Building risk monitoring and evaluation method and system based on neural network
The BP neural network conducts construction risk monitoring and evaluation, which solves the problems of inefficient and difficult real-time monitoring of traditional manual monitoring, and realizes efficient and accurate risk assessment and real-time monitoring, improving the risk management capabilities of construction projects.
Patent Information
- Application Number
- CN202411061370.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-05
- Publication Date
- 2024-12-03
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional artificial construction risk monitoring and evaluation are inefficient, difficult to monitor in real time, and it is impossible to detect and deal with potential risk events in a timely manner, which increases project uncertainty.
BP neural network is used for construction risk monitoring and evaluation. By building a multi-layer feedforward neural network, using backpropagation algorithm to train the model, and automatically process risk evaluation indicators to achieve real-time monitoring and evaluation.
It improves the accuracy and efficiency of risk assessment, realizes real-time monitoring, enhances the comprehensiveness and in-depthness of risk assessment, and can quickly respond to and adapt to different environments and conditions to ensure the accuracy and reliability of monitoring results.
Smart Images

Figure CN119067439A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of construction engineering, and in particular to a method and system for monitoring and evaluating construction risks based on a neural network. Background Art
[0002] In construction projects, risk monitoring and assessment are key links to ensure the smooth progress of projects, protect personnel safety and improve project quality. Through risk monitoring and assessment, potential risk factors can be discovered and identified in a timely manner, so that measures can be taken in advance to prevent and respond, ensure that the project proceeds smoothly as planned, and avoid delays or interruptions caused by risk events. Conducting a comprehensive risk assessment at the beginning of a project helps identify risk factors that may lead to cost overruns. By formulating targeted preventive measures, project costs can be effectively controlled and unnecessary waste can be avoided.
[0003] Although traditional manual construction risk monitoring and assessment has played a certain role in the past, with the rapid development of the construction industry and technological advancement, its drawbacks have become increasingly prominent, mainly in the following aspects:
[0004] (1) Inefficiency: Manual monitoring requires on-site inspections and record keeping by personnel, which is not only time-consuming and labor-intensive, but also difficult to cover all potential risk points. In addition, the frequency and depth of manual monitoring are also limited by the number and capabilities of personnel.
[0005] (2) Difficulty in real-time monitoring: Manual monitoring can usually only be performed through regular inspections and records, and real-time monitoring cannot be achieved. This means that if a risk event occurs between two inspections, it may not be discovered and responded to in a timely manner, increasing the uncertainty of the project.
[0006] To this end, the present invention proposes a construction risk monitoring and assessment method and system based on neural network. Summary of the invention
[0007] In view of this, the embodiment of the present invention hopes to provide a method and system for monitoring and assessing building risks based on a neural network to solve or alleviate the technical problems existing in the prior art, that is, to solve the technical defects of low efficiency and difficulty in real-time monitoring of manual detection and assessment, and at least provide a beneficial option for this; the technical solution of the present invention is implemented as follows:
[0008] First, the construction risk monitoring and assessment method based on neural network:
[0009] 1. Overview:
[0010] The present invention aims to solve the above technical problems. This solution chooses to build and deploy a BP neural network (BackPropagation Neural Network), which is a multi-layer feedforward neural network that is trained through a back propagation algorithm and has strong nonlinear mapping and self-learning capabilities, making it very suitable for dealing with complex nonlinear problems in building risk monitoring and assessment.
[0011] (II) Technical solution:
[0012] To achieve the above technical objectives, the following steps S1 to S3 need to be performed.
[0013] 2.1 Step S1, read the risk assessment index system:
[0014] Read N risk assessment indicators and quantify them.
[0015] Suppose there are N risk assessment indicators, represented by x1, x2,…, x N ; Then each indicator x i Quantize to [0,1]. The quantization method is:
[0016]
[0017] Among them, min(x i ) and max(x i ) are the minimum and maximum values of the index xi in all samples. In this way, each index x i Will be converted to a value between 0 and 1, thus achieving quantization.
[0018] 2.2 Step S2, perform BP neural network operation:
[0019] The input vector X of the BP neural network model is set to X = [x1, x2, ..., x N ] T , where T is the transpose and the output is a specific index x i 'The predicted value corresponds to the risk level.
[0020] BP neural network includes:
[0021] 1) Input layer: The input layer has N nodes, corresponding to N risk assessment indicators.
[0022] 2) Hidden layer: M hidden layer nodes, the activation function is the sigmoid function, then the output of the hidden layer is:
[0023]
[0024] Among them, w ijis the weight from the input layer to the hidden layer, b j is the bias of the hidden layer node, is the sigmoid activation function.
[0025] 3) Output layer: The output layer has K nodes, corresponding to K different risk levels or risk probabilities. The output prediction value y of the output layer is k for:
[0026]
[0027] Among them, v jk is the weight from the hidden layer to the output layer, c k is the bias of the output layer node. Output y k Corresponding to a specific indicator x i '.
[0028] 2.3 Training Neural Network:
[0029] Although the BP neural network model is used in the above steps S1 to S2, the neural network model still needs to be trained before it can be actually deployed. The training method includes the introduction of the k-fold cross-validation method; the overall training steps are:
[0030] 1) Division:
[0031] Construct a data set D and randomly divide it into k subsets of equal size, namely D1, D2, ..., D k .
[0032] 2) Loop execution:
[0033] 2.1) Select D i As the test set D test , and the remaining subsets are combined as training set D train ;
[0034] 2.2) Use mean square error as the loss function E:
[0035]
[0036] Among them, t k is the target output, y k is the actual output of the network.
[0037] Then, the weights and biases are updated by gradient descent to minimize the loss function E.
[0038] 2.3) Use the test set D test Test the trained BP neural network model and record the recall rate.
[0039] 2.4) Calculate the average of k test results as the final evaluation of model performance.
[0040] 2.5) After the model converges or reaches the predetermined number of iterations, stop training. And select the weight v that minimizes the loss function E jk , bias c k , weight w ij , bias b j , the number of nodes N and the number of hidden layer nodes M are taken as the optimal model parameters.
[0041] 2.4 Step S3, Analytical Prediction Value y k :
[0042] The predicted value y k The corresponding specific indicator x i 'Analyze and obtain monitoring and evaluation information I.
[0043] I=f(y k ,x i ');
[0044] Among them, f is an analytical function, which is based on the output y of the neural network k and the corresponding specific index x i ' to generate monitoring and evaluation information. The specific form of this function depends on what information is to be extracted from the output and what is defined as "monitoring and evaluation information".
[0045] The implementation logic of the parsing function f includes:
[0046] 1) Determine y k With a specific indicator x i ′;
[0047] 2) According to y k The value of and / or its relationship to x i ′ to assess the risk level or probability;
[0048] 3) Generate monitoring and evaluation reports including risk level, probability, recommended measures, etc.
[0049] (III) Mechanism for solving technical problems:
[0050] 3.1 Improving detection and evaluation efficiency
[0051] BP neural network can automatically learn and extract features from input data without the need for tedious manual feature extraction and rule setting. This automated processing capability significantly improves the efficiency of detection and evaluation. Neural networks have the ability to process information in parallel and can perform calculations on multiple nodes simultaneously, thereby speeding up processing. This makes BP neural networks perform well when processing large-scale data and improves evaluation efficiency. Through training, BP neural networks can quickly learn the mapping relationship between input and output and respond quickly to new input data. This fast learning ability enables neural networks to quickly perform detection and evaluation in a real-time environment.
[0052] 3.2 Realize real-time monitoring:
[0053] BP neural network can receive input data in real time and process and respond quickly. This enables neural networks to be used in real-time monitoring scenarios. Neural networks have strong adaptability and can maintain stable performance in different environments and conditions. This enables BP neural networks to run stably for a long time in real-time monitoring systems, ensuring the accuracy and reliability of monitoring results. The network structure and parameters of the neural network can be adjusted and optimized according to actual needs. As the complexity and scale of monitoring tasks change, the processing capacity of the neural network can be expanded by increasing the number of network layers and the number of neurons to meet the needs of real-time monitoring.
[0054] Second, the building risk monitoring and assessment system based on neural network:
[0055] like Figure 2 As shown, the system includes a processor and a register coupled to the processor, wherein the register stores program instructions, and when the program instructions are executed by the processor, the processor executes the building risk monitoring and assessment method as described above.
[0056] Compared with the prior art, the present invention has the following beneficial effects:
[0057] (1) Improved accuracy and objectivity of risk assessment: By quantifying risk assessment indicators to the interval [0,1], the dimension and scale differences between different indicators are eliminated, so that each indicator has the same comparison basis in the neural network, which improves the accuracy of the assessment. The BP neural network has a strong nonlinear mapping ability and can handle complex nonlinear relationships, making risk assessment more accurate and objective. Even if there is a complex nonlinear relationship between risk assessment indicators and risk levels, the neural network can still effectively learn and predict.
[0058] (2) Improved efficiency and real-time performance of risk assessment: BP neural network can automatically learn and extract features from input data, without the need for tedious manual feature extraction and rule setting. This automated processing capability significantly improves the efficiency of risk assessment. Neural networks have the ability to process information in parallel and can perform calculations on multiple nodes simultaneously, thereby speeding up processing. This makes BP neural network perform well in processing large-scale risk assessment data and improves the real-time performance of the assessment.
[0059] (3) Enhanced comprehensiveness and depth of risk assessment: BP neural network can consider multiple risk assessment indicators at the same time and realize comprehensive evaluation of multiple indicators. This makes risk assessment more comprehensive and in-depth, and can more accurately reflect the overall risk status of the evaluated object. Through the learning of the hidden layer, BP neural network can extract deep features in the input data, which may have an important indicative role in risk assessment. This makes risk assessment more in-depth and detailed.
[0060] (4) Improved interpretability and scalability of the model: By analyzing the function f, the output yk of the neural network can be associated with the specific indicator xi' to generate a monitoring and evaluation report containing risk level, probability, recommended measures, etc. This makes the results of the model easier to understand and apply. Adjustability of model parameters: The network structure and parameters of the BP neural network can be adjusted and optimized according to actual needs. This enables the model to adapt to different risk assessment scenarios and needs, and improves the scalability of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0062] Figure 1 It is a schematic diagram of the method flow of the present invention;
[0063] Figure 2 It is a schematic diagram of the system composition of the present invention;
[0064] Figure 3 It is a schematic diagram of the BP neural network training method of the present invention. DETAILED DESCRIPTION
[0065] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below;
[0066] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.
[0067] Embodiment 1: Figure 1 As shown, this embodiment discloses a construction risk monitoring and assessment method based on a neural network, which includes the following steps S1 to S3.
[0068] In this embodiment, regarding step S1, read the risk assessment index system: determine all risk assessment indicators involved in building risk monitoring and assessment, which may include but are not limited to the safety of building structures, the durability of building materials, the operating conditions of building equipment, etc. Suppose there are N risk assessment indicators, and these risk assessment indicators are represented as x1, x2, ..., x N ;
[0069] In order to eliminate the dimension and scale differences between different indicators, each indicator x i Quantized to [0,1].
[0070] The quantification method is:
[0071]
[0072] Among them, min(x i ) and max(x i ) are the minimum and maximum values of the index xi in all samples. In this way, each index x i Will be converted to a value between 0 and 1, thus achieving quantization.
[0073] It can be understood that quantization is to convert the original data into a form that is easier to process and compare. Here, we use the linear proportional transformation method to map the value of each risk assessment indicator to the interval [0,1]. The purpose of this is to eliminate the dimension and scale differences between different indicators so that they have the same basis for comparison. Through quantization, it can be ensured that the neural network model receives unified and standardized input data. This helps to improve the training efficiency and prediction accuracy of the model. At the same time, quantization also makes it easier for us to compare the risk levels between different samples.
[0074] Through quantification, we eliminate the dimension and scale differences between different indicators, so that each indicator has the same basis for comparison in the neural network. This helps to improve the accuracy and objectivity of risk assessment. Quantification makes it easier for the neural network model to learn the common features between different samples, thereby enhancing the generalization ability of the model. This means that the model can also make accurate predictions and evaluations when faced with new and unseen data.
[0075] Furthermore, the python execution program of step S1 is as follows:
[0076]
[0077]
[0078] Among them, an empty array quantized_data with the same shape as the original data data is created to store the quantized risk assessment index values. The numpy library is used here to handle array operations because it provides efficient and convenient numerical calculation functions. Then a loop is used to traverse each risk assessment indicator. For each indicator, we need to calculate its minimum and maximum values in all samples for quantification. For each indicator, we use the formula of linear proportional transformation for quantification. This formula maps the original indicator value to the interval [0,1], thereby eliminating the dimension and scale differences between different indicators. The results of the quantization process are stored in the quantized_data array. Finally, an array quantized_data containing the quantized risk assessment index values is obtained. This array will be used as input data for subsequent neural network models.
[0079] In this embodiment, regarding step S2, BP neural network operation is performed: BP neural network operation is performed to realize monitoring and assessment of building risks. BP neural network is a multi-layer feedforward neural network, which is trained by back propagation algorithm and can learn the complex mapping relationship between input data and output data.
[0080] The input vector X of the BP neural network model is set to X = [x1, x2, ..., x N ] T , where T is the transpose, x1,x2,…,x N is the quantitative value of N risk assessment indicators. The output is a specific indicator x i The predicted value of ' corresponds to the risk level. In practical applications, we can set multiple output nodes according to needs to predict multiple risk indicators or risk levels.
[0081] Specifically, the BP neural network includes:
[0082] 1) Input layer: The input layer has N nodes, corresponding to N risk assessment indicators. The input of each node is the quantitative value of the risk assessment indicator.
[0083] 2) Hidden layer: M hidden layer nodes, the activation function is sigmoid function. The output of the hidden layer is obtained by weighted summation and adding bias, and then applying sigmoid activation function. The output of the hidden layer is:
[0084]
[0085] Among them, w ij is the weight from the input layer to the hidden layer, b j is the bias of the hidden layer node, is the sigmoid activation function.
[0086] 3) Output layer: The output layer has K nodes, corresponding to K different risk levels or risk probabilities. The output of the output layer is also obtained by weighted summation and bias, and then applying the sigmoid activation function; the output prediction value y of the output layer k for:
[0087]
[0088] Among them, v jk is the weight from the hidden layer to the output layer, c k is the bias of the output layer node. Output y k Corresponding to a specific indicator x i '.
[0089] It can be understood that the BP neural network is learned and trained through two processes: forward propagation and back propagation. In the forward propagation process, the input data passes through the input layer, hidden layer, and output layer to obtain the predicted result. In the back propagation process, the weights and biases in the network are adjusted according to the error between the predicted result and the actual result to reduce the error. Through multiple iterations of training, the BP neural network can learn the mapping relationship between the input data and the output data. In the building risk monitoring and assessment method, the BP neural network, as the core algorithm, is responsible for processing the quantified risk assessment indicators and outputting the predicted risk level or risk probability. Through training and learning, the BP neural network can capture the complex relationship between the risk assessment indicators and the risk level, thereby achieving accurate prediction and assessment. Since the BP neural network has a fast operation speed and can process a large amount of data, it can realize real-time monitoring and assessment of building risks. Compared with traditional manual detection and assessment methods, this greatly improves the efficiency and real-time performance of monitoring and assessment.
[0090] Furthermore, the Python execution program of step S2 is as follows:
[0091]
[0092]
[0093]
[0094] In the above program, the parameters of the neural network are set, including the size of the input layer, hidden layer, and output layer, and the weights and biases are randomly initialized. Then the forward propagation function is defined, which accepts the input data X and calculates the predicted output through the neural network. During the forward propagation process, the sigmoid activation function is used to calculate the output of the hidden layer and the output layer.
[0095] A prediction function is defined that calls the forward propagation function to calculate the predicted risk level for a given input data. Finally, a loop is used to iterate over all samples and make a prediction for each sample. We reshape the individual samples into a two-dimensional array to fit the input of the neural network and print out the predicted risk level.
[0096] In this embodiment, regarding step S3, the analytical prediction value y k : The predicted value y k The corresponding specific indicator x i 'Analyze and obtain monitoring and evaluation information I.
[0097] I=f(y k ,x i ');
[0098] Among them, f is an analytical function, which is based on the output y of the neural networkk and the corresponding specific index x i ' to generate monitoring and evaluation information. The specific form of this function depends on what information is to be extracted from the output and what is defined as "monitoring and evaluation information".
[0099] The implementation logic of the parsing function f includes:
[0100] S300, determine the corresponding relationship: for each predicted value y output by the neural network k You need to know which specific monitoring indicator x it corresponds to. i '. The implementation of this step relies on the preprocessing and labeling of input data during the neural network training phase, ensuring that each input feature is associated with a specific monitoring indicator.
[0101] S301, assess the risk level or probability: according to the predicted value y k and / or its relationship to a specific indicator x i 'relationship to assess the risk level or probability. This usually involves converting continuous prediction values into discrete risk levels (such as "low", "medium", "high"), or calculating the probability of a specific event occurring. The implementation of this step may rely on preset thresholds or mapping relationships to convert prediction values into risk levels or probabilities.
[0102] S302, Generate Monitoring and Assessment Report: Generate a monitoring and assessment report containing risk level, probability, recommended measures, etc. This report is the output that the end user will see, and it should clearly and accurately convey the current risk status of the building and the recommended response measures. Report generation involves formatting the assessment results, adding necessary explanatory text, and providing specific recommended measures based on risk level or probability.
[0103] It is understandable that by automatically parsing the predicted values of the neural network, we can quickly and accurately generate monitoring and evaluation reports, greatly improving the efficiency of traditional manual detection and evaluation. Since the prediction of the neural network is real-time, our monitoring and evaluation reports can also be generated in real time, solving the technical defects of traditional methods that are difficult to monitor in real time. By analyzing the function f, we can provide customized risk response suggestions based on the specific predicted values and specific indicator relationships, making monitoring and evaluation more targeted and practical.
[0104] Furthermore, the Python execution program of step S3 is as follows:
[0105]
[0106]
[0107]
[0108] In the above program, a parse_prediction function is defined, which accepts the predicted value yk of the neural network, the specific indicator xi_prime corresponding to the predicted value, the risk level threshold risk_thresholds, and the recommended measures for different risk levels as input. Then, the function determines the risk level based on the predicted value and the risk level threshold, calculates the risk probability, obtains the recommended measures, and generates a monitoring and evaluation information dictionary containing specific indicators, risk levels, risk probabilities, and recommended measures.
[0109] Embodiment 2: Figure 3 As shown. This embodiment discloses a training method for the BP neural network model in the first embodiment. The neural network model still needs to be trained before it can be actually deployed. Its training method includes introducing the k-fold cross-validation method; the overall training steps are:
[0110] 1) Division:
[0111] Construct a dataset D that covers all possible situations of building risks and randomly divides it into k subsets of equal size, namely D1, D2, ..., D k .
[0112] 2) Loop execution:
[0113] 2.1) Select D i As the test set D test , and the remaining subsets are combined as training set D train ;
[0114] 2.2) Use mean square error as the loss function E:
[0115]
[0116] Among them, t k is the target output, y k is the actual output of the network. This formula can accurately reflect the gap between the model prediction value and the actual value.
[0117] Then, the weights and biases are updated by gradient descent to minimize the loss function E.
[0118] 2.3) Use the test set D test Test the trained BP neural network model and record the recall rate.
[0119] 2.4) Calculate the average of k test results as the final evaluation of model performance.
[0120] 2.5) After the model converges or reaches the predetermined number of iterations, stop training. And select the weight v that minimizes the loss function E jk , bias c k , weight w ij , bias b j , the number of nodes N and the number of hidden layer nodes M are taken as the optimal model parameters.
[0121] It is understandable that through the k-fold cross-validation method, the model can be trained and validated on multiple different data subsets, thereby enhancing its performance on new data. During the training process, by continuously monitoring the changes in the recall rate and loss function, we can ensure that the model remains stable until it reaches a convergence state. Using the mean square error as the loss function and combining the gradient descent method to optimize the weights and biases can significantly improve the accuracy of the model's predictions. By selecting the optimal model parameters and validating them on multiple data subsets, we can provide strong guarantees for the actual deployment of the model.
[0122] Furthermore, the Python training program for this step is as follows:
[0123]
[0124]
[0125]
[0126]
[0127] In the above program, the dataset D and its corresponding label set labels are first prepared. Then the relevant parameters of the neural network are set, including the size of the input layer, hidden layer, and output layer, as well as the learning rate and number of iterations.
[0128] The weights and biases of the neural network are randomly initialized. During the forward propagation process, we calculate the predicted output of the neural network. Then we use the mean squared error function to calculate the loss between the predicted output and the actual label.
[0129] In the back-propagation process, we first calculate the error terms of the output layer and the hidden layer. Then we update the weights and biases of the neural network according to the error terms to minimize the loss function.
[0130] The k-fold cross validation method is used to divide the dataset and perform training and testing separately. In each iteration, we print out the current loss value for monitoring. Finally, the performance of the model is evaluated on the test set and the test loss is printed.
[0131] For those skilled in the art, it can be further appreciated that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0132] At the same time, those skilled in the art can understand that all or part of the processes in all the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0133] The above embodiments only express the implementation methods of the relevant practical applications of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.
Claims
1. A method for monitoring and assessing building risks based on a neural network, characterized in that: The steps include: S1, read the risk assessment indicator system: read N risk assessment indicators and quantify them; S2, execute BP neural network operation: the input vector X of the BP neural network model is expressed as X = [x1, x2, ..., x N ] T , where T is the transpose and the output is a specific index x i 'The predicted value y k , which corresponds to the risk level; S3, analytical prediction value y k : The predicted value y k The corresponding specific indicator x i 'Analyze and obtain monitoring and evaluation information I.
2. The building risk monitoring and assessment method according to claim 1, characterized in that: In S1, N risk assessment indicators are represented by x1, x2, ..., x N ; Then each indicator x i Quantize to [0,1]: Among them, min(x i ) and max(x i ) are the minimum and maximum values of the indicator xi in all samples respectively.
3. The building risk monitoring and assessment method according to claim 1, characterized in that: In S2, the BP neural network includes an input layer corresponding to N risk assessment indicators, a hidden layer corresponding to M hidden layer nodes, and an output layer corresponding to K different risk levels or risk probabilities.
4. The building risk monitoring and assessment method according to claim 3 is characterized in that: In the hidden layer, the activation function is the sigmoid function, and the output of the hidden layer is: Among them, w ij is the weight from the input layer to the hidden layer, b j is the bias of the hidden layer node, is the sigmoid activation function.
5. The building risk monitoring and assessment method according to claim 3, characterized in that: In the output layer, the output predicted value y k for: Among them, v jk is the weight from the hidden layer to the output layer, c k is the bias of the output layer node; output y k Corresponding to a specific indicator x i '.
6. The building risk monitoring and assessment method according to any one of claims 1 to 5, characterized in that: The training method of the BP neural network is to update the model parameters by the gradient descent method to minimize the loss function E.
7. The building risk monitoring and assessment method according to claim 6, characterized in that: The form of the loss function E is: Among them, t k is the target output, y k is the actual output of the network; weights v are chosen to minimize the loss function E jk , bias c k , weight w ij , bias b j , the number of nodes N and the number of hidden layer nodes M are taken as the optimal model parameters.
8. The building risk monitoring and assessment method according to any one of claims 1 to 5, characterized in that: In S3, the method for obtaining the monitoring and evaluation information I is: I = f(y k ,x i '); Among them, f is an analytical function, which is based on the output y of the neural network k and the corresponding specific index x i ' to generate monitoring and evaluation information.
9. The building risk monitoring and assessment method according to claim 8, characterized in that: The implementation logic of the analytical function f includes: Determine y k With a specific indicator x i The corresponding relationship between ′; According to y k The value of and / or its relationship to x i ′ to assess the risk level or probability; Generate monitoring and assessment reports including risk level, probability and recommended actions.
10. The construction risk monitoring and assessment system based on neural network is characterized by: The system includes a processor and a register coupled to the processor, wherein the register stores program instructions, and when the program instructions are executed by the processor, the processor executes the building risk monitoring and assessment method as described in any one of claims 1-9.
