Construction project safety intelligent risk assessment system and method based on artificial intelligence

Abstract

The invention discloses a construction engineering safety intelligent risk assessment system and method based on artificial intelligence, and relates to the technical field of artificial intelligence, and the system comprises a preprocessing module, a monitoring prediction module, an anomaly analysis module and a classification assessment module. The preprocessing module is used for collecting historical state data of a construction project, performing manual labeling and forming a data set; the monitoring and predicting module is used for monitoring construction project state data in real time and predicting according to historical data; the exception analysis module is used for calculating the LOF value of each sample and setting an exception threshold in combination with a labeling condition; and the classification evaluation module is used for calculating and monitoring the LOF value of the new state data and predicting the state data, and judging classification according to an abnormal threshold so as to carry out risk evaluation. The invention also provides a method for realizing the system. The method can effectively improve the situation that abnormal sample threshold setting cannot adapt to continuous changes in an engineering environment in the prior art.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based construction engineering safety intelligent risk assessment system and method. Background Art

[0002] Safety is of vital importance in construction projects. Intelligent risk assessment can help identify potential safety hazards and take measures to prevent and control them in advance, thereby reducing the occurrence of accidents and protecting the lives and property of construction workers. At the same time, it can also help improve the quality and progress of the project and reduce the cost of the project.

[0003] The existing intelligent risk assessment technology for construction project safety first collects data information on various aspects such as changes in engineering structures and on-site environment by deploying various sensors at the construction site and using drones for inspection. Then, the collected data is transmitted to the data processing center, and the massive data is mined and analyzed using big data analysis technology. At the same time, the data is trained and processed with the help of machine learning and artificial intelligence algorithms to establish a risk assessment model.

[0004] In the existing intelligent risk assessment process for construction project safety, fixed thresholds are used to judge anomalies. This static threshold setting cannot adapt to the ever-changing situation in the engineering environment, resulting in unstable assessment results. In addition, the existing classification models rely more on the accuracy of manual labeling, and the subjectivity will affect the training effect and generalization ability of the model, and it also lacks good interpretability. Summary of the invention

[0005] The purpose of the present invention is to provide an artificial intelligence-based construction project safety intelligent risk assessment system and method to solve the problems raised in the prior art.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: an intelligent risk assessment system for construction project safety based on artificial intelligence, the system comprising a preprocessing module, a monitoring and prediction module, an abnormality analysis module and a classification and assessment module; The preprocessing module is used to collect historical status data of construction projects, perform manual annotation and form a data set; the monitoring and prediction module is used to monitor the status data of construction projects in real time and make predictions based on historical data; the abnormal analysis module is used to calculate the LOF value of each sample and set the abnormal threshold based on the annotation situation; the classification and evaluation module is used to calculate the LOF value of the new monitoring status data and the predicted status data and judge the classification based on the abnormal threshold to make a risk assessment; The output end of the preprocessing module is connected to the input end of the monitoring and prediction module and the abnormality analysis module; the output end of the monitoring and prediction module is connected to the input end of the abnormality analysis module; the output end of the abnormality analysis module is connected to the input end of the classification and evaluation module.

[0007] The preprocessing module includes a data collection unit, a manual annotation unit and a database unit; The data collection unit is used to collect the historical status data of the construction project; the manual annotation unit is used to manually annotate the collected data samples; the database unit is used to form a data set based on the data and its corresponding annotations and continuously update it; The output end of the data collection unit is connected to the input end of the manual labeling unit; the output end of the manual labeling unit is connected to the input end of the database unit; the output end of the database unit is connected to the input ends of the monitoring and prediction module and the abnormality analysis module.

[0008] The monitoring and prediction module includes a real-time monitoring unit and a time series prediction unit; The real-time monitoring unit is used to monitor the construction project status data in real time; the time series prediction unit is used to predict the status data based on historical data; The output end of the real-time monitoring unit is connected to the input end of the time series prediction unit; the output end of the time series prediction unit is connected to the input end of the abnormality analysis module.

[0009] The abnormal analysis module includes a LOF calculation unit and a threshold setting unit; The LOF calculation unit is used to calculate the LOF values ​​of all samples; the threshold setting unit is used to set the abnormal threshold in combination with the LOF value and the annotation situation of each sample; The output end of the LOF calculation unit is connected to the input end of the threshold setting unit; the output end of the threshold setting unit is connected to the input end of the classification evaluation module.

[0010] The classification and assessment module includes a classification unit and a risk assessment unit; The classification unit is used to calculate the LOF value of the monitoring new state data and the predicted state data and judge the classification according to the abnormal threshold; the risk assessment unit is used to make risk assessment according to different classification situations; The output end of the classification unit is connected to the input end of the risk assessment unit.

[0011] The construction project safety intelligent risk assessment method based on artificial intelligence includes the following steps: S1. Collect historical status data of construction projects and annotate each sample manually; S2, perform anomaly analysis on the sample, calculate the sample LOF value, and set the anomaly threshold based on the labeled category; S3, monitor construction project characteristic data in real time and make abnormal judgments to assess risks; S4. Combine historical data and real-time monitoring data to predict future data, and perform anomaly analysis on future data to assess risks.

[0012] In step S1, the engineering history status feature data is collected: [F1, F2, …, F n ]; Where n represents the number of features, F n represents the nth feature; The characteristic data refers to characteristic parameters reflecting the status of the construction project in different aspects; Each sample is annotated manually and the annotation categories are [A, B1, B2, …, B m ]; Among them, A indicates that the sample is a normal sample; B1~B m Indicates that the sample is an abnormal sample, m is a positive integer, indicating the number of abnormal types, B m Represents the mth abnormal sample category; The labeling classification refers to the qualitative classification of the engineering history status samples; A database is formed, which includes historical status data of construction projects and their corresponding annotation categories.

[0013] In step S2, the feature data in the database are standardized and digitized; the LOF value of each sample is calculated: for each sample p, the distance between it and other samples is calculated using the Euclidean distance; d(p,o) represents the distance between sample p and sample o, ; Select the k nearest neighbor samples, marked as N k (p); Where k is a positive integer; N k (p) represents the set of k samples closest to sample p; i is a positive integer, 1≤i≤n, F i p represents the i-th feature of sample p; F i o represents the i-th feature of sample o; Calculate the reachable density ρk(p) of sample p: ; Among them, d(q,N k (q)) is the distance from sample q to its neighbor samples; Calculate the local outlier factor: ; Among them, |N k (p)| represents the number of samples in the set; Calculate the LOF value of each sample and mark it as B1~B m The samples are integrated as abnormal samples, and the samples marked as A are normal samples; the average value of the LOF values ​​of all normal samples is marked as U, and the average value of the LOF values ​​of all abnormal samples is marked as V; Divide U and V into equal parts by I, and take the threshold [U,U+(VU) / I,U+2(VU) / I,…,U+(VU)(I-1) / I,V]; All samples are compared with different thresholds according to their LOF values, and samples with LOF values ​​greater than the threshold are marked as abnormal, otherwise they are marked as normal; Calculate the accuracy corresponding to different thresholds: E=(TP+TN) / (TP+TN+FP+FN); Among them, TP represents true positive examples, which are actually abnormal and marked as abnormal samples; TN represents true negative examples, which are actually normal and marked as normal samples; FP represents false positive examples, which are actually normal but marked as abnormal samples; FN represents false negative examples, which are actually abnormal but marked as normal samples. Get different accuracy rates corresponding to different thresholds, build a regression model with the threshold as input and the accuracy as output to obtain the polynomial regression equation: E(T)=a0+a1T+a2T 2 +…+a r T r ; Where T is the threshold; E(T) is the accuracy corresponding to different thresholds T; r is the polynomial order; a0, a1, …, a r For the parameters to be estimated, the least squares method is used to estimate the error sum between the actual accuracy and the model prediction accuracy to be minimized; Derivate the polynomial equation and set the derivative equal to 0, solve the equation to obtain possible extreme points, substitute the extreme points into the original polynomial regression equation, compare the accuracies corresponding to different extreme points, and select the extreme point corresponding to the highest accuracy as the abnormal threshold z.

[0014] In step S3, the construction project characteristics are monitored in real time, and the acquired new sample data is put into the database for updating; the LOF value of the new sample data is calculated and compared with the abnormal threshold z; when the LOF value of the new sample data is less than z, the new sample is judged to be a normal sample, the sample is marked as a normal sample, a risk assessment is made and output to the user, and the upload database is updated; when the LOF value of the new sample data is greater than or equal to z, the new sample is judged to be an abnormal sample, the Euclidean distance between the new sample and all samples in the database is calculated, the abnormal type corresponding to the sample with the closest distance is selected as the abnormal type of the new sample, a risk assessment is made and output to the user, and the upload database is updated.

[0015] In step S4, a time series model is constructed based on historical data to predict the characteristic data of construction projects at future times: the historical data and the characteristic data of real-time monitoring are used as input to construct an LSTM model, and the predicted characteristic values ​​at future times are output; The LOF value of the sample at the future moment is calculated based on the predicted data and compared with the abnormal threshold z; when the LOF value of the new sample data is less than z, the new sample is judged to be a normal sample, a risk assessment is made and output to the user; when the LOF value of the new sample data is greater than or equal to z, the new sample is judged to be an abnormal sample, the Euclidean distance between the new sample and all samples in the database is calculated, the abnormal type corresponding to the sample with the closest distance is selected as the abnormal type of the new sample, a risk assessment is made and output to the user.

[0016] Continuously update the model as the database is updated.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses Euclidean distance calculation to match new samples with historical abnormal samples and determine the most similar abnormal types. Refined classification helps to more accurately identify the specific source of risk, so as to take targeted countermeasures; 2. The present invention can dynamically calculate the degree of abnormality of each sample through the local abnormality factor and combined with the polynomial regression model, thereby improving the accuracy of risk assessment; LOF is based on the local density of samples and neighboring samples, which can better detect local abnormalities and reduce missed judgments and misjudgments; 3. The present invention adaptively adjusts the abnormal threshold through a polynomial regression model, and automatically determines the optimal threshold according to the distribution of LOF values ​​of normal samples and abnormal samples; this enables the system to dynamically adapt to the risk levels in different environments, thereby improving the flexibility and adaptability of risk judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of the construction engineering safety intelligent risk assessment system based on artificial intelligence of the present invention; Figure 2It is a schematic diagram of the steps of the construction project safety intelligent risk assessment method based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0019] 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.

[0020] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, an intelligent risk assessment system for construction project safety based on artificial intelligence, the system includes a preprocessing module, a monitoring and prediction module, an abnormality analysis module and a classification assessment module; The preprocessing module is used to collect historical status data of construction projects, perform manual annotation and form a data set; the monitoring and prediction module is used to monitor the status data of construction projects in real time and make predictions based on historical data; the abnormal analysis module is used to calculate the LOF value of each sample and set the abnormal threshold based on the annotation situation; the classification and evaluation module is used to calculate the LOF value of the new monitoring status data and the predicted status data and judge the classification based on the abnormal threshold to make a risk assessment; The output end of the preprocessing module is connected to the input end of the monitoring and prediction module and the abnormality analysis module; the output end of the monitoring and prediction module is connected to the input end of the abnormality analysis module; the output end of the abnormality analysis module is connected to the input end of the classification and evaluation module.

[0021] The preprocessing module includes a data collection unit, a manual annotation unit and a database unit; The data collection unit is used to collect the historical status data of the construction project; the manual annotation unit is used to manually annotate the collected data samples; the database unit is used to form a data set based on the data and its corresponding annotations and continuously update it; The output end of the data collection unit is connected to the input end of the manual labeling unit; the output end of the manual labeling unit is connected to the input end of the database unit; the output end of the database unit is connected to the input ends of the monitoring and prediction module and the abnormality analysis module.

[0022] The monitoring and prediction module includes a real-time monitoring unit and a time series prediction unit; The real-time monitoring unit is used to monitor the construction project status data in real time; the time series prediction unit is used to predict the status data based on historical data; The output end of the real-time monitoring unit is connected to the input end of the time series prediction unit; the output end of the time series prediction unit is connected to the input end of the abnormality analysis module.

[0023] The abnormal analysis module includes a LOF calculation unit and a threshold setting unit; The LOF calculation unit is used to calculate the LOF values ​​of all samples; the threshold setting unit is used to set the abnormal threshold in combination with the LOF value and the annotation situation of each sample; The output end of the LOF calculation unit is connected to the input end of the threshold setting unit; the output end of the threshold setting unit is connected to the input end of the classification evaluation module.

[0024] The classification and assessment module includes a classification unit and a risk assessment unit; The classification unit is used to calculate the LOF value of the monitoring new state data and the predicted state data and judge the classification according to the abnormal threshold; the risk assessment unit is used to make risk assessment according to different classification situations; The output end of the classification unit is connected to the input end of the risk assessment unit.

[0025] The construction project safety intelligent risk assessment method based on artificial intelligence includes the following steps: S1. Collect historical status data of construction projects and annotate each sample manually; S2, perform anomaly analysis on the sample, calculate the sample LOF value, and set the anomaly threshold based on the labeled category; S3, monitor construction project characteristic data in real time and make abnormal judgments to assess risks; S4. Combine historical data and real-time monitoring data to predict future data, and perform anomaly analysis on future data to assess risks.

[0026] In step S1, the engineering history status feature data is collected: [F1, F2, …, F n ]; Where n represents the number of features, F n represents the nth feature; The characteristic data refers to characteristic parameters reflecting the status of the construction project in different aspects; Each sample is annotated manually and the annotation categories are [A, B1, B2, …, B m ]; Among them, A indicates that the sample is a normal sample; B1~B m Indicates that the sample is an abnormal sample, m is a positive integer, indicating the number of abnormal types, B m Represents the mth abnormal sample category; The labeling classification refers to the qualitative classification of the engineering history status samples; A database is formed, which includes historical status data of construction projects and their corresponding annotation categories.

[0027] In step S2, the feature data in the database are standardized and digitized; the LOF value of each sample is calculated: for each sample p, the distance between it and other samples is calculated using the Euclidean distance; d(p,o) represents the distance between sample p and sample o, ; Select the k nearest neighbor samples, marked as N k (p); Where k is a positive integer; N k (p) represents the set of k samples closest to sample p; i is a positive integer, 1≤i≤n, F i p represents the i-th feature of sample p; F i o represents the i-th feature of sample o; Calculate the reachable density ρk(p) of sample p: ; Among them, d(q,N k (q)) is the distance from sample q to its neighbor samples; Calculate the local outlier factor: ; Among them, |N k (p)| represents the number of samples in the set; Calculate the LOF value of each sample and mark it as B1~B m The samples are integrated as abnormal samples, and the samples marked as A are normal samples; the average value of the LOF values ​​of all normal samples is marked as U, and the average value of the LOF values ​​of all abnormal samples is marked as V; Divide U and V into equal parts by I, and take the threshold [U,U+(VU) / I,U+2(VU) / I,…,U+(VU)(I-1) / I,V]; All samples are compared with different thresholds according to their LOF values, and samples with LOF values ​​greater than the threshold are marked as abnormal, otherwise they are marked as normal; Calculate the accuracy corresponding to different thresholds: E=(TP+TN) / (TP+TN+FP+FN); Among them, TP represents true positive examples, which are actually abnormal and marked as abnormal samples; TN represents true negative examples, which are actually normal and marked as normal samples; FP represents false positive examples, which are actually normal but marked as abnormal samples; FN represents false negative examples, which are actually abnormal but marked as normal samples. Get different accuracy rates corresponding to different thresholds, build a regression model with the threshold as input and the accuracy as output to obtain the polynomial regression equation: E(T)=a0+a1T+a2T 2 +…+a r T r ; Where T is the threshold; E(T) is the accuracy corresponding to different thresholds T; r is the polynomial order; a0, a1, …, a r For the parameters to be estimated, the least squares method is used to estimate the error sum between the actual accuracy and the model prediction accuracy to be minimized; Derivate the polynomial equation and set the derivative equal to 0, solve the equation to obtain possible extreme points, substitute the extreme points into the original polynomial regression equation, compare the accuracies corresponding to different extreme points, and select the extreme point corresponding to the highest accuracy as the abnormal threshold z.

[0028] In step S3, the construction project characteristics are monitored in real time, and the acquired new sample data is put into the database for updating; the LOF value of the new sample data is calculated and compared with the abnormal threshold z; when the LOF value of the new sample data is less than z, the new sample is judged to be a normal sample, the sample is marked as a normal sample, a risk assessment is made and output to the user, and the upload database is updated; when the LOF value of the new sample data is greater than or equal to z, the new sample is judged to be an abnormal sample, the Euclidean distance between the new sample and all samples in the database is calculated, the abnormal type corresponding to the sample with the closest distance is selected as the abnormal type of the new sample, a risk assessment is made and output to the user, and the upload database is updated.

[0029] In step S4, a time series model is constructed based on historical data to predict the characteristic data of construction projects at future times: the historical data and the characteristic data of real-time monitoring are used as input to construct an LSTM model, and the predicted characteristic values ​​at future times are output; Calculate the LOF value of the sample at the future moment based on the predicted data and compare it with the abnormal threshold z; when the LOF value of the new sample data is less than z, judge the new sample as a normal sample, make a risk assessment and output it to the user; when the LOF value of the new sample data is greater than or equal to z, judge the new sample as an abnormal sample, calculate the Euclidean distance between the new sample and all samples in the database, select the abnormal type corresponding to the sample with the closest distance as the abnormal type of the new sample, make a risk assessment and output it to the user. The model is continuously updated as the database is updated.

[0030] In this example: A large-scale construction project needs to implement a safety risk assessment to ensure the safety and quality of the construction process. The project has collected historical status data for multiple stages, involving structural strength, environmental factors, and construction progress; Step S1: Data collection and annotation: Data collection: Collect characteristic data, including but not limited to the following: structural strength, ambient temperature, ambient humidity, construction progress; Collect 1000 sample data; Manual labeling: Manually label each sample: normal sample, structural problem, environmental anomaly, construction delay; Each sample and its feature data and annotations are recorded in the database, forming the following structure: Sample ID, structural strength, ambient temperature, ambient humidity, construction progress, normal samples / structural problems / environmental abnormalities / construction delays; Step S2: Abnormal analysis: Data standardization: standardize the feature data; Calculate the LOF value: For each sample, calculate the Euclidean distance to other samples; select the k nearest neighbor samples; calculate the reachable density of the sample; calculate the local anomaly factor; Calculate the LOF value of each sample, integrate the samples marked as structural problems, environmental anomalies, and construction delays into abnormal samples, and the samples marked as normal samples into normal samples; calculate the average value of the LOF values ​​of all normal samples and mark it as U, and calculate the average value of the LOF values ​​of all abnormal samples and mark it as V; Divide U and V into equal parts by I, and take the threshold [U,U+(VU) / I,U+2(VU) / I,…,U+(VU)(I-1) / I,V]; All samples are compared with different thresholds according to their LOF values, and samples with LOF values ​​greater than the threshold are marked as abnormal, otherwise they are marked as normal; Calculate the accuracy corresponding to different thresholds: E=(TP+TN) / (TP+TN+FP+FN); Among them, TP represents true positive examples, which are actually abnormal and marked as abnormal samples; TN represents true negative examples, which are actually normal and marked as normal samples; FP represents false positive examples, which are actually normal but marked as abnormal samples; FN represents false negative examples, which are actually abnormal but marked as normal samples. Different accuracy rates corresponding to different thresholds are obtained, and a regression model is constructed with the threshold as input and the accuracy rate as output to obtain a polynomial regression equation; Derivate the polynomial equation and set the derivative equal to 0. Solve the equation to get the possible extreme points. Substitute the extreme points into the original polynomial regression equation, compare the accuracy rates corresponding to different extreme points, and select the extreme point corresponding to the highest accuracy rate as the abnormal threshold z; Step S3: Monitor the construction project characteristics in real time, and update the acquired new sample data in the database; calculate the LOF value of the new sample data and compare it with the abnormal threshold z; when the LOF value of the new sample data is less than z, judge the new sample as a normal sample, mark the sample as a normal sample, make a risk assessment and output it to the user, and update the uploaded database; when the LOF value of the new sample data is greater than or equal to z, judge the new sample as an abnormal sample, calculate the Euclidean distance between the new sample and all samples in the database, select the abnormal type corresponding to the sample with the closest distance as the abnormal type of the new sample, make a risk assessment and output it to the user, and update the uploaded database.

[0031] Step S4: Build a time series model based on historical data to predict construction project feature data at future times; calculate the LOF value of samples at future times based on the predicted data, and compare it with the abnormal threshold z; when the LOF value of the new sample data is less than z, the new sample is judged to be a normal sample, and a risk assessment is made and output to the user; when the LOF value of the new sample data is greater than or equal to z, the new sample is judged to be an abnormal sample, and the Euclidean distance between the new sample and all samples in the database is calculated, and the abnormal type corresponding to the sample with the closest distance is selected as the abnormal type of the new sample, and a risk assessment is made and output to the user. The model is continuously updated as the database is updated.

[0032] Through this method, the construction team can promptly identify potential risks and take preventive measures before problems occur, thus improving overall construction safety. The system will continue to optimize the accuracy and reliability of risk assessment as data is continuously updated.

[0033] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An artificial intelligence-based construction project safety intelligent risk assessment system, characterized by: The system includes a preprocessing module, a monitoring and prediction module, an abnormality analysis module, and a classification and evaluation module; The preprocessing module is used to collect the historical status data of the construction project, perform manual annotation and form a data set; the monitoring and prediction module is used to monitor the status data of the construction project in real time and make predictions based on the historical data; The anomaly analysis module is used to calculate the LOF value of each sample and set the anomaly threshold in combination with the annotation situation; the classification evaluation module is used to calculate the LOF value of the new monitoring state data and the predicted state data and judge the classification according to the anomaly threshold to make a risk assessment; The output end of the preprocessing module is connected to the input end of the monitoring and prediction module and the abnormality analysis module; the output end of the monitoring and prediction module is connected to the input end of the abnormality analysis module; the output end of the abnormality analysis module is connected to the input end of the classification and evaluation module.

2. The construction engineering safety intelligent risk assessment system based on artificial intelligence according to claim 1 is characterized by: The preprocessing module includes a data collection unit, a manual annotation unit and a database unit; The data collection unit is used to collect the historical status data of the construction project; the manual annotation unit is used to manually annotate the collected data samples; the database unit is used to form a data set based on the data and its corresponding annotations and continuously update it; The output end of the data collection unit is connected to the input end of the manual labeling unit; the output end of the manual labeling unit is connected to the input end of the database unit; the output end of the database unit is connected to the input ends of the monitoring and prediction module and the abnormality analysis module.

3. The construction engineering safety intelligent risk assessment system based on artificial intelligence according to claim 2 is characterized by: The monitoring and prediction module includes a real-time monitoring unit and a time series prediction unit; The real-time monitoring unit is used to monitor the construction project status data in real time; the time series prediction unit is used to predict the status data based on historical data; The output end of the real-time monitoring unit is connected to the input end of the time series prediction unit; the output end of the time series prediction unit is connected to the input end of the abnormality analysis module.

4. The construction engineering safety intelligent risk assessment system based on artificial intelligence according to claim 3 is characterized by: The abnormal analysis module includes a LOF calculation unit and a threshold setting unit; The LOF calculation unit is used to calculate the LOF values ​​of all samples; the threshold setting unit is used to set the abnormal threshold in combination with the LOF value and the annotation situation of each sample; The output end of the LOF calculation unit is connected to the input end of the threshold setting unit; the output end of the threshold setting unit is connected to the input end of the classification evaluation module.

5. The construction engineering safety intelligent risk assessment system based on artificial intelligence according to claim 4 is characterized by: The classification and assessment module includes a classification unit and a risk assessment unit; The classification unit is used to calculate the LOF value of the monitoring new state data and the predicted state data and judge the classification according to the abnormal threshold; the risk assessment unit is used to make risk assessment according to different classification situations; The output end of the classification unit is connected to the input end of the risk assessment unit.

6. An artificial intelligence-based intelligent risk assessment method for construction project safety, characterized by: The method comprises the following steps: S1. Collect historical status data of construction projects and annotate each sample manually; S2, perform anomaly analysis on the sample, calculate the sample LOF value, and set the anomaly threshold based on the labeled category; S3, monitor construction project characteristic data in real time and make abnormal judgments to assess risks; S4. Combine historical data and real-time monitoring data to predict future data, and perform anomaly analysis on future data to assess risks.

7. The construction engineering safety intelligent risk assessment method based on artificial intelligence according to claim 6 is characterized by: In step S1, the engineering history status feature data is collected: [F1, F2, …, F n ]; Where n represents the number of features, F n represents the nth feature; The characteristic data refers to characteristic parameters reflecting the status of the construction project in different aspects; Each sample is annotated manually and the annotation categories are [A, B1, B2, …, B m ]; Among them, A indicates that the sample is a normal sample; B1~B m Indicates that the sample is an abnormal sample, m is a positive integer, indicating the number of abnormal types, B m Represents the mth abnormal sample category; The labeling classification refers to the qualitative classification of the engineering history status samples; A database is formed, which includes historical status data of construction projects and their corresponding annotation categories.

8. The construction engineering safety intelligent risk assessment method based on artificial intelligence according to claim 7 is characterized by: In step S2, the feature data in the database are standardized and digitized; the LOF value of each sample is calculated: for each sample p, the distance between it and other samples is calculated using the Euclidean distance; d(p,o) represents the distance between sample p and sample o, ; Select the nearest k neighbor samples, marked as N k (p); Where k is a positive integer; N k (p) represents the set of k samples closest to sample p; i is a positive integer, 1≤i≤n, F i p represents the i-th feature of sample p; F i o represents the i-th feature of sample o; Calculate the reachable density ρk(p) of sample p: ; Among them, d(q,N k (q)) is the distance from sample q to its neighbor samples; Calculate the local outlier factor: ; Among them, |N k (p)| represents the number of samples in the set; Calculate the LOF value of each sample and mark it as B1~B m The samples are integrated as abnormal samples, and the samples marked as A are normal samples; the average value of the LOF values ​​of all normal samples is marked as U, and the average value of the LOF values ​​of all abnormal samples is marked as V; Divide U and V into equal parts by I, and take the threshold value [U,U+(VU) / I,U+2(VU) / I,...,U+(VU)(I-1) / I,V]; All samples are compared with different thresholds according to their LOF values, and samples with LOF values ​​greater than the threshold are marked as abnormal, otherwise they are marked as normal; Calculate the accuracy corresponding to different thresholds: E=(TP+TN) / (TP+TN+FP+FN); Among them, TP represents true positive examples, which are actually abnormal and marked as abnormal samples; TN represents true negative examples, which are actually normal and marked as normal samples; FP represents false positive examples, which are actually normal but marked as abnormal samples; FN represents false negative examples, which are actually abnormal but marked as normal samples. Get different accuracy rates corresponding to different thresholds, build a regression model with the threshold as input and the accuracy as output to obtain the polynomial regression equation: E(T)=a0+a1T+a2T 2 +…+a r T r ; Where T is the threshold; E(T) is the accuracy corresponding to different thresholds T; r is the polynomial order; a0, a1, …, a r For the parameters to be estimated, the least squares method is used to estimate the error sum between the actual accuracy and the model prediction accuracy to be minimized; Derivate the polynomial equation and set the derivative equal to 0, solve the equation to obtain possible extreme points, substitute the extreme points into the original polynomial regression equation, compare the accuracies corresponding to different extreme points, and select the extreme point corresponding to the highest accuracy as the abnormal threshold z.

9. The construction engineering safety intelligent risk assessment method based on artificial intelligence according to claim 8 is characterized by: In step S3, the construction project characteristics are monitored in real time, and the acquired new sample data is put into the database for updating; the LOF value of the new sample data is calculated and compared with the abnormal threshold z; when the LOF value of the new sample data is less than z, the new sample is judged to be a normal sample, the sample is marked as a normal sample, a risk assessment is made and output to the user, and the upload database is updated; when the LOF value of the new sample data is greater than or equal to z, the new sample is judged to be an abnormal sample, the Euclidean distance between the new sample and all samples in the database is calculated, the abnormal type corresponding to the sample with the closest distance is selected as the abnormal type of the new sample, a risk assessment is made and output to the user, and the upload database is updated.

10. The construction engineering safety intelligent risk assessment method based on artificial intelligence according to claim 9 is characterized by: In step S4, a time series model is constructed based on historical data to predict the characteristic data of construction projects at future moments: the historical data and the characteristic data of real-time monitoring are used as input to construct an LSTM model, and the predicted characteristic values ​​at future moments are output; The LOF value of the sample at the future moment is calculated based on the predicted data and compared with the abnormal threshold z; when the LOF value of the new sample data is less than z, the new sample is judged to be a normal sample, a risk assessment is made and output to the user; when the LOF value of the new sample data is greater than or equal to z, the new sample is judged to be an abnormal sample, the Euclidean distance between the new sample and all samples in the database is calculated, the abnormal type corresponding to the sample with the closest distance is selected as the abnormal type of the new sample, a risk assessment is made and output to the user.

Images