaboration capabilities, and realizes accurate tracking, compliance management and low-carbon optimization of the carbon footprint of construction machinery products throughout their entire life cycle.

[0005] This invention provides a method for carbon footprint tracking of engineering machinery products based on a large model, comprising: Receive user requests for carbon footprint tracking of target construction machinery products; The carbon footprint tracking request is input into a large model in the field of construction machinery for parsing, generating a carbon footprint calculation logic chain for the entire life cycle of the target construction machinery product; Based on the carbon footprint calculation logic chain, the scheduling data layer and knowledge layer acquire the corresponding carbon emission parameter data of the target construction machinery product; the data layer stores the energy and carbon dataset of the construction machinery product; the knowledge layer stores a knowledge graph specific to the construction machinery industry. The carbon emission parameter data is input into the mathematical calculation engine, and calculations are performed based on the carbon footprint calculation logic chain to obtain the carbon footprint quantification result. Based on the carbon footprint quantification results, a visual report containing carbon footprint details and optimization suggestions is generated and output.

[0006] In some embodiments, the large model in the field of engineering machinery is obtained through training via the following steps: Construct a training dataset; the training dataset includes at least one of domain question-answer pairs, computation rule instruction sets, and structured data text pairs; A low-rank adaptive fine-tuning method is used to fine-tune the pre-trained large language model based on the training dataset to obtain the fine-tuned adapter weights. The adapter weights are merged with the original weights of the pre-trained large language model to generate the large model for the engineering machinery domain.

[0007] In some embodiments, the energy carbon dataset is obtained based on the following steps: Acquire multi-source heterogeneous carbon data and preprocess the multi-source heterogeneous carbon data; The preprocessed multi-source heterogeneous energy carbon data is standardized to obtain standardized energy carbon data. The standardized energy and carbon data are semantically aligned using a cross-modal similarity method to construct the energy and carbon dataset.

[0008] In some embodiments, the construction machinery industry-specific knowledge graph is obtained based on the following steps: Representative entities are extracted from standard documents and product manuals; the representative entities include at least one of standard regulatory entities, product entities, process entities, and emission factor entities. Based on the semantic relationships between the representative entities, a semantic relationship network is established to obtain the knowledge graph specific to the engineering machinery industry; the semantic relationships include at least one of following, containing, consuming, corresponding, and supplying.

[0009] In some embodiments, the method further includes: Attribute information data is attached to the representative entity; the attribute information data includes at least one of numerical data, unit data, source data, and effective status data. Based on the newly released policy and regulatory texts, update the attribute information data and semantic relationships of the corresponding entities in the construction machinery industry-specific knowledge graph.

[0010] In some embodiments, generating and outputting a visualization report containing carbon footprint details and optimization suggestions based on the carbon footprint quantification results includes: Generate a heat map of the carbon footprint of the target engineering machinery product throughout its entire life cycle; In response to a user's drilling operation on any stage of the full lifecycle carbon footprint heatmap, fine-grained data for that stage is displayed.

[0011] This invention provides a carbon footprint tracking device for engineering machinery products based on a large model, comprising: The response module is used to receive carbon footprint tracking requests from users for target construction machinery products; The parsing module is used to input the carbon footprint tracking request into a large model in the field of construction machinery for parsing, and generate a carbon footprint calculation logic chain for the entire life cycle of the target construction machinery product; The retrieval module is used to schedule the data layer and knowledge layer to obtain the corresponding carbon emission parameter data of the target construction machinery product based on the carbon footprint calculation logic chain; the data layer stores the energy and carbon dataset of the construction machinery product; the knowledge layer stores the construction machinery industry-specific knowledge graph. The calculation module is used to input the carbon emission parameter data into the mathematical calculation engine, perform calculations based on the carbon footprint calculation logic chain, and obtain the carbon footprint quantification result. The presentation module is used to generate and output a visual report containing carbon footprint details and optimization suggestions based on the carbon footprint quantification results.

[0012] The present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the carbon footprint tracking method for engineering machinery products based on a large model.

[0013] The present invention provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the carbon footprint tracking method for engineering machinery products based on a large model.

[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the carbon footprint tracking method for engineering machinery products based on a large model.

[0015] This invention provides a method and apparatus for carbon footprint tracking of construction machinery products based on a large model. By constructing an energy carbon dataset and a knowledge graph specific to the construction machinery industry, it achieves automated cleaning and semantic alignment of multi-source heterogeneous data throughout the entire lifecycle, solving the problems of fragmented industry data and lagging standard updates, thus making carbon accounting based on evidence. Through a hybrid architecture of "domain large model + mathematical calculation engine," it responds to users' carbon footprint tracking requests in real time. It uses the large model to generate accurate logical chains and schedules a professional engine for numerical calculations, which not only overcomes the illusions and errors that may be generated by simple large model calculations, but also realizes the transformation from passive "tracking" to proactive "optimization" by generating a visual report containing carbon footprint details and optimization suggestions. By deeply integrating industry scenarios and the advantages of large model technology, it achieves accurate tracking, compliance management, and low-carbon optimization of the carbon footprint of construction machinery products throughout their entire lifecycle, filling the gap in industry-specific large model carbon tracking technology. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the carbon footprint tracking method for engineering machinery products based on a large model provided by the present invention.

[0019] Figure 2 This is the overall architecture diagram of the carbon footprint tracking method for engineering machinery products based on a large model provided by the present invention.

[0020] Figure 3 This is an application operation flowchart of the carbon footprint tracking method for engineering machinery products based on a large model provided by the present invention.

[0021] Figure 4 This is a diagram of the overall architecture of the model layer of the carbon footprint tracking method for engineering machinery products based on a large model provided by the present invention.

[0022] Figure 5 This is a knowledge layer workflow diagram of the carbon footprint tracking method for engineering machinery products based on a large model provided by the present invention.

[0023] Figure 6 This is a schematic diagram of the carbon footprint tracking device for engineering machinery products based on a large model provided by the present invention.

[0024] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first," "second," etc., used in this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps, units, or modules is not necessarily limited to those explicitly listed, but may include other steps, units, or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0027] Figure 1 This is a flowchart illustrating the carbon footprint tracking method for engineering machinery products based on a large model provided by the present invention, as shown below. Figure 1 As shown, the method includes steps 110, 120, 130, 140 and 150.

[0028] Step 110: Receive the user's carbon footprint tracking request for the target construction machinery product.

[0029] Specifically, the execution entity of the carbon footprint tracking method for construction machinery products based on a large model provided in this embodiment of the invention is a carbon footprint tracking device for construction machinery products based on a large model. This device can be implemented in software, such as a carbon footprint tracking program for construction machinery products based on a large model running on a computer; or it can be implemented in hardware, such as a computer or server that executes the carbon footprint tracking method for construction machinery products based on a large model.

[0030] The carbon footprint tracking method for construction machinery products based on a large model provided in this invention is applicable to construction machinery manufacturers or discrete manufacturing industries. Its core objective is to achieve precise carbon footprint tracking and management throughout the entire lifecycle of construction machinery products such as earthmoving machinery, aviation equipment, and special equipment (including stages such as raw material acquisition and pretreatment, production, product distribution and storage, use, and end-of-life). The method focuses on overcoming key technical challenges in construction machinery carbon footprint tracking, including the difficulty of automatic fusion and standardization of multimodal heterogeneous data throughout the entire lifecycle; unreliable calculations due to the lack of product-specific carbon footprint tracking standards; weak real-time performance and large errors in carbon emission calculations under highly dynamic scenarios; cross-entity data collaboration and reliable interaction across the entire lifecycle; and the lack of carbon footprint "tracking-optimization" feedback and insufficient personalized adaptation.

[0031] Considering that the same equipment from different construction machinery manufacturers has common characteristics in the entire life cycle process and core carbon emission links, the method of this invention can form a common technical solution with strong generalization, which can be flexibly adapted to different manufacturers, different models of construction machinery and diverse operating scenarios. Ultimately, it provides the industry with integrated support of "precise tracking - compliance management - low carbon optimization", helping the construction machinery field to achieve carbon footprint transparency and energy conservation and emission reduction goals.

[0032] Figure 2 This is an overall architecture diagram of the carbon footprint tracking method for engineering machinery products based on a large model provided by the present invention, as shown in the diagram. Figure 2 As shown, the carbon footprint tracking method for engineering machinery products based on a large model provided in this embodiment of the invention adopts a four-layer architecture of "data layer - knowledge layer - model layer - application layer", forming a technical link of "automatic data integration - dynamic standard construction - accurate carbon emission calculation - optimization scheme feedback", which is suitable for the needs of multiple models and all stages of the entire life cycle of earthmoving machinery, aviation equipment and special equipment.

[0033] The four-layer architecture of "data layer - knowledge layer - model layer - application layer" and its collaborative working mechanism proposed in this invention constitute a complete technical framework for realizing the technical chain of "automatic data integration - dynamic standard construction - accurate carbon emission calculation - optimization closed-loop implementation". It includes: a data layer module for realizing automatic integration of multi-source data; a knowledge layer module for storing and dynamically updating industry carbon accounting knowledge; a model layer module containing a domain-wide large model and mathematical calculation engine finely tuned by low-rank adaptation (LoRA); and an application layer module for providing interactive carbon footprint reports and optimization suggestions.

[0034] The data layer addresses the heterogeneous nature of data throughout the entire lifecycle of construction machinery products by constructing a modular data processing chain. The knowledge layer builds a knowledge graph specific to the construction machinery industry, ensuring that carbon footprint tracking is "rule-based" and providing a complete carbon footprint traceability chain. The model layer constructs a domain-specific fine-tuned large model to deeply reason and integrate the facts from the data layer and the rules from the knowledge layer, combining this with a mathematical calculation engine to output accurate carbon footprint quantification results. The application layer is user-facing, providing an interactive application that offers a clear operational process and accurate, reliable equipment carbon footprint tracking report.

[0035] Figure 3 This is a flowchart illustrating the application operation of the carbon footprint tracking method for engineering machinery products based on a large model provided by this invention. Figure 3 As shown, the application layer is user-facing, responding to users' carbon footprint tracking requests through interactive applications.

[0036] In this embodiment of the invention, a carbon footprint tracking request from a user for a target construction machinery product is received. This carbon footprint tracking request is the starting signal that triggers the entire system to perform data scheduling, logical parsing, and numerical calculations.

[0037] Taking a certain excavator as an example, a feasible implementation method is as follows: Step 1: After logging into the system, users can obtain the basic carbon footprint of a specific device by entering its serial number (SN). Users can then ask questions such as, "Please generate a carbon footprint heatmap for my SWE60E mini excavator."

[0038] Step 120: Input the carbon footprint tracking request into the large model of the construction machinery field for parsing, and generate a carbon footprint calculation logic chain for the entire life cycle of the target construction machinery product.

[0039] Specifically, in the field of engineering machinery, the large model is the core component of the model layer, responsible for understanding user intent, parsing unstructured data, and generating accurate carbon footprint calculation logic chains.

[0040] Traditional carbon accounting software typically uses hard-coded pre-defined calculation rules, which are difficult to adapt to the needs of different equipment, operating conditions, or standard changes. This application, however, utilizes a large-scale engineering machinery domain model that has been fine-tuned to dynamically generate the calculation logic.

[0041] In this embodiment of the invention, the user's carbon footprint tracking request is input into a large model in the field of construction machinery for parsing, and a carbon footprint calculation logic chain for the entire life cycle of the target construction machinery product is generated.

[0042] Taking the aforementioned excavator as an example, after receiving input, the large-scale model in the construction machinery field first performs semantic understanding and intent recognition. Based on its training knowledge in the construction machinery field, the model can identify that "SWE60E" is a specific excavator model, and that "full life cycle carbon footprint" means covering all stages, including raw material acquisition, production and manufacturing, distribution and storage, use, and end-of-life recycling. Furthermore, the large-scale model in the construction machinery field does not directly provide a potentially incorrect numerical result, but rather generates a step-by-step, structured computational logic chain.

[0043] Specifically, the Large Language Model (LLM) at the model layer receives instructions, performs intent recognition and task parsing, and generates a thought process, such as "The user requests to generate a carbon footprint heatmap for the SWE60E mini excavator. The full lifecycle carbon footprint of the SWE60E mini excavator must first be calculated. The calculation logic chain for the SWE60E mini excavator is as follows:" CFP stands for product lifecycle carbon footprint. It refers to greenhouse gas emissions during the raw material acquisition and pretreatment stage; It refers to greenhouse gas emissions during the production phase; It refers to greenhouse gas emissions during the distribution and storage phase; It refers to greenhouse gas emissions during the usage phase; This refers to greenhouse gas emissions during the recycling and processing stages; This represents greenhouse gas activity data for the j-th activity; This represents the greenhouse gas emission factor corresponding to the j-th activity; This represents the global warming potential value corresponding to the j-th activity.

[0044] Step 130: Based on the carbon footprint calculation logic chain, schedule the data layer and knowledge layer to obtain the corresponding carbon emission parameter data of the target construction machinery product; the data layer stores the energy and carbon dataset of the construction machinery product; the knowledge layer stores the construction machinery industry-specific knowledge graph.

[0045] Specifically, the data layer stores a carbon dataset of engineering machinery products after multimodal preprocessing and semantic alignment. The knowledge layer stores a knowledge graph specific to the engineering machinery industry.

[0046] In this embodiment of the invention, based on the carbon footprint calculation logic chain, the data layer and knowledge layer can be scheduled to obtain the corresponding carbon emission parameter data of the target engineering machinery product.

[0047] Taking the aforementioned excavator as an example, after obtaining the carbon footprint calculation logic chain through the large model in the field of construction machinery, a precise call request is initiated to the data layer and knowledge layer based on the calculation logic chain, such as "calling the material and weight of all parts of SWE60E, calling the working hours and energy consumption records of SWE60E during the manufacturing process, calling the operating data such as working time and fuel consumption of SWE60E, and obtaining emission factors from the knowledge graph, etc." Step 140: Input the carbon emission parameter data into the mathematical calculation engine, perform calculations based on the carbon footprint calculation logic chain, and obtain the carbon footprint quantification result.

[0048] Specifically, unlike existing technologies that directly rely on large language models to generate final numerical results, this invention designs a hybrid architecture of "domain-wide large model + mathematical calculation engine". The fine-tuning of the large model is responsible for handling unstructured semantic understanding and logical reasoning, while the calculation engine ensures the accurate solution of structured mathematical formulas. This solves the problems of poor adaptability and insufficient industry-specificity of traditional general carbon management systems. The final output carbon footprint results are accurate, transparent, and traceable, greatly enhancing the credibility and compliance of the report.

[0049] Large models are essentially generative models based on probabilistic prediction. When dealing with complex floating-point operations, multi-level accumulations, and unit conversions, they are prone to precision loss. To address this issue, this invention introduces a mathematical computation engine to handle the specific numerical calculation tasks.

[0050] Carbon emission parameter data is input into the mathematical calculation engine, and calculations are performed based on the carbon footprint calculation logic chain to obtain the carbon footprint quantification result.

[0051] Step 150: Based on the carbon footprint quantification results, generate and output a visualization report containing carbon footprint details and optimization suggestions.

[0052] Specifically, after the mathematical calculation engine completes the numerical calculation, the system not only outputs a single total carbon emission value, but also assembles the calculation results into a structured and visualized comprehensive report.

[0053] Furthermore, unlike traditional software that only displays static data, this embodiment of the invention utilizes the reasoning capabilities of a large domain model to diagnose the calculation results and generate concrete optimization suggestions.

[0054] Taking the aforementioned excavator as an example, after the mathematical calculation engine completes the numerical calculation, it synthesizes the calculation results and generates a structured detailed report. Table 1 shows an example of core content, namely the SWE60E full life cycle carbon footprint calculation report, as shown in Table 1.

[0055] Table 1. SWE60E Life Cycle Carbon Footprint Calculation Report

[0056] The carbon footprint tracking method for construction machinery products based on a large model provided in this invention achieves automated cleaning and semantic alignment of multi-source heterogeneous data throughout the entire lifecycle by constructing an energy carbon dataset and a knowledge graph specific to the construction machinery industry. This solves the problems of fragmented industry data and lagging standard updates, making carbon accounting based on evidence. Through a hybrid architecture of "domain large model + mathematical calculation engine", it responds to users' carbon footprint tracking requests in real time. It uses the large model to generate accurate logical chains and schedules a professional engine for numerical calculations. This not only overcomes the illusions and errors that may be generated by simple large model calculations, but also realizes the transformation from passive "tracking" to proactive "optimization" by generating a visual report containing carbon footprint details and optimization suggestions. By deeply integrating industry scenarios and the advantages of large model technology, it achieves accurate tracking, compliance management, and low-carbon optimization of the carbon footprint of construction machinery products throughout their entire lifecycle, filling the gap in industry-specific large model carbon tracking technology.

[0057] In some embodiments, the large model in the field of engineering machinery is obtained through training via the following steps: Construct a training dataset; the training dataset includes at least one of domain question-answer pairs, computation rule instruction sets, and structured data text pairs; A low-rank adaptive fine-tuning method is used to fine-tune the pre-trained large language model based on the training dataset to obtain the fine-tuned adapter weights. The adapter weights are merged with the original weights of the pre-trained large language model to generate the large model for the engineering machinery domain.

[0058] Specifically, in this embodiment of the invention, in order to ensure the professionalism and inference accuracy of the large model in the field of carbon footprint of engineering machinery, this embodiment of the invention constructs a fine-tuning training set and uses a parameter-efficient fine-tuning method for domain-adaptive fine-tuning.

[0059] Figure 4 This is a diagram of the overall architecture of the model layer of the carbon footprint tracking method for engineering machinery products based on a large model provided by this invention, as shown below. Figure 4 As shown in the figure, this illustrates the overall architecture of the model layer. The efficient fine-tuning method for large domain models based on low-rank adaptation specifically includes a method for constructing a training dataset that integrates domain question-and-answer pairs, computational rule instruction sets, and structured data text pairs for fine-tuning; and a specific fine-tuning process, parameter configuration, and merged deployment scheme using LoRA technology, through low-rank matrix injection and incremental updates, to efficiently adapt a general large model to a domain expert in engineering machinery carbon footprint tracking. The specific operations are as follows: Step 1: Select a base model. Choose a large language model with strong logical reasoning capabilities as the base model, i.e., a pre-trained large language model. Freeze all parameters of the original LLM. During fine-tuning, the original parameters... It remains unchanged. Here, d is the input feature dimension, and k is the output feature dimension.

[0060] Step 2: Construct the training dataset. The training set consists of three parts: domain question-answer pairs, computation rule instruction sets, and structured data text pairs.

[0061] Domain-specific question answering is used to construct engineering machinery expertise based on carbon accounting standards, for example: Q: "How to calculate the carbon emissions during the use of an excavator?" answer:" in, This refers to carbon emissions during the usage phase. This refers to diesel fuel consumption. It is a diesel carbon emission factor, and for diesel-powered excavators, the main basis is diesel consumption.

[0062] The computation rule instruction set is used to train the model to identify problems and generate computational logic, for example: Instruction: "Please generate calculation steps for 'Calculate the carbon emissions of an excavator during the manufacturing phase'"; Output: 1. Obtain the weight and material of all components from the Bill of Materials (BOM). 2. Query the upstream production emission factors for each material. 3. Obtain the energy consumption of each manufacturing process from the Manufacturing Execution System (MES). 4. Query the emission factors for electricity and natural gas. 5. Calculate: in, This refers to carbon emissions during the manufacturing phase; It represents the carbon emissions generated by material consumption during the manufacturing process; n is the total amount of materials. Let be the weight of the i-th material; It is the carbon emission factor corresponding to the material; It represents the carbon emissions generated by energy consumption during the manufacturing process; k represents the total energy consumption. This is the consumption of the j-th type of energy; It is the carbon emission factor corresponding to energy.

[0063] Structured text data is used to train models to understand the semantics behind structured data. For example: JavaScript Object Notation (JSON) data: {"Device ID":"SWE60E","Working Time":"8.5","Fuel Consumption":"156"} Text description: "The SWE60E excavator worked for 8.5 hours today, consuming 156L of diesel." Step 3: Inject the rank decomposition matrix into some linear layers of the base model using a low-rank adaptive fine-tuning method, and perform incremental updates. The specific formula is as follows: Where A and B are low-rank matrices, satisfying , , r is the rank that satisfies In this embodiment of the invention, A is initialized as a zero matrix, and B is initialized as a random Gaussian distribution to ensure that at the start of LLM training, It is zero.

[0064] Step 4: Fine-tuning and Optimization of Training Configuration. The LoRA parameter fine-tuning method is employed to fine-tune the pre-trained large language model using this training dataset. The LoRA injection parameters are updated via gradient descent. The number of trainable parameters has increased from the original... Significantly reduced to The loss function minimized in this invention is as follows: in, It is the cross-entropy loss function. This is the forward computation process of LLM. , These are the input and output parameters for fine-tuning the training set, where N is the number of data sets. Finally, key hyperparameters such as rank, scaling factor, learning rate, and training period are set to update the parameters during training.

[0065] Step 5: Fine-tuning, saving, and applying weights. Only save the LoRA adapter weight files obtained during training. During inference deployment, the original weights of the base model are... With training The models will be merged to form a single, large-scale domain model specifically for tracking the carbon footprint of construction machinery. The merging method is as follows: in, These are the final model weights.

[0066] The carbon footprint tracking method for construction machinery products based on a large model provided in this invention constructs a training dataset and employs LoRA technology. Through low-rank matrix injection and incremental updates, it efficiently adapts a general large model to a large model for the field of carbon footprint tracking in construction machinery. This not only significantly reduces the computational cost and deployment threshold of model training, but also enables the general model to accurately grasp the complex BOM structure, operating terminology, and accounting standards of the construction machinery industry. As a result, it can accurately interpret user intent and generate a carbon footprint calculation logic chain that conforms to industry standards. This effectively solves the core technical problems of general large models, such as a lack of professional knowledge in vertical fields, the susceptibility of reasoning logic to illusions, and the inability to adapt to dynamic standards. It significantly improves the intelligence level and reliability of the carbon accounting process.

[0067] In some embodiments, the energy carbon dataset is obtained based on the following steps: Acquire multi-source heterogeneous carbon data and preprocess the multi-source heterogeneous carbon data; The preprocessed multi-source heterogeneous energy carbon data is standardized to obtain standardized energy carbon data. The standardized energy and carbon data are semantically aligned using a cross-modal similarity method to construct the energy and carbon dataset.

[0068] Specifically, to support accurate calculations in subsequent model layers, a high-quality energy and carbon dataset needs to be constructed. The data layer collects energy and carbon data from different products through methods such as cameras, sensors, and tables, and fuses the heterogeneous data to obtain the energy and carbon dataset.

[0069] In this embodiment of the invention, the data layer, taking into account the heterogeneous characteristics of the data throughout the entire life cycle of engineering machinery products, adopts a modular approach to construct a full-process data processing link, including a multi-source data access module, a multi-modal preprocessing module, a data fusion and standardization module, etc., to realize the automatic collection, cleaning and unification of energy and carbon data, providing highly reliable data support for carbon footprint tracking.

[0070] First, multi-source heterogeneous energy carbon data is acquired. The multi-source data access module uses technologies such as protocol compatibility and format conversion to achieve full coverage access of energy carbon data.

[0071] Taking excavator products as an example, for time-series IoT data, it connects to equipment sensors such as load cells and fuel flow meters through recommended standard 485 + Modbus protocol (RS485+Modbus), Controller Area Network (CAN), and Message Queuing Telemetry Transport (MQTT) protocols, achieving a data collection frequency of up to 1 time / second to acquire real-time operating condition data. For structured production data, it connects to Manufacturing Execution System (MES) and Enterprise Resource Planning (ERP) systems through interfaces such as Open Platform Communications (OPC) and Java Database Connectivity (JDBC) to automatically extract data such as equipment energy consumption and raw material usage. For unstructured policy / technical data, it uses Optical Character Recognition (OCR) + Portable Document Format (PDO) technology. The system utilizes OCR+PDF parsing technology to read text data such as the "Greenhouse Gases - Product Carbon Footprint - Quantification Requirements and Guidelines" series of standards and supplier technical manuals. Through an innovative multi-source access combination of "MQTT + JDBC + OCR + Application Programming Interface (API) + mobile forms," ​​it achieves standardized integration of heterogeneous data such as IoT data, production system data, and policy texts, thus solving the problem of fragmented data in the construction machinery industry.

[0072] Then, the acquired multi-source heterogeneous carbon data are preprocessed. Multimodal preprocessing utilizes techniques such as machine learning to clean the data and extract features, eliminating noise and redundancy. For example, using 3... The principle is to use the isolated forest algorithm to filter out energy consumption anomalies in the equipment, and to use the sliding window method to downsample the data, retaining the relevant characteristics of energy and carbon data and reducing data pressure.

[0073] Finally, the preprocessed multi-source heterogeneous energy carbon data is standardized to obtain standardized energy carbon data. Semantic alignment of this standardized energy carbon data is then performed based on the cross-modal similarity method to construct an energy carbon dataset. The data fusion and standardization module employs techniques such as format normalization and association mapping to construct the energy carbon dataset, including methods for unifying data units and time formats, and semantic alignment based on cosine similarity and weighted fusion. Semantic alignment is achieved through the cross-modal similarity method, calculating similarity and performing weighted fusion, as shown in the following formula: in, , There are two eigenvectors; This indicates the calculation of cosine similarity. Similarity; k is the eigenvector and Different similarity measures between them; F is the weighted average fusion.

[0074] The carbon footprint tracking method for engineering machinery products based on a large model provided in this invention achieves automatic collection and fusion of multi-source heterogeneous data through a "data layer," significantly improving efficiency compared to traditional manual collection. It employs a 3D model. The outlier filtering scheme combining criteria and the isolated forest algorithm significantly reduces the data anomaly rate, overcoming the bottlenecks of low efficiency and error-proneness in traditional manual data collection and processing; it also constructs a high-quality energy and carbon dataset, providing data support for accurate calculations in subsequent model layers.

[0075] In some embodiments, the construction machinery industry-specific knowledge graph is obtained based on the following steps: Representative entities are extracted from standard documents and product manuals; the representative entities include at least one of standard regulatory entities, product entities, process entities, and emission factor entities. Based on the semantic relationships between the representative entities, a semantic relationship network is established to obtain the knowledge graph specific to the engineering machinery industry; the semantic relationships include at least one of following, containing, consuming, corresponding, and supplying.

[0076] Specifically, the knowledge layer module is responsible for constructing a knowledge graph specific to the construction machinery industry. This knowledge graph is not a simple database, but rather a semantic network with logical reasoning capabilities that structures discrete carbon accounting elements in a machine-readable format.

[0077] In this embodiment of the invention, the knowledge layer constructs a knowledge graph specific to the construction machinery industry. It stores data such as standards and regulations, product components, life cycle processes and corresponding emission factors related to carbon footprint accounting in a structured form that is machine readable, and establishes semantic relationships between them to achieve "rules to follow" for carbon footprint tracking, providing a complete carbon footprint traceability chain and enhancing the transparency and credibility of the carbon footprint tracking process. Figure 5 This is a knowledge layer workflow diagram of the carbon footprint tracking method for engineering machinery products based on a large model provided by the present invention, such as... Figure 5 As shown in the diagram, this diagram illustrates the knowledge layer structure and workflow.

[0078] Taking hydraulic excavators as an example, a dedicated knowledge graph is constructed, and the specific implementation is as follows: Step 1: Entity Definition and Extraction. Representative entities are extracted from relevant standard documents and product manuals, including standard and regulatory entities, product entities, process entities, emission factor entities, etc.

[0079] Among them, standard regulatory entities represent the various norms that must be followed in carbon footprint accounting.

[0080] Product entities represent the physical objects being tracked and their components, including complete machines and parts. Complete machines include SWE60E mini excavators, while parts include six-cylinder diesel engines, hydraulic main pumps, track assemblies, platforms, and so on.

[0081] Process entities represent all activities that occur throughout a product's entire lifecycle, including the raw material acquisition and pretreatment stage, the production stage, the product distribution and storage stage, the usage stage, and the end-of-life stage. Specific activities include steel rolling, welding, painting, transportation, etc.

[0082] The emission factor entity is the core calculation factor that converts activity data into carbon dioxide equivalents, including diesel CO2 emission factors, grid average emission factors, high-strength steel smelting factors, and so on.

[0083] Step 2: Relationship Definition and Association. A relation is a directed edge connecting entities, defining the semantic association between them and serving as the path for logical reasoning in a knowledge graph. This invention establishes a semantic relationship network for entities in a graph database, including following, containing, consuming, corresponding, supplying, etc. Table 2 shows examples of semantic relationships in a knowledge graph, as shown in Table 2: Table 2. Examples of semantic relations in knowledge graphs

[0084] In this embodiment of the invention, a semantic relationship network is established based on the semantic relationships between representative entities to obtain a knowledge graph specific to the engineering machinery industry.

[0085] The carbon footprint tracking method for construction machinery products based on a large model provided in this invention constructs a knowledge graph specific to the construction machinery industry. This transforms discrete industry standards and regulations, complex product lifecycle processes, and dynamically changing emission factors into a machine-readable structured semantic network, providing precise logical constraints and real-time parameter support for the inference and calculation of the large model. This mechanism effectively overcomes the shortcomings of general large models, such as the tendency to produce illusions and unclear accounting boundaries in professional fields. It achieves "rule-based" and transparent and traceable carbon accounting processes, thereby significantly improving the compliance and calculation accuracy of carbon footprint tracking results.

[0086] In some embodiments, the method further includes: Attribute information data is attached to the representative entity; the attribute information data includes at least one of numerical data, unit data, source data, and effective status data. Based on the newly released policy and regulatory texts, update the attribute information data and semantic relationships of the corresponding entities in the construction machinery industry-specific knowledge graph.

[0087] Specifically, the knowledge layer module is responsible for constructing a knowledge graph specific to the engineering machinery industry. After constructing the skeleton of entities and relationships, in order to support subsequent high-precision numerical calculations and logical reasoning, the method in this embodiment of the invention further includes the step of attaching attribute information data to the representative entities. Attributes are metadata stored on the graph nodes in the form of "key-value pairs," realizing a refined expression of knowledge.

[0088] Taking the aforementioned hydraulic excavator as an example, after constructing the dedicated knowledge graph and building the skeleton of entities and relationships, the following steps are also included: Attribute definition and classification. Attributes are specific feature information attached to an entity in the form of key-value pairs. They enable refined expression of knowledge, supporting subsequent carbon footprint calculation and reasoning decisions.

[0089] The physical attributes of standards and regulations include publication date, effective status, scope of application, specific requirements, and revision history.

[0090] Product attributes include technical parameters, production batch, weight, and material composition.

[0091] The process entity attribute report includes energy consumption parameters, process parameters, duration, distance, etc.

[0092] The entity attributes of emission factors include numerical value, unit, update time, source, usage scenario, confidence level, etc.

[0093] Furthermore, addressing the pain point of frequent updates to carbon accounting standards and rapid policy changes in the construction machinery industry—namely, the inability of static rule statistics to adapt in real time—this method also includes a dynamic graph update mechanism based on newly released policy and regulatory texts. Utilizing the natural language understanding capabilities of a large model, it parses newly released policy and regulatory texts and automatically updates entities, attributes, and relationships in the graph, thereby achieving dynamic updates of the knowledge graph along with standards and policies.

[0094] The carbon footprint tracking method for construction machinery products based on a large model provided in this invention greatly improves the accuracy of carbon emission calculation and significantly shortens the policy adaptation cycle by constructing an industry knowledge graph, combining a dynamic factor update mechanism and a carbon accounting knowledge system specific to the construction machinery industry, thus solving the problem of standard lag.

[0095] In some embodiments, generating and outputting a visualization report containing carbon footprint details and optimization suggestions based on the carbon footprint quantification results includes: Generate a heat map of the carbon footprint of the target engineering machinery product throughout its entire life cycle; In response to a user's drilling operation on any stage of the full lifecycle carbon footprint heatmap, fine-grained data for that stage is displayed.

[0096] Specifically, in this embodiment of the invention, a user can perform a full lifecycle carbon footprint calculation by inputting a device identifier and obtain an interactive report containing detailed data, a visual heatmap, and specific optimization suggestions.

[0097] Once the system has completed the calculation of the carbon footprint, the application layer will transform the complex data results into an intuitive visualization report for users to use for decision analysis.

[0098] The system first aggregates the total carbon emissions of the target construction machinery product, such as the SWE60E excavator, across all stages, including raw material acquisition, manufacturing, distribution and storage, product use, and end-of-life recycling. Based on this data, the system uses a data visualization engine to generate a full lifecycle carbon footprint heatmap.

[0099] The system loads a full lifecycle carbon footprint heatmap, allowing users to click on any stage in the graph for drill-down analysis and view more granular data for that stage. The model provides suggestions such as, "The carbon emissions from this equipment's usage phase are higher than average, the equipment's idling time is high, and it is recommended to strengthen energy-saving training for operators," etc.

[0100] The carbon footprint tracking method for engineering machinery products based on a large model provided in this invention provides intuitive visualization reports and drill-down analysis functions through solution feedback from "tracking" to "optimization". Based on intelligent diagnosis at the model level, it automatically locates carbon emission hotspots and generates concrete optimization suggestions to promote the low-carbon transformation of the industry.

[0101] Figure 6 This is a schematic diagram of the carbon footprint tracking device for engineering machinery products based on a large model provided by the present invention, as shown below. Figure 6 As shown, the device includes a response module 610, a parsing module 620, a retrieval module 630, a calculation module 640, and a display module 650 connected in sequence.

[0102] Response module 610 is used to receive carbon footprint tracking requests from users for target construction machinery products; The parsing module 620 is used to input the carbon footprint tracking request into a large model in the field of construction machinery for parsing, and generate a carbon footprint calculation logic chain for the entire life cycle of the target construction machinery product; The retrieval module 630 is used to schedule the data layer and knowledge layer to obtain the corresponding carbon emission parameter data of the target construction machinery product based on the carbon footprint calculation logic chain; the data layer stores the energy and carbon dataset of the construction machinery product; the knowledge layer stores the construction machinery industry-specific knowledge graph. The calculation module 640 is used to input the carbon emission parameter data into the mathematical calculation engine, perform calculations based on the carbon footprint calculation logic chain, and obtain the carbon footprint quantification result. The display module 650 is used to generate and output a visual report containing carbon footprint details and optimization suggestions based on the carbon footprint quantification results.

[0103] The carbon footprint tracking device for construction machinery products based on a large model provided in this invention achieves automated cleaning and semantic alignment of multi-source heterogeneous data throughout the entire lifecycle by constructing an energy carbon dataset and a knowledge graph specific to the construction machinery industry. This solves the problems of fragmented industry data and lagging standard updates, making carbon accounting based on evidence. Through a hybrid architecture of "domain large model + mathematical calculation engine", it responds to users' carbon footprint tracking requests in real time. It uses the large model to generate accurate logical chains and schedules a professional engine for numerical calculations. This not only overcomes the illusions and errors that may be generated by simple large model calculations, but also realizes the transformation from passive "tracking" to proactive "optimization" by generating a visual report containing carbon footprint details and optimization suggestions. By deeply integrating industry scenarios and the advantages of large model technology, it achieves accurate tracking, compliance management, and low-carbon optimization of the carbon footprint of construction machinery products throughout their entire lifecycle, filling the gap in industry-specific large model carbon tracking technology.

[0104] Figure 7This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 7 As shown, the electronic device may include a processor 710, a communications interface 720, a memory 730, and a communications bus 740, wherein the processor 710, communications interface 720, and memory 730 communicate with each other via the communications bus 740. The processor 710 can call logical commands stored in the memory 730 to execute the methods described in the above embodiments, for example: The system receives a carbon footprint tracking request from a user for a target construction machinery product; it inputs the carbon footprint tracking request into a large-scale model for the construction machinery field for parsing, generating a carbon footprint calculation logic chain for the entire lifecycle of the target construction machinery product; based on the carbon footprint calculation logic chain, it schedules the data layer and knowledge layer to obtain the corresponding carbon emission parameter data of the target construction machinery product; the data layer stores the energy and carbon dataset of the construction machinery product; the knowledge layer stores a knowledge graph specific to the construction machinery industry; it inputs the carbon emission parameter data into a mathematical calculation engine, performs calculations based on the carbon footprint calculation logic chain, and obtains the carbon footprint quantification result; based on the carbon footprint quantification result, it generates and outputs a visual report containing carbon footprint details and optimization suggestions.

[0105] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0106] The processor in the electronic device provided in this embodiment of the invention can call logical instructions in the memory to implement the above method. Its specific implementation method is the same as the aforementioned method implementation method and can achieve the same beneficial effects, which will not be repeated here.

[0107] This invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the methods provided in the above embodiments.

[0108] The specific implementation method is the same as the aforementioned method implementation method and can achieve the same beneficial effects, so it will not be repeated here.

[0109] This invention provides a computer program product, including a computer program that, when executed by a processor, implements the method described above.

[0110] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0111] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for tracking the carbon footprint of engineering machinery products based on a large model, characterized in that, include: Receive user requests for carbon footprint tracking of target construction machinery products; The carbon footprint tracking request is input into a large model in the field of construction machinery for parsing, generating a carbon footprint calculation logic chain for the entire life cycle of the target construction machinery product; Based on the carbon footprint calculation logic chain, the scheduling data layer and knowledge layer acquire the corresponding carbon emission parameter data of the target engineering machinery product; the data layer stores the energy and carbon dataset of the engineering machinery product. The knowledge layer stores a knowledge graph specific to the engineering machinery industry; The carbon emission parameter data is input into the mathematical calculation engine, and calculations are performed based on the carbon footprint calculation logic chain to obtain the carbon footprint quantification result. Based on the carbon footprint quantification results, a visual report containing carbon footprint details and optimization suggestions is generated and output.

2. The carbon footprint tracking method for engineering machinery products based on a large model according to claim 1, characterized in that, The large-scale model in the field of engineering machinery was obtained through the following steps: Construct a training dataset; the training dataset includes at least one of domain question-answer pairs, computation rule instruction sets, and structured data text pairs; A low-rank adaptive fine-tuning method is used to fine-tune the pre-trained large language model based on the training dataset to obtain the fine-tuned adapter weights. The adapter weights are merged with the original weights of the pre-trained large language model to generate the large model for the engineering machinery domain.

3. The carbon footprint tracking method for engineering machinery products based on a large model according to claim 1, characterized in that, The energy carbon dataset was obtained based on the following steps: Acquire multi-source heterogeneous carbon data and preprocess the multi-source heterogeneous carbon data; The preprocessed multi-source heterogeneous energy carbon data is standardized to obtain standardized energy carbon data. The standardized energy and carbon data are semantically aligned using a cross-modal similarity method to construct the energy and carbon dataset.

4. The carbon footprint tracking method for engineering machinery products based on a large model according to claim 1, characterized in that, The knowledge graph specific to the construction machinery industry was obtained based on the following steps: Representative entities are extracted from standard documents and product manuals; the representative entities include at least one of standard regulatory entities, product entities, process entities, and emission factor entities. Based on the semantic relationships between the representative entities, a semantic relationship network is established to obtain the knowledge graph specific to the engineering machinery industry; the semantic relationships include at least one of following, containing, consuming, corresponding, and supplying.

5. The carbon footprint tracking method for engineering machinery products based on a large model according to claim 4, characterized in that, The method further includes: Attribute information data is attached to the representative entity; the attribute information data includes at least one of numerical data, unit data, source data, and effective status data. Based on the newly released policy and regulatory texts, update the attribute information data and semantic relationships of the corresponding entities in the construction machinery industry-specific knowledge graph.

6. The carbon footprint tracking method for engineering machinery products based on a large model according to claim 1, characterized in that, Based on the carbon footprint quantification results, a visual report containing carbon footprint details and optimization suggestions is generated and output, including: Generate a heat map of the carbon footprint of the target engineering machinery product throughout its entire life cycle; In response to a user's drilling operation on any stage of the full lifecycle carbon footprint heatmap, fine-grained data for that stage is displayed.

7. A carbon footprint tracking device for engineering machinery products based on a large model, characterized in that, include: The response module is used to receive carbon footprint tracking requests from users for target construction machinery products; The parsing module is used to input the carbon footprint tracking request into a large model in the field of construction machinery for parsing, and generate a carbon footprint calculation logic chain for the entire life cycle of the target construction machinery product; The retrieval module is used to schedule the data layer and knowledge layer to obtain the corresponding carbon emission parameter data of the target construction machinery product based on the carbon footprint calculation logic chain; the data layer stores the energy and carbon dataset of the construction machinery product. The knowledge layer stores a knowledge graph specific to the engineering machinery industry; The calculation module is used to input the carbon emission parameter data into the mathematical calculation engine, perform calculations based on the carbon footprint calculation logic chain, and obtain the carbon footprint quantification result. The presentation module is used to generate and output a visual report containing carbon footprint details and optimization suggestions based on the carbon footprint quantification results.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the carbon footprint tracking method for engineering machinery products based on a large model as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the carbon footprint tracking method for engineering machinery products based on a large model as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the carbon footprint tracking method for engineering machinery products based on a large model as described in any one of claims 1 to 6.

Images