Engineering management system and method based on AI large model

Abstract

The invention relates to the technical field of AI processing, and discloses an engineering management system and method based on an AI large model, and the management system is composed of an intelligent image collection module, a multi-dimensional data analysis module, a real-time environment monitoring module, a self-inspection error correction module and a man-machine cooperation decision support module. The system has a periodical self-checking function, the accuracy of the model can be automatically checked every time after certain iteration, when it is detected that the confidence coefficient output by the model is lower than a preset threshold value, the system can be automatically switched to a standby model, it is ensured that high accuracy can still be kept, and through deep learning and transfer learning, the accuracy of the model is improved. The system can continuously obtain knowledge and adaptability from new data, improves model performance, combines images, sensor data and environment information, provides intelligent suggestions based on historical data analysis and a current environment, enhances reliability in a decision making process, and improves the decision making efficiency. The beneficial effect that the error detection on the engineering construction site caused by the algorithm error or bug is avoided is achieved.

Description

Technical Field

[0001] The present invention relates to the field of AI processing technology, and specifically to an engineering management system and method based on an AI big model. Background Art

[0002] Engineering projects are projects that are carried out through engineering construction. They are one-time engineering construction tasks that are managed. Generally, when conducting construction safety supervision, it is necessary to set up multiple supervisors at the construction site, which causes labor costs. In addition, when supervisors conduct construction safety supervision at the construction site, they usually adopt the method of patrol inspection, which makes it difficult to conduct comprehensive safety supervision of the construction site and improve supervision efficiency.

[0003] In order to address the defects of the existing technology, an AI-based construction management method and system with publication number CN 118504962 A is proposed. An AI detection model is introduced to obtain construction images, and construction risk detection is performed according to the AI ​​detection model. There is no need to set up multiple supervisors at the construction site, which reduces labor costs and improves the comprehensiveness and efficiency of construction safety supervision at the construction site.

[0004] Although the aforementioned technical solution uses an AI detection model to obtain construction images and perform construction risk detection based on the AI ​​detection model, construction images also require photographic equipment and operators to input data. Even if the method of installing photographic equipment at a fixed workstation + automatic input is used, it is impossible to avoid erroneous detection caused by algorithm errors or logical bugs. In particular, the model cannot adapt to new construction environments or changes, resulting in missed reports or false alarms. Summary of the invention

[0005] Technical issues solved:

[0006] In view of the deficiencies in the prior art, the present invention provides an engineering management system and method based on an AI large model, which has the advantages of not causing incorrect detection of the engineering construction site due to algorithm errors or bugs, and the model can adaptively adjust as the construction site environment changes to avoid errors and omissions, thereby solving the problems of the above-mentioned technologies.

[0007] Technical solution:

[0008] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: an engineering management system based on an AI big model, the management system consisting of an intelligent image acquisition module, a multi-dimensional data analysis module, a real-time environment monitoring module, a self-checking and error correction module, and a human-machine collaborative decision support module;

[0009] The intelligent image acquisition module uses high-resolution cameras and drone technology to collect construction site images in real time, with multiple video inputs to obtain comprehensive data from different angles and heights;

[0010] The multi-dimensional data analysis module uses deep learning and transfer learning technology to build a large deep neural network model to detect construction risks. The model automatically optimizes according to new construction data and environmental changes to reduce manual intervention;

[0011] The real-time environment monitoring module integrates multiple sensors: temperature, humidity, light, and gas detection to monitor environmental changes at the construction site. Through sensor data and combined with image information, the model is dynamically adjusted to ensure real-time adaptation to the actual conditions of the construction site;

[0012] The self-checking and error correction module is used to regularly check the model and identify potential algorithm errors or bugs. When an abnormal detection occurs, the system automatically switches to the redundant backup model for verification;

[0013] The human-machine collaborative decision support module is used to provide a visual interface to display risk detection results and environmental monitoring data, support engineering managers in making decisions, and provide intelligent suggestions by combining historical data analysis with application scenarios;

[0014] The model checking expression in the self-checking and error correction module is:

[0015]

[0016] Where: A is the model accuracy; TP is the number of true positives, the number of positive examples correctly identified by the model; FP is the number of false positives, the number of negative examples incorrectly identified as positive by the model; TN is the number of true negatives, which refers to the number of instances correctly identified as negative by the model, and FN is the number of false positives, which refers to the number of instances that the model incorrectly identifies as positive.

[0017] The abnormal judgment expression is:

[0018]

[0019] Where: θ is the threshold, C is the specific confidence level, and if the confidence level of the system output is lower than the threshold, it is considered that an abnormality has occurred.

[0020] Preferably, the autonomous inspection description expression in the self-checking and error correction module is:

[0021] Self-checking process: every f=10 iterations, check the model accuracy

[0022] Where: f is the self-check cycle, which means that the model performs an accuracy check every f iterations;

[0023] The redundant model verification expression is:

[0024]

[0025] Among them: Abackup To ensure the accuracy of the backup model; Tp backup is the number of true positive examples of the backup model; FP backup is the number of false positive examples of the backup model; TN backup FN is the accuracy of the backup model in identifying negative classes. The higher the value, the better the backup model performs in identifying negative examples. backup How many positive classes are missed by the backup model. The lower this value is, the more effective the backup model is in identifying positive samples.

[0026] Preferably, the intelligent image acquisition module is expressed as:

[0027]

[0028] Where: i is a different camera or video source, C is the resolution of the high-resolution camera; A i is the angle of the camera, indicating the configuration of the camera at different angles; H i The altitude of the drone.

[0029] Preferably, the expression of the learning model in the multi-dimensional data analysis module is:

[0030] M(D1)=σ(W·DD+b)

[0031] Where: M(D1) is the predicted output of the deep learning model for the input data DD; b is the bias term; W is the weight matrix; σ is the activation function;

[0032] The knowledge transfer expression of the transfer learning is:

[0033]

[0034] in: Represents the process of transfer learning, using the source domain data D source To train the model to optimize the target domain data D target Model M target ;

[0035] The optimization objective expression is:

[0036]

[0037] Among them: θ is the model parameter, is the loss function, which measures the difference between the model output and the true value.

[0038] Preferably, the risk detection indicator expression in the multi-dimensional data analysis module is:

[0039]

[0040] Where: TP is the number of true positive cases; f(D,E,T,R) is a function that uses construction data D, environmental changes E, transfer learning T, and model parameters R to calculate the risk score;

[0041] The dynamic model update expression based on environmental factors in the multi-dimensional data analysis module is:

[0042] M updated (D,E)=M(D)+δ(E)

[0043] Where: M updated (D, E) is the updated model based on the new construction data D and environmental change E; δ(E) is a function that represents the response or correction to the environmental change factor E.

[0044] Preferably, the expression of the environment detection model in the human-machine collaborative decision support module is:

[0045] M adjusted =M(D2)+δ(S,I)

[0046] Where: M adjusted is the adjusted model; δ(S,I) is the dynamic adjustment function based on sensor data and image information; M(D2) is the original model; S is a vector containing data from multiple sensors: temperature data, humidity data, light data, and gas detection data.

[0047] Preferably, the δ(S,I)) is further expressed as a weighted sum:

[0048] δ(S,I)=w S ·S+w I I

[0049] Where: w S is the weight vector of sensor data; I represents the image information of the construction site, which is the image data captured by the camera in real time; w S is a weight vector used to weight the impact of sensor data S; w I It is the weighting coefficient of the image information I, which indicates the importance of the image information in the model adjustment.

[0050] Preferably, the comprehensive expression of the real-time environment monitoring module is: adjusted =M(D)+(w S ·S+w I ·I).

[0051] Preferably, the intelligent suggestion expression in the human-machine collaborative decision support module is:

[0052]

[0053] Where: f i (H,S) represents the specific function that affects the intelligent suggestion; w i The weight of a specific factor is used to reflect the relative importance of each factor to the intelligent recommendation; H is historical data, which refers to various data collected during the project life cycle; S is the application scenario, which refers to the specific situation or context in the current construction environment;

[0054] f i (H,S) = δ i H+γ i ·S

[0055] Where: i is a parameter related to historical data, indicating the specific impact of historical data on intelligent suggestions; i It is a parameter related to the application scenario, indicating the specific impact of the application scenario on the intelligent suggestion;

[0056] The weight w of each factor in the intelligent suggestion expression i is a positive number: w i >0.

[0057] The engineering management system and method based on AI big model include the following steps:

[0058] Step 1: Data collection: Start the intelligent image acquisition module, use drones and high-resolution cameras to collect images from multiple angles and heights in real time, monitor the environmental data of the construction site in real time, and use integrated sensors: temperature, humidity, light, and gas detection for environmental monitoring;

[0059] Step 2: Data analysis: Use deep learning and transfer learning techniques to analyze the collected image data, conduct construction risk detection, and combine image information and sensor data to conduct real-time environmental monitoring and analysis;

[0060] Step 3: Self-check and error correction: Periodically self-check the model accuracy to identify potential algorithm errors. When the confidence level of the model output is lower than the preset threshold, automatically switch to the backup model for verification.

[0061] Step 4: Decision support: Through the decision support module, risk detection results and environmental monitoring data are displayed on a visual interface, and intelligent suggestions are provided in combination with historical data analysis and specific conditions of the construction site;

[0062] Step 5: Dynamic adjustment: Use environmental monitoring data to dynamically adjust model weights to adapt to real-time changes at the construction site;

[0063] Step 6: Continuous operation: Ensure the system continues to operate, collect real-time data and process it;

[0064] Step 7: Continuous optimization: Continuously optimize the model and system based on new data and new algorithms to improve system performance.

[0065] Compared with the prior art, the present invention provides an engineering management system and method based on an AI big model, which has the following beneficial effects:

[0066] 1. The present invention provides a periodic self-checking function through the system. After each iteration, the accuracy of the model will be automatically checked. When the confidence of the model output is detected to be lower than the preset threshold value, the system will automatically switch to the backup model to ensure that a high accuracy can be maintained. The main model and the backup model are adopted. Even if the main model has an error, the backup model can take over its function, thereby reducing the risk of single point failure. Through deep learning and transfer learning, the system can continuously obtain knowledge and adaptability from new data to improve model performance. Combining images, sensor data and environmental information, the system makes decisions through multi-source information to reduce the risks caused by single-source errors, monitors changes in the external environment in real time, and integrates these data into risk assessment to ensure that the model can be dynamically adjusted according to real-time conditions. By visually displaying risk detection results and environmental data, managers can directly understand the system output and improve their understanding of decision-making. The intelligent suggestions provided are based on historical data analysis and the current environment, which enhances the reliability of the decision-making process and achieves the beneficial effect of preventing incorrect detection of engineering construction sites due to algorithm errors or bugs.

[0067] 2. The present invention uses a variety of sensors to collect real-time data through the system, which can reflect the environmental changes at the construction site in real time, ensuring that the model obtains the latest information for calculation and judgment. By integrating image data and sensor data, the model can be analyzed from multiple dimensions, and the diversity of environmental changes will be reflected in real time through different data sources, thereby enhancing the model's adaptability to changes and reducing errors caused by a single information source. The system is designed with a dynamic adjustment mechanism that can automatically update the model's parameters based on real-time environmental data input, ensuring that the model can better handle new environmental conditions and improve the accuracy of the prediction, achieving the beneficial effect of the model adaptively adjusting to changes in the construction site environment to avoid errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 Schematic diagram of the system of the present invention. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] See also Figure 1 , an engineering management system based on an AI big model, the management system consists of an intelligent image acquisition module, a multi-dimensional data analysis module, a real-time environment monitoring module, a self-checking and error correction module, and a human-machine collaborative decision support module;

[0071] The intelligent image acquisition module uses high-resolution cameras and drone technology to collect construction site images in real time, with multiple video inputs to obtain comprehensive data from different angles and heights;

[0072] The multi-dimensional data analysis module uses deep learning and transfer learning technology to build a large deep neural network model to detect construction risks. The model automatically optimizes according to new construction data and environmental changes to reduce manual intervention;

[0073] The real-time environment monitoring module integrates multiple sensors: temperature, humidity, light, and gas detection to monitor environmental changes at the construction site. Through sensor data and combined with image information, the model is dynamically adjusted to ensure real-time adaptation to the actual conditions of the construction site;

[0074] The self-checking and error correction module is used to regularly check the model and identify potential algorithm errors or bugs. When an abnormal detection occurs, the system automatically switches to the redundant backup model for verification;

[0075] The human-machine collaborative decision support module is used to provide a visual interface to display risk detection results and environmental monitoring data, support engineering managers in making decisions, and provide intelligent suggestions by combining historical data analysis with application scenarios;

[0076] The model checking expression in the self-checking and error correction module is:

[0077]

[0078] Where: A is the model accuracy; TP is the number of true positives, the number of positive examples correctly identified by the model; FP is the number of false positives, the number of negative examples incorrectly identified as positive by the model; TN is the number of true negatives, which refers to the number of instances correctly identified as negative by the model, and FN is the number of false positives, which refers to the number of instances that the model incorrectly identifies as positive.

[0079] The abnormal judgment expression is:

[0080]

[0081] Where: θ is the threshold, C is the specific confidence level, and if the confidence level of the system output is lower than the threshold, it is considered that an abnormality has occurred.

[0082] The system has a periodic self-check function, which automatically checks the accuracy of the model after a certain number of iterations (such as every 10 times). This regular evaluation can detect potential problems in a timely manner.

[0083] When it is detected that the confidence of the model output is lower than the preset threshold, the system will automatically switch to the backup model to ensure that a high level of accuracy is maintained.

[0084] By adopting the main model and backup model approach, even if the main model fails, the backup model can take over its function, thus reducing the risk of single point failure.

[0085] Through deep learning and transfer learning, the system can continuously acquire knowledge and adaptability from new data and improve model performance.

[0086] By combining images, sensor data, and environmental information, the system makes decisions based on multi-source information, reducing the risk of single-source errors.

[0087] Monitor changes in the external environment in real time and integrate this data into risk assessment to ensure that the model can be dynamically adjusted according to real-time conditions.

[0088] By visually displaying risk detection results and environmental data, managers can directly understand system outputs and improve their understanding of decision-making.

[0089] The intelligent recommendations provided are based on historical data analysis and the current environment, enhancing reliability in the decision-making process.

[0090] The system has the ability to receive and process new data, continuously update models and adjust algorithms to adapt to the current environment, and reduce errors caused by outdated models.

[0091] Through feedback from the construction site, the algorithm is continuously adjusted and optimized to improve the adaptability and accuracy of the model.

[0092] 6. Data backup and storage

[0093] Historical data storage: By storing historical data, it is easy to analyze and compare later, discover potential trends and anomalies in time, and ensure the long-term stability of the system.

[0094] Regular review and optimization: Use historical data to conduct regular reviews and optimizations to identify past algorithm errors or defects and make targeted repairs.

[0095] The engineering management system and method based on AI big model include the following steps:

[0096] Step 1: Data collection: Start the intelligent image acquisition module, use drones and high-resolution cameras to collect images from multiple angles and heights in real time, monitor the environmental data of the construction site in real time, and use integrated sensors: temperature, humidity, light, and gas detection for environmental monitoring;

[0097] Step 2: Data analysis: Use deep learning and transfer learning techniques to analyze the collected image data, conduct construction risk detection, and combine image information and sensor data to conduct real-time environmental monitoring and analysis;

[0098] Step 3: Self-check and error correction: Periodically self-check the model accuracy to identify potential algorithm errors. When the confidence level of the model output is lower than the preset threshold, automatically switch to the backup model for verification.

[0099] Step 4: Decision support: Through the decision support module, risk detection results and environmental monitoring data are displayed on a visual interface, and intelligent suggestions are provided in combination with historical data analysis and specific conditions of the construction site;

[0100] Step 5: Dynamic adjustment: Use environmental monitoring data to dynamically adjust model weights to adapt to real-time changes at the construction site;

[0101] Step 6: Continuous operation: Ensure the system continues to operate, collect real-time data and process it;

[0102] Step 7: Continuous optimization: Continuously optimize the model and system based on new data and new algorithms to improve system performance.

[0103] By integrating image data and sensor data, the model can analyze from multiple dimensions. The diversity of environmental changes will be reflected in real time through different data sources, thereby enhancing the model's ability to adapt to changes and reducing errors caused by a single information source.

[0104] A dynamic adjustment mechanism is designed to automatically update the model’s parameters based on real-time environmental data input (such as temperature changes or changes in construction conditions). This adaptive adjustment ensures that the model can better handle new environmental conditions and improves the accuracy of predictions.

[0105] The model has the ability to learn online and update in real time based on newly acquired data. This capability allows the model to continuously improve to cope with new environmental factors, thereby reducing the occurrence of errors and omissions.

[0106] When errors occur in the output results of the model, the model parameters are adjusted and optimized in a timely manner through the feedback mechanism to ensure the accuracy of subsequent judgments and predictions.

[0107] By displaying the environmental data and risk detection results of the construction site through a visual interface, managers can quickly identify changes and make more accurate decisions. This support enhances the ability to respond to environmental changes in a timely manner.

[0108] Regularly conduct self-checks to evaluate the model performance and determine whether adjustments are needed to adapt to the changing construction environment, ensuring that the model always maintains a high level of effectiveness and accuracy.

[0109] Store historical construction site data for subsequent analysis, trace trends, and understand which environmental factors may lead to potential errors and omissions, which can be used as a basis for adjusting the model

[0110] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. The engineering management system based on AI big model is characterized by: The management system consists of an intelligent image acquisition module, a multi-dimensional data analysis module, a real-time environment monitoring module, a self-checking and error correction module, and a human-machine collaborative decision support module; The intelligent image acquisition module uses high-resolution cameras and drone technology to collect construction site images in real time, with multiple video inputs to obtain comprehensive data from different angles and heights; The multi-dimensional data analysis module uses deep learning and transfer learning technology to build a large deep neural network model to detect construction risks. The model automatically optimizes according to new construction data and environmental changes to reduce manual intervention; The real-time environment monitoring module integrates multiple sensors: temperature, humidity, light, and gas detection to monitor environmental changes at the construction site. Through sensor data and combined with image information, the model is dynamically adjusted to ensure real-time adaptation to the actual conditions of the construction site; The self-checking and error correction module is used to regularly check the model and identify potential algorithm errors or bugs. When an abnormal detection occurs, the system automatically switches to the redundant backup model for verification; The human-machine collaborative decision support module is used to provide a visual interface to display risk detection results and environmental monitoring data, support engineering managers in making decisions, and provide intelligent suggestions by combining historical data analysis with application scenarios; The model checking expression in the self-checking and error correction module is: Where: A is the model accuracy; TP is the number of true positives, the number of positive examples correctly identified by the model; FP is the number of false positives, the number of negative examples incorrectly identified as positive by the model; TN is the number of true negatives, which refers to the number of instances correctly identified as negative by the model, and FN is the number of false positives, which refers to the number of instances that the model incorrectly identifies as positive. The abnormal judgment expression is: Where: θ is the threshold, C is the specific confidence level, and if the confidence level of the system output is lower than the threshold, it is considered that an abnormality has occurred.

2. The engineering management system based on AI big model according to claim 1 is characterized by: The autonomous inspection description expression in the self-checking and error correction module is: Self-checking process: every f=10 iterations, check the model accuracy Where: f is the self-check cycle, which means that the model performs an accuracy check every f iterations; The redundant model verification expression is: Among them: A backup To back up the accuracy of the model; TP backup is the number of true positive examples of the backup model; FP backup is the number of false positive examples of the backup model; TN backup FN is the accuracy of the backup model in identifying negative classes. The higher the value, the better the backup model performs in identifying negative examples. backup How many positive classes are missed by the backup model. The lower this value is, the more effective the backup model is in identifying positive samples.

3. The engineering management system based on AI big model according to claim 2 is characterized by: The expression of the intelligent image acquisition module is: Where: i is a different camera or video source, C is the resolution of the high-resolution camera; A i is the angle of the camera, indicating the configuration of the camera at different angles; H i The altitude of the drone.

4. The engineering management system based on AI big model according to claim 3 is characterized by: The expression of the learning model in the multi-dimensional data analysis module is: M(D1)=σ(W·DD+b) Where: M(D1) is the predicted output of the deep learning model for the input data DD; b is the bias term; W is the weight matrix; σ is the activation function; The knowledge transfer expression of the transfer learning is: in: Represents the process of transfer learning, using the source domain data D source To train the model to optimize the target domain data D target Model M target ; The optimization objective expression is: Among them: θ is the model parameter, is the loss function, which measures the difference between the model output and the true value.

5. The engineering management system based on AI big model according to claim 4 is characterized in that: The risk detection index expression in the multi-dimensional data analysis module is: Where: TP is the number of true positive cases; f(D,E,T,R) is a function that uses construction data D, environmental changes E, transfer learning T, and model parameters R to calculate the risk score; The dynamic model update expression based on environmental factors in the multi-dimensional data analysis module is: M updated (D,E)=M(D)+δ(E) Where: M updated (D, E) is the updated model based on the new construction data D and environmental change E; δ(E) is a function that represents the response or correction to the environmental change factor E.

6. The engineering management system based on AI big model according to claim 5 is characterized by: The expression of the environment detection model in the human-machine collaborative decision support module is: M adjusted =M(D2)+δ(S,I) Where: M adjusted is the adjusted model; δ(S,I) is the dynamic adjustment function based on sensor data and image information; M(D2) is the original model; S is a vector containing data from multiple sensors: temperature data, humidity data, light data, and gas detection data.

7. The engineering management system based on AI big model according to claim 6 is characterized by: The δ(S, I)) is further expressed as a weighted sum: δ(S,I)=w S ·S+w I ·I Where: w S is the weight vector of sensor data; I represents the image information of the construction site, which is the image data captured by the camera in real time; w S is a weight vector used to weight the impact of sensor data S; w I It is the weighting coefficient of the image information I, which indicates the importance of the image information in the model adjustment.

8. The engineering management system based on AI big model according to claim 7 is characterized by: The comprehensive expression of the real-time environment monitoring module is: adjusted =M(D)+(w S ·S+w I ·I).

9. The engineering management system based on AI big model according to claim 8 is characterized in that: The intelligent suggestion expression in the human-machine collaborative decision support module is: Where: f i (H, S) represents the specific function that affects the intelligent suggestion; w i The weight of a specific factor is used to reflect the relative importance of each factor to the intelligent recommendation; H is historical data, which refers to various data collected during the project life cycle; S is the application scenario, which refers to the specific situation or context in the current construction environment; f i (H, S)=δ i ·H+γ i ·S Where: i is a parameter related to historical data, indicating the specific impact of historical data on intelligent suggestions; i It is a parameter related to the application scenario, indicating the specific impact of the application scenario on the intelligent suggestion; The weight w of each factor in the intelligent suggestion expression i is a positive number: w i >0.

10. The engineering management method based on AI big model is characterized by: The management method is applied to the management system according to claim 1, and the steps of the management method are: Step 1: Data collection: Start the intelligent image acquisition module, use drones and high-resolution cameras to collect images from multiple angles and heights in real time, monitor the environmental data of the construction site in real time, and use integrated sensors: temperature, humidity, light, and gas detection for environmental monitoring; Step 2: Data analysis: Use deep learning and transfer learning techniques to analyze the collected image data, conduct construction risk detection, and combine image information and sensor data to conduct real-time environmental monitoring and analysis; Step 3: Self-check and error correction: Periodically self-check the model accuracy to identify potential algorithm errors. When the confidence level of the model output is lower than the preset threshold, automatically switch to the backup model for verification. Step 4: Decision support: Through the decision support module, risk detection results and environmental monitoring data are displayed on a visual interface, and intelligent suggestions are provided in combination with historical data analysis and specific conditions of the construction site; Step 5: Dynamic adjustment: Use environmental monitoring data to dynamically adjust model weights to adapt to real-time changes at the construction site; Step 6: Continuous operation: Ensure the system continues to operate, collect real-time data and process it; Step 7. Continuous optimization: Continuously optimize the model and system based on new data and new algorithms to improve system performance.

Images