CN-122022479-A - Intelligent management system and method for building safety based on artificial intelligence

CN122022479ACN 122022479 ACN122022479 ACN 122022479ACN-122022479-A

Abstract

The invention discloses a construction safety intelligent management system and a construction safety intelligent management method based on artificial intelligence, which relate to the technical field of construction safety, wherein an intelligent terminal is used for collecting body motion data, space positioning data and environment interaction data of operators to construct a multi-source data set, the system is used for generating a multi-scale feature set through extraction of micro-motion features, macro-motion features and environment interaction features, further calculating unbalanced falling risks, high falling risks and collision injury risks of the operators, comprehensively evaluating instantaneous risk values, combining a risk attenuation function and an environment static risk value, calculating total risk value, carrying out grading early warning according to the total risk value and the change trend of the total risk value, and informing the operators through the intelligent terminal, and aims at monitoring the safety risk of the operators in real time, timely sending early warning, effectively reducing the safety accident risks in construction operation and improving the safety of the operators.

Inventors

MENG XIANGLONG
LU BO
ZHANG JINTAO
LI WENTING
ZHANG FEIFEI

Assignees

上海欣发建设工程监理有限公司

Dates

Publication Date: 20260512
Application Date: 20260131

Claims (9)

1. An artificial intelligence based construction safety intelligent management method, which is characterized by comprising the following steps: Firstly, acquiring body motion data, space positioning data and environment interaction data related to an operator based on an intelligent terminal worn by the operator, and performing air alignment to construct a multi-source data set; Step two, based on the constructed multi-source data set, micro-motion feature extraction, macro-behavior feature extraction and environment interaction feature extraction are synchronously executed, and a multi-scale feature set is generated; Step three, calculating an unbalance falling risk component, a high falling risk component and a collision injury risk component which are currently associated with an operator based on a multi-scale feature set, and comprehensively evaluating an instantaneous comprehensive risk value of the operator; Step four, constructing a risk attenuation function bound with an operator, evaluating an interactive risk value of the operator in a three-dimensional space, and calculating a total risk value of the operator by combining an environment static risk value and an instantaneous comprehensive risk value; and fifthly, executing hierarchical early warning operation based on the total risk value and the change trend thereof, and informing the operator of early warning notification through the intelligent terminal.
2. The method according to claim 1, wherein in the first step, the body motion data is collected based on a nine-axis inertial measurement unit integrated by the intelligent terminal, including a three-axis acceleration ACC (t), a three-axis angular velocity GYRO (t), and a three-axis magnetic field strength MAG (t), wherein (t) represents a relationship with time; the space positioning data are collected based on a UWB ultra-wideband positioning unit and an inertial navigation module integrated by the intelligent terminal, and the space positioning data comprise three-dimensional coordinates [ X (t), Y (t), Z (t) ]; The environment interaction data is based on millimeter wave radar collection integrated by the intelligent terminal, and comprises a relative distance D_obj (t), a relative speed V_obj (t) and an azimuth angle theta_obj (t) of an obstacle in a 5-meter distance with an operator as a center; The body motion data, the space positioning data and the environment interaction data are collected based on a preset uniform sampling time interval.
3. The method according to claim 2, wherein in the first step, the specific manner of constructing the multi-source data set is: time alignment is carried out based on the time bound by the body motion data, the space positioning data and the environment interaction data; converting the body motion data, the space positioning data and the environment interaction data which are subjected to time alignment into a pre-constructed three-dimensional panoramic model of the construction site, and mapping the three-dimensional panoramic model into a body coordinate system taking the mass center of an operator as an origin; And extracting body motion data, space positioning data and environment interaction data in the body coordinate system, and combining the body motion data, the space positioning data and the environment interaction data into a multi-source data set.
4. A method according to claim 3, wherein in the second step, the specific manner of generating the multiscale feature set is: s1, extracting micro-motion characteristics: S11, calculating a standard deviation sigma_A and kurt_A of an acceleration vector sum A_sum=sqrt (ACC_x2+ACC_y2+ACC_z2) in a time window T1, wherein sqrt () is a square root function, and the time window T1 is a time period preset by an operator; S12, resolving a real-time pitch angle alpha_pitch (t) and a roll angle alpha_roll (t) of the trunk of the operator relative to the ground based on a quaternion method; S13, calculating an attitude stability index s_ posture =1/(1+σ_α_pitch+σ_α_roll), wherein σ_α is a standard deviation of the angle within 1 second, σ_α_pitch represents a standard deviation of the real-time pitch angle within 1 second, and σ_α_roll represents a standard deviation of the roll angle within 1 second; s2, extracting macroscopic behavior characteristics: S21, determining a moving track of the operator in a time window T1 based on three-dimensional coordinates [ X (T), Y (T) and Z (T) ] of the operator, and calculating a curvature change rate of the moving track, wherein C_curve (T) = |delta theta_head/delta S|, delta theta_head is a course angle change amount, and delta S is a moving arc length; S22, calculating motion entropy H_motion= -sigma (p_i. Log2 (p_i)) of an operator in a three-dimensional space, wherein p_i is the probability of each discrete direction interval of a velocity vector on a unit sphere; S3, extracting environment interaction characteristics: s31, when the millimeter wave radar detects d_obj (t) < a preset safety distance d_s, adopting: γ_approach=(D_s-D_obj(t))/D_s+(V_obj(t)/V_s); Calculating a proximity risk coefficient gamma_app, wherein V_s is a preset reference speed; Extracting micro-motion characteristics, including an acceleration vector sum A_sum, a standard deviation sigma_A, kurt_A, a real-time pitch angle alpha_pitch (t), a roll angle alpha_roll (t) and a posture stability index S_ posture; Extracting macroscopic behavior characteristics including curvature change rate C_curve (t) and motion entropy H_motion; Extracting environment interaction characteristics including a proximity risk coefficient gamma_app; summarizing micro-action features, macro-action features and environmental interaction features to generate a multi-scale feature set F (t).
5. The method according to claim 4, wherein in the third step, the specific way to comprehensively evaluate the instantaneous comprehensive risk value of the operator is as follows: s51, calculating an unbalanced fall risk component r_stumble (t): R_stumble(t)=w1*(1-S_posture(t))+w2*(σ_A(t)/σ_A_max)+w3*(|Kurt_A(t)-3|/Kurt_norm); wherein w1, w2, w3 are preset weight coefficients, w1+w2+w3=1, σ_a_max is a preset acceleration standard deviation threshold, and kurt_norm is a kurtosis reference value; S52, calculating a high-falling risk component R_fall (t): R_fall(t)=f_height(Z(t))*[λ1*g_dist(D_edge(t))+λ2*(H_motion(t)/H_max)+λ3*(C_curve(t)/C_max)]; Wherein f_height (Z (t)) is a height penalty function, f_height (Z (t))=1+log (1+z (t)/z_ref), z_ref is a preset reference height; d_edge (t) is the projection horizontal distance from the operator to the nearest dangerous edge calculated in real time based on the three-dimensional coordinates [ X (t), Y (t), Z (t) ] of the operator and the pre-constructed three-dimensional panoramic model of the construction site; g_dist (d_edge (t)) is an edge distance risk function, g_dist (d_edge (t))=exp (-d_edge (t)/d_safe), d_safe is a preset safe distance threshold; λ1, λ2, λ3 are preset weight coefficients, λ1+λ2+λ3=1, and h_max and c_max are preset motion entropy threshold and curvature change rate threshold respectively; S53, calculating a collision injury risk component R_collision (t): R_collision(t)=max(γ_approach(t))*[1+η*(||V_worker(t)||/V_worker_max)]; wherein max (gamma_app reach (t)) is the maximum value in the proximity risk coefficients of all obstacles calculated from millimeter wave radar data at time t, V_worker (t) is the motion speed of an operator himself calculated based on space positioning data at time t, V_worker_max is a preset operator safety speed upper limit, and eta is a preset adjustment coefficient; S54, calculating an instantaneous integrated risk value r_instance (t) based on the unbalanced fall risk component r_stable (t), the high fall risk component r_fall (t), and the collision injury risk component r_collision (t): R_instant(t)=β1*R_stumble(t)+β2*R_fall(t)+β3*R_collision(t); wherein β1, β2, β3 are preset comprehensive weight coefficients, and β1+β2+β3=1.
6. The method according to claim 5, wherein in the fourth step, a risk attenuation function bound to the operator is constructed, and the specific way of assessing the interactive risk value of the operator in the three-dimensional space is as follows: Establishing a personal historical risk position record list L_risk for an operator, wherein any record contains three-dimensional coordinates of a risk position [ X_risk, Y_risk, Z_risk ]; the three-dimensional coordinates of the risk positions correspond to an initial high-falling risk component R_fall_init and a record generation timestamp T_risk; When the instantaneous high-falling risk component R_fall (t) exceeds a preset activation threshold R_fall_th, taking the space positioning data [ X (t), Y (t), Z (t) ] of the current moment of an operator as a three-dimensional coordinate of a risk position, taking R_fall (t) as an initial high-falling risk component R_fall_init, forming a new record with a current time stamp, and adding the new record into an L_risk; constructing the risk attenuation function phi (delta t, R_init) bound with operators, and calculating a risk value of any risk position three-dimensional coordinate attenuated along with time: R_decayed=Φ(Δt,R_init)=R_init*exp(-k*Δt) Wherein Δt is the time difference between the current evaluation time and the risk record timestamp t_risk, r_init is the initial high-fall risk component r_fall_init, k is a preset attenuation coefficient, exp () is an exponential function; Calculating distances d_i=sqrt ((X (t) -x_risk) 2+ (Y (t) -y_risk) 2+ (Z (t) -z_risk) 2) from all historical risk positions in l_risk based on spatial location data [ X (t), Y (t), Z (t) ]atthe current moment of the operator; traversing L_risk, and calculating a risk value R_ decayed after the attenuation of each record with the distance d_i from the current position being smaller than the preset interaction radius R_int based on a risk attenuation function phi (delta t, R_init); the maximum value of all risk values R_ decayed is defined as an interaction risk value R_ interaction (t) at the current moment of the operator: If there is no record satisfying d_i < r_int, let r_ interaction (t) =0.
7. The method according to claim 6, wherein in the fourth step, the total risk value of the operator is calculated by combining the environmental static risk value and the instantaneous integrated risk value in the following specific manner: s71, inquiring an environment static risk map pre-constructed by an operator based on three-dimensional coordinates [ X (t), Y (t), Z (t) ] of the operator at the current moment, and acquiring an environment static risk value R_ environment (t) corresponding to the three-dimensional coordinates; The method comprises the steps that an environment static risk map is built based on a three-dimensional panoramic model of a construction site, and risk grade values related to static environment risk sources are predefined for different space areas in the model; S72, calculating a total risk value R_total (t) of the operator by combining the instantaneous comprehensive risk value R_instance (t), the interactive risk value R_ interaction (t) and the environmental static risk value R_ environment (t): R_total(t)=γ1*R_instant(t)+γ2*R_interaction(t)+γ3*R_environment(t) wherein, gamma 1, gamma 2 and gamma 3 are preset fusion weight coefficients, and γ1+γ2+γ3=1.
8. The method according to claim 7, wherein in the fifth step, based on the total risk value and the trend thereof, the specific manner of performing the hierarchical early warning operation is: S81, extracting a risk early warning threshold value, wherein the risk early warning threshold value comprises a first threshold value Th_low, a second threshold value Th_medium and a third threshold value Th_high, th_low < Th_medium < Th_high <1, and [0,1] represents a risk early warning threshold value interval; S82, calculating the change rate V_risk (T) of the total risk value R_total (T) in a preset time window T2, wherein the time length of T2 is larger than T1; S83, determining an early warning level according to a threshold interval to which the total risk value R_total (t) belongs and a change rate V_risk (t), wherein the early warning level comprises an attention level, a warning level and a critical level: if R_total (t) is not less than Th_high or R_total (t) is not less than Th_medium and V_risk (t) is greater than a preset rise rate threshold V_up_th, judging as critical-level early warning; if Th_medium is less than or equal to R_total (t) < Th_high and V_risk (t) is less than or equal to V_up_th, or Th_low is less than or equal to R_total (t) < Th_medium and V_risk (t) > V_up_th, judging as warning level early warning; If Th_low is less than or equal to R_total (t) < Th_medium and V_risk (t) is less than or equal to V_up_th, judging as attention level early warning; If R_total (t) < Th_low, judging that the early warning state is not available; S84, when the attention level early warning is judged, a low-frequency intermittent vibration and a yellow visual prompt are sent out through the intelligent terminal; When the warning level early warning is judged, the intelligent terminal sends out continuous vibration, red visual prompt and level-one beeping sound; When the emergency early warning is judged, the intelligent terminal sends out high-intensity continuous vibration, flashing red visual prompts and secondary sharp beeping sounds, and alarm information containing the three-dimensional coordinates of the operators and the early warning level is synchronously sent to the operators.
9. An artificial intelligence based construction safety intelligent management system, the system comprising: The multi-source space-time alignment acquisition module is used for acquiring body motion data, space positioning data and environment interaction data related to an operator based on an intelligent terminal worn by the operator, and performing space alignment to construct a multi-source data set; the multi-scale feature fusion extraction module is used for synchronously executing micro-motion feature extraction, macro-behavior feature extraction and environment interaction feature extraction based on the constructed multi-source data set to generate a multi-scale feature set; The instantaneous risk component calculation module is used for calculating an unbalance falling risk component, a high falling risk component and a collision injury risk component which are currently associated with the operator based on the multi-scale feature set, and comprehensively evaluating the instantaneous comprehensive risk value of the operator; The space-time interaction risk integrating module is used for constructing a risk attenuation function bound with an operator, evaluating the interaction risk value of the operator in a three-dimensional space, and calculating the total risk value of the operator by combining the environment static risk value and the instantaneous comprehensive risk value; and the grading early warning pushing module is used for executing grading early warning operation based on the total risk value and the change trend thereof and notifying an operator of early warning notification through the intelligent terminal.

Description

Intelligent management system and method for building safety based on artificial intelligence Technical Field The invention belongs to the technical field of construction safety, and particularly relates to an artificial intelligence-based construction safety intelligent management system and method. Background The construction site environment is complex and changeable, and operators face a plurality of safety risks, such as unbalanced falling, high falling, collision injury and the like. The existing construction safety technical scheme generally has the obvious defects and disadvantages of low integration level, single perception dimension, static risk assessment, rough early warning mechanism and the like, the existing system depends on a single data source, such as whether personnel enter a dangerous area or not by using positioning data or monitoring falling by using an inertial sensor, the perception mode of the splitting is difficult to comprehensively capture complex and dynamic risk coupling states of an operation site, composite risks formed by personnel fine posture instability and abnormal behavior modes and environment dynamic obstacle interaction cannot be effectively identified, the risk assessment model of the existing technology is usually static and lagged, judgment is usually carried out on the basis of preset fixed rules or simple thresholds, such as alarming is carried out only when personnel pass through an electronic fence, quantitative modeling and dynamic calculation capability of potential influences on the current state of personnel are lacked due to continuous change of the personnel body state, space-time law of a behavior track and historical risk position, the early prediction and progressive evolution identification of risks are insufficient, early warning and intervention windows are often lost when the risks are obviously raised or accidents occur, in addition, the existing early warning mode and the existing warning mode is single and the system is more important, and the danger is easily ignored due to the fact that the danger is prone to serious and serious fatigue is caused by the fact that the personnel are prone to change. In order to solve the problems, the invention provides an artificial intelligence-based construction safety intelligent management system and method. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a construction safety intelligent management system and method based on artificial intelligence, which solve the problems of incomplete dynamic composite risk perception, evaluation lag and extensive early warning existing in the construction safety technical field in the prior art. The aim of the invention can be achieved by the following technical scheme: An artificial intelligence based construction safety intelligent management method, the method comprising: Firstly, acquiring body motion data, space positioning data and environment interaction data related to an operator based on an intelligent terminal worn by the operator, and performing air alignment to construct a multi-source data set; Step two, based on the constructed multi-source data set, micro-motion feature extraction, macro-behavior feature extraction and environment interaction feature extraction are synchronously executed, and a multi-scale feature set is generated; Step three, calculating an unbalance falling risk component, a high falling risk component and a collision injury risk component which are currently associated with an operator based on a multi-scale feature set, and comprehensively evaluating an instantaneous comprehensive risk value of the operator; Step four, constructing a risk attenuation function bound with an operator, evaluating an interactive risk value of the operator in a three-dimensional space, and calculating a total risk value of the operator by combining an environment static risk value and an instantaneous comprehensive risk value; and fifthly, executing hierarchical early warning operation based on the total risk value and the change trend thereof, and informing the operator of early warning notification through the intelligent terminal. In the first step, body motion data is collected based on a nine-axis inertial measurement unit integrated by an intelligent terminal, and the nine-axis inertial measurement unit comprises a three-axis acceleration ACC (t), a three-axis angular velocity GYRO (t) and a three-axis magnetic field strength MAG (t), wherein (t) represents a relation with time; the space positioning data are collected based on a UWB ultra-wideband positioning unit and an inertial navigation module integrated by the intelligent terminal, and the space positioning data comprise three-dimensional coordinates [ X (t), Y (t), Z (t) ]; The environment interaction data is based on millimeter wave radar collection integrated by the intelligent terminal, and comprises a relative distance D_obj (t), a relative speed