Visualized supervision system and method for engineering quality applied to municipal construction

Abstract

The present invention discloses a system and method for visualizing and supervising engineering quality applied to municipal construction, and relates to the field of intelligent construction technology. The present invention comprises the following steps: constructing a construction parameter set and an effect parameter set; obtaining historical bridge construction sample data, processing geographic features to form a spatial feature vector; completing parameter encoding and normalization; calculating the effect parameter decay rate, building a multi-output regression model, combining construction parameters and spatial features as input, and training the model with a WMSE loss function; constructing an optimization function with the decay rate as the target, combining constraints, and solving the optimal parameter combination with an improved genetic algorithm; building a visualization platform, building a digital twin, predicting effect parameters based on sensors and LSTM models, comparing predictions from different models, and issuing an early warning when the threshold is exceeded. The present invention can effectively improve the situation in which the existing technology lacks dynamic adaptation to the environment and insufficient utilization of the parameter decay process.

Description

Technical Field

[0001] The present invention relates to the field of intelligent construction technology, and in particular to a system and method for visually supervising engineering quality applied to municipal construction. Background Art

[0002] In municipal bridge construction, the scientific configuration of construction parameters and dynamic quality control are key to ensuring project performance throughout its lifecycle. Traditional technical systems typically construct a set of construction parameters (such as material strength, structural dimensions, and construction techniques) based on empirical rules, and use single-variable regression models to analyze the relationship between these parameters and performance indicators (such as bearing capacity and durability). With the advancement of machine learning technology, multi-output regression models are increasingly being applied to multi-parameter coupled impact analysis, enabling simultaneous prediction of the evolution of multiple performance indicators.

[0003] Existing traditional approaches to municipal engineering quality supervision often rely on static parameter configuration, ignoring the diverse demands of geographical characteristics on construction parameters. For example, the same concrete mix and curing process are used in high-altitude areas as in plains, without dynamically adjusting parameters to address environmental factors such as low pressure on the plateau and large temperature swings between day and night. This can lead to quality risks such as rapid strength degradation in bridges early in their service life. Furthermore, these approaches rely on a single regression model to predict performance parameters, lacking a phased analysis of parameter degradation over the service life. Summary of the Invention

[0004] The purpose of the present invention is to provide a system and method for visually monitoring engineering quality applied to municipal construction, so as to solve the problems raised in the prior art.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a method for visually supervising engineering quality applied to municipal construction, the method comprising the following steps:

[0006] Step 1: Construct construction parameter sets and effect parameter sets; obtain historical bridge construction sample data, process geographic features to form spatial feature vectors; complete parameter encoding and normalization;

[0007] Step 2: Calculate the attenuation rate of the effect parameters, build a multi-output regression model, concatenate the construction parameters and spatial features as input, use ReLU activation to extract the association, and train the model with the WMSE loss function;

[0008] Step 3: Construct an optimization function with the attenuation rate as the target, and use the improved genetic algorithm to solve the optimal parameter combination in combination with the constraints;

[0009] Step 4: Build a visualization platform, construct a digital twin based on BIM, predict effect parameters based on sensors and LSTM models, compare the prediction values ​​of different models, and issue an early warning when the threshold is exceeded.

[0010] In step 1, a construction parameter set is constructed to cover all elements of bridge construction, which is expressed as: [C1, C2, ..., C n ]; among them, C1, C2, …, C n Represent the 1st, 2nd,…, nth construction parameters respectively;

[0011] Construction parameters include material parameters, such as main material properties and auxiliary material ratios, structural parameters, such as geometric dimensions and mechanical configuration, and process parameters, such as construction technology and maintenance conditions.

[0012] Construct the effect parameter set, expressed as: [E1,E2,…,E m ]; where E1, E2, …, E m Represent the 1st, 2nd,…,mth effect parameters respectively;

[0013] Obtain historical bridge construction samples and corresponding parameter data, effect parameter data corresponding to different stages, and geographical features;

[0014] Geographical features include latitude and longitude (φ, λ) and altitude h;

[0015] Parameters are coded using a preset unified format (e.g., C30 concrete is coded as C-30, and cast-in-place process is coded as TC) to facilitate data management and analysis.

[0016] For categorical parameters, one-hot encoding is performed, and for numerical parameters, normalization is performed;

[0017] Convert the longitude and latitude (φ, λ) into UTM projection coordinates (x, y) to form a spatial feature vector S = (x, y, h).

[0018] In step 2, the bridge service life is divided into N stages, and the time for obtaining effect parameter data is expressed as: [T1, T2, …, T N+1 ];

[0019] For the effect parameter E i , define the attenuation index of each stage:

[0020] For the nth stage, calculate the decay rate D n,i =(E n,i -E (n+1),i ) / E n,i ;

[0021] Where i∈{1,2,…,m}; n∈{1,2,…,N}; D n,i Indicates the effect parameter E of the nth stage i The decay rate of E n,i Indicates the effect parameter E at the nth moment i Value; E (n+1),iIndicates the effect parameter E at the (n+1)th moment i value;

[0022] Calculate the comprehensive decay rate: For the nth stage, the decay rate weight is the ratio of stage length to duration, w n =(T n+1 -T n ) / (T N+1 -T1);

[0023] Comprehensive attenuation rate D i =Σ n=1 N D n,i w n ;

[0024] A multi-output regression model is constructed to achieve nonlinear mapping from construction parameters, spatial features to effect parameters. The model architecture is divided into three layers:

[0025] Integrate the input layer features, concatenate the construction parameters and the spatial feature vector to form the input layer feature X with a dimension of (n+3);

[0026] Use the fully connected layer to extract the nonlinear relationship between parameters, and the activation function is ReLU;

[0027] Set the spatial feature weight matrix W s , calculate the spatial impact coefficient; focus on geographical features such as high altitudes and extreme climate zones;

[0028] For m effect parameters, construct independent output heads; each output head corresponds to the prediction of one effect parameter;

[0029] The historical bridge sample is represented as (X k ,E k ), where X k is the input layer feature of the kth sample; E k =[E k1 ,E k2 ,…,E km ] is the corresponding effect parameter value;

[0030] Divide the training set and test set, and use K-fold cross validation to evaluate the generalization ability

[0031] The loss function uses weighted multi-output mean square error (WMSE) to highlight the importance of core effect parameters: Loss=Σ i=1 m β i (1 / N)Σ k=1 N (E ki '-E ki )2 ;

[0032] Among them, β i Indicates the effect parameter E i Weight; N represents the number of batch samples; E ki ' represents the kth sample effect parameter E i Predicted value; E ki Indicates the kth sample effect parameter E i Actual value.

[0033] In step 3, under the given spatial features [x*, y*, h*], the optimization function is constructed: minF=Σ i=1 m γ i D i ;

[0034] Among them, γ i Indicates the effect parameter E i corresponding weights;

[0035] Set constraints: E i ≥E i,min Among them, E i,min Indicates the effect parameter E set i The minimum value to be achieved; p j ∈[p j,min ,p j,max ]; where j∈{1,2,…,n}; p i represents the jth construction parameter; p j,min ,p j,max Represent the construction parameters p j The preset minimum and maximum values ​​of

[0036] Based on the optimization function and constraints, an improved genetic algorithm (GA) is used to solve the optimal combination of construction parameter values:

[0037] The inverse of the objective function F is used as the fitness, so that the individual with a smaller decay rate has a higher fitness, and GA optimization is supported by converting the minimum problem to the maximum problem;

[0038] Adding penalty items to individuals that do not meet the normative constraints;

[0039] Use simulated binary crossover (SBX), set the crossover probability, set a larger crossover weight for highly sensitive parameters, and enhance the search efficiency of key parameters;

[0040] Use polynomial mutation with a mutation probability of 1 / m, automatically reducing the mutation amplitude for parameters close to the constraint boundary to avoid parameter out of bounds;

[0041] The preset A1% optimal solutions before each generation are directly transferred to the next generation. The "survival of the fittest" mechanism is used to prevent the loss of excellent solutions and maintain the stability of population evolution.

[0042] Set termination conditions: the mean change rate of the objective function for a preset number of B1 generations is less than the preset threshold A2%, or the number of iterations reaches a preset number of B2 generations;

[0043] The optimal combination of construction parameter values ​​is obtained, and the corresponding effect parameter prediction values ​​are obtained based on the multi-output regression model.

[0044] In step 4, the bridge is constructed based on the optimal combination of construction parameter values, and a digital twin of the bridge is built based on BIM technology to achieve real-time synchronous display of the physical bridge and the virtual model;

[0045] Sensors are deployed to obtain real-time data on bridge construction effect parameters, and the values ​​of each effect parameter are predicted based on the LSTM time series model. Data is continuously collected to update the time series model to obtain time-series-based effect parameter predictions.

[0046] Compare the corresponding effect parameter prediction values ​​based on the multi-output regression model with the effect parameter prediction values ​​based on time series; when one or more errors between the two are greater than the set threshold, an early warning is issued and the specific effect parameters are displayed for the administrator's reference.

[0047] A visual supervision system for engineering quality applied to municipal construction, which includes a data preprocessing module, a model building module, a parameter optimization module, and a monitoring and early warning module;

[0048] The data preprocessing module is used to construct a construction parameter set and an effect parameter set; obtain historical bridge construction sample data, process geographic features to form a spatial feature vector; and complete parameter encoding and normalization. The model construction module is used to calculate the effect parameter attenuation rate, build a multi-output regression model, combine the construction parameters and spatial features as input, and train the model using the WMSE loss function. The parameter optimization module is used to construct an optimization function with the attenuation rate as the target, and use an improved genetic algorithm to solve the optimal parameter combination in combination with the constraints.

[0049] The monitoring and early warning module is used to build a visualization platform, construct a digital twin based on BIM, predict effect parameters based on sensors and LSTM models, compare prediction values ​​of different models, and issue an early warning when the threshold is exceeded.

[0050] The data preprocessing module includes a parameter encoding unit and a space conversion unit;

[0051] The parameter encoding unit is used to encode the construction parameter set and the effect parameter set; the space conversion unit is used to convert the geographic features into UTM coordinates.

[0052] The model building module includes a decay calculation unit, a model training unit and a verification and evaluation unit;

[0053] The attenuation calculation unit is used to calculate the stage attenuation rate and the weighted comprehensive attenuation rate; the model training unit is used to construct a multi-output regression model; the verification and evaluation unit is used to evaluate the model based on K-fold cross-validation and the test set verifies the generalization ability.

[0054] The parameter optimization module includes a constraint construction unit and a genetic algorithm unit;

[0055] The constraint construction unit is used to set parameter boundary constraints and effect parameter threshold constraints; the genetic algorithm unit is used to use improved GA to solve the optimal parameter combination.

[0056] The monitoring and early warning module includes a BIM modeling unit, a time series prediction unit, a model comparison unit and an early warning trigger unit;

[0057] The BIM modeling unit is used to construct a digital twin of the bridge based on the optimal parameter combination to achieve visualization; the time series prediction unit is used to obtain real-time data through sensors, and the LSTM model predicts the time series changes of effect parameters; the model comparison unit is used to compare the predicted value of the multi-output regression model with the predicted value of the LSTM; the early warning trigger unit is used to trigger an early warning when the error in the comparison is greater than a preset threshold, and locate abnormal parameters.

[0058] Compared with the existing technology, the beneficial effects of the present invention are as follows: the present invention constructs a construction parameter set and an effect parameter set including materials, structures, and processes, converts geographical features (latitude and longitude, altitude) into spatial feature vectors of UTM projection coordinates, and solves the problem of insufficient quantification of regional difference factors in traditional methods; the present invention calculates the attenuation rate of effect parameters by stage, and calculates the comprehensive attenuation rate with the stage duration as the weight. Compared with the traditional method that only focuses on single time point indicators, the present invention can more dynamically reflect the degradation process of bridge performance over time; the present invention takes the comprehensive attenuation rate as the optimization target, combines it with construction parameter constraints, and solves the optimal parameter combination under given geographical features (such as high altitude and severe cold areas); and can automatically adjust construction parameters for different regions. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A schematic diagram of the steps of the method for visually supervising engineering quality of municipal construction according to the present invention;

[0060] Figure 2 The figure is a flow chart of the application of the present invention to the visual supervision system of engineering quality in municipal construction. DETAILED DESCRIPTION

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0062] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution for a method for visually supervising engineering quality of municipal construction, which includes the following steps:

[0063] Step 1: Construct construction parameter sets and effect parameter sets; obtain historical bridge construction sample data, process geographic features to form spatial feature vectors; complete parameter encoding and normalization;

[0064] Step 2: Calculate the attenuation rate of the effect parameters, build a multi-output regression model, concatenate the construction parameters and spatial features as input, use ReLU activation to extract the association, and train the model with the WMSE loss function;

[0065] Step 3: Construct an optimization function with the attenuation rate as the target, and use the improved genetic algorithm to solve the optimal parameter combination in combination with the constraints;

[0066] Step 4: Build a visualization platform, construct a digital twin based on BIM, predict effect parameters based on sensors and LSTM models, compare the prediction values ​​of different models, and issue an early warning when the threshold is exceeded.

[0067] In step 1, a construction parameter set is constructed to cover all elements of bridge construction, which is expressed as: [C1, C2, ..., C n ]; among them, C1, C2, …, C n Represent the 1st, 2nd,…, nth construction parameters respectively;

[0068] Construction parameters include material parameters, such as main material properties and auxiliary material ratios, structural parameters, such as geometric dimensions and mechanical configuration, and process parameters, such as construction technology and maintenance conditions.

[0069] Construct the effect parameter set, expressed as: [E1,E2,…,E m ]; where E1, E2, …, E m Represent the 1st, 2nd,…,mth effect parameters respectively;

[0070] Obtain historical bridge construction samples and corresponding parameter data, effect parameter data corresponding to different stages, and geographical features;

[0071] Geographical features include latitude and longitude (φ, λ) and altitude h;

[0072] Parameters are coded using a preset unified format (e.g., C30 concrete is coded as C-30, and cast-in-place process is coded as TC) to facilitate data management and analysis.

[0073] For categorical parameters, one-hot encoding is performed, and for numerical parameters, normalization is performed;

[0074] Convert the longitude and latitude (φ, λ) into UTM projection coordinates (x, y) to form a spatial feature vector S = (x, y, h).

[0075] In step 2, the bridge service life is divided into N stages, and the time for obtaining effect parameter data is expressed as: [T1, T2, …, T N+1 ];

[0076] For the effect parameter E i , define the attenuation index of each stage:

[0077] For the nth stage, calculate the decay rate D n,i =(E n,i -E (n+1),i ) / E n,i ;

[0078] Where i∈{1,2,…,m}; n∈{1,2,…,N}; D n,i Indicates the effect parameter E of the nth stage i The decay rate of E n,i Indicates the effect parameter E at the nth moment i Value; E (n+1),i Indicates the effect parameter E at the (n+1)th moment i value;

[0079] Calculate the comprehensive decay rate: For the nth stage, the decay rate weight is the ratio of stage length to duration, w n =(T n+1 -T n ) / (T N+1 -T1);

[0080] Comprehensive attenuation rate D i =Σ n=1 N D n,i w n ;

[0081] A multi-output regression model is constructed to achieve nonlinear mapping from construction parameters, spatial features to effect parameters. The model architecture is divided into three layers:

[0082] Integrate the input layer features, concatenate the construction parameters and the spatial feature vector to form the input layer feature X with a dimension of (n+3);

[0083] Use the fully connected layer to extract the nonlinear relationship between parameters, and the activation function is ReLU;

[0084] Set the spatial feature weight matrix W s , calculate the spatial impact coefficient; focus on geographical features such as high altitudes and extreme climate zones;

[0085] For m effect parameters, construct independent output heads; each output head corresponds to the prediction of one effect parameter;

[0086] The historical bridge sample is represented as (X k ,E k ), where X k is the input layer feature of the kth sample; E k =[E k1 ,E k2 ,…,E km ] is the corresponding effect parameter value;

[0087] Divide the training set and test set, and use K-fold cross validation to evaluate the generalization ability

[0088] The loss function uses weighted multi-output mean square error (WMSE) to highlight the importance of core effect parameters: Loss=Σ i=1 m β i (1 / N)Σ k=1 N (E ki '-E ki ) 2 ;

[0089] Among them, β i Indicates the effect parameter E i Weight; N represents the number of batch samples; E ki ' represents the kth sample effect parameter E i Predicted value; E ki Indicates the kth sample effect parameter E i Actual value.

[0090] In step 3, under the given spatial features [x*, y*, h*], the optimization function is constructed: minF=Σ i=1 m γ i D i ;

[0091] Among them, γ i Indicates the effect parameter E i corresponding weights;

[0092] Set constraints: E i ≥E i,min Among them, Ei,min Indicates the effect parameter E set i The minimum value to be achieved; p j ∈[p j,min ,p j,max ]; where j∈{1,2,…,n}; p i represents the jth construction parameter; p j,min ,p j,max Represent the construction parameters p j The preset minimum and maximum values ​​of

[0093] Based on the optimization function and constraints, an improved genetic algorithm (GA) is used to solve the optimal combination of construction parameter values:

[0094] The inverse of the objective function F is used as the fitness, so that the individual with a smaller decay rate has a higher fitness, and GA optimization is supported by converting the minimum problem to the maximum problem;

[0095] Adding penalty items to individuals that do not meet the normative constraints;

[0096] Use simulated binary crossover (SBX), set the crossover probability, set a larger crossover weight for highly sensitive parameters, and enhance the search efficiency of key parameters;

[0097] Use polynomial mutation with a mutation probability of 1 / m, automatically reducing the mutation amplitude for parameters close to the constraint boundary to avoid parameter out of bounds;

[0098] The preset A1% optimal solutions before each generation are directly transferred to the next generation. The "survival of the fittest" mechanism is used to prevent the loss of excellent solutions and maintain the stability of population evolution.

[0099] Set termination conditions: the mean change rate of the objective function for a preset number of B1 generations is less than the preset threshold A2%, or the number of iterations reaches a preset number of B2 generations;

[0100] The optimal combination of construction parameter values ​​is obtained, and the corresponding effect parameter prediction values ​​are obtained based on the multi-output regression model.

[0101] In step 4, the bridge is constructed based on the optimal combination of construction parameter values, and a digital twin of the bridge is built based on BIM technology to achieve real-time synchronous display of the physical bridge and the virtual model;

[0102] Sensors are deployed to obtain real-time data on bridge construction effect parameters, and the values ​​of each effect parameter are predicted based on the LSTM time series model. Data is continuously collected to update the time series model to obtain time-series-based effect parameter predictions.

[0103] Compare the corresponding effect parameter prediction values ​​based on the multi-output regression model with the effect parameter prediction values ​​based on time series; when one or more errors between the two are greater than the set threshold, an early warning is issued and the specific effect parameters are displayed for the administrator's reference.

[0104] A visual supervision system for engineering quality applied to municipal construction, which includes a data preprocessing module, a model building module, a parameter optimization module, and a monitoring and early warning module;

[0105] The data preprocessing module is used to construct a construction parameter set and an effect parameter set; obtain historical bridge construction sample data, process geographic features to form a spatial feature vector; and complete parameter encoding and normalization. The model construction module is used to calculate the effect parameter attenuation rate, build a multi-output regression model, combine the construction parameters and spatial features as input, and train the model using the WMSE loss function. The parameter optimization module is used to construct an optimization function with the attenuation rate as the target, and use an improved genetic algorithm to solve the optimal parameter combination in combination with the constraints.

[0106] The monitoring and early warning module is used to build a visualization platform, construct a digital twin based on BIM, predict effect parameters based on sensors and LSTM models, compare prediction values ​​of different models, and issue an early warning when the threshold is exceeded.

[0107] The data preprocessing module includes a parameter encoding unit and a space conversion unit;

[0108] The parameter encoding unit is used to encode the construction parameter set and the effect parameter set; the space conversion unit is used to convert the geographic features into UTM coordinates.

[0109] The model building module includes a decay calculation unit, a model training unit and a verification and evaluation unit;

[0110] The attenuation calculation unit is used to calculate the stage attenuation rate and the weighted comprehensive attenuation rate; the model training unit is used to construct a multi-output regression model; the verification and evaluation unit is used to evaluate the model based on K-fold cross-validation and the test set verifies the generalization ability.

[0111] The parameter optimization module includes a constraint construction unit and a genetic algorithm unit;

[0112] The constraint construction unit is used to set parameter boundary constraints and effect parameter threshold constraints; the genetic algorithm unit is used to use improved GA to solve the optimal parameter combination.

[0113] The monitoring and early warning module includes a BIM modeling unit, a time series prediction unit, a model comparison unit and an early warning trigger unit;

[0114] The BIM modeling unit is used to construct a digital twin of the bridge based on the optimal parameter combination to achieve visualization; the time series prediction unit is used to obtain real-time data through sensors, and the LSTM model predicts the time series changes of effect parameters; the model comparison unit is used to compare the predicted value of the multi-output regression model with the predicted value of the LSTM; the early warning trigger unit is used to trigger an early warning when the error in the comparison is greater than a preset threshold, and locate abnormal parameters.

[0115] In this example, a city plans to construct a prestressed concrete continuous box girder bridge across a river. The bridge is 850 meters long, with a main span of 150 meters and a six-lane, bidirectional design. To ensure the quality of the bridge's construction, a visual monitoring method based on parametric modeling and digital twin technology is employed to precisely control the entire process, from material selection to lifecycle performance prediction.

[0116] Step 1: Construct parameter set and spatial feature processing; Construction parameter set construction: Cover all elements of bridge construction, and define the construction parameter set as: C = [C1, C2, …, C15];

[0117] Specific parameters include: Material parameters (C1-C4): C1 is the main material properties (coding rule: concrete strength grade uses "C-XX", such as C30 concrete is coded as "C-30"; steel type uses "ST-XXX", such as Q345qD steel is coded as "ST-Q345qD"), including the main beam concrete strength grade (C-30) and prestressed steel strand model (ST-15.2); C2 is the auxiliary material ratio (coding rule: admixture dosage uses "AD-%d", such as a water reducer dosage of 1.5% is coded as "AD-1.5"), including the fly ash dosage (AD-20%) and water reducer dosage (AD-1.2%).

[0118] Structural parameters (C5-C9): C5 is the geometric dimensions (coding rule: length unit m uses "L-%d", height unit m uses "H-%d"), including the main beam span (L-150), beam height (H-2.8), and web thickness (H-0.6); C6 is the mechanical configuration (coding rule: prestressed tension control stress uses "σ-%dMPa"), including the prestressed tension control stress (σ-1302MPa) and ordinary steel reinforcement ratio (R-2.5%);

[0119] Process parameters (C10-C15): C10 represents the construction process (coding rule: cast-in-place process is "TC", prefabricated assembly is "TP"), including the main beam construction process (TC) and pile foundation construction process (TD, bored pile); C11 represents the curing conditions (coding rule: standard curing is "CS", steam curing is "C-Steam"), including the concrete curing method (C-Steam, steam curing temperature 45°C) and curing time (D-14, 14 days);

[0120] Effect parameter set construction: define the effect parameter set as: E=[E1,E2,E3,E4];

[0121] Specifically, E1 is the compressive strength of concrete (MPa), E2 is the steel corrosion rate (μm / year), E3 is the thickness deviation of the bridge deck pavement (mm), and E4 is the dynamic response frequency of the structure (Hz);

[0122] Obtain the latitude and longitude (φ, λ) and altitude h of the bridge site, convert them into plane coordinates (x, y) through UTM projection, and form a spatial feature vector S;

[0123] Perform one-hot encoding on categorical parameters (such as construction technology and maintenance method). For example, cast-in-place technology "TC" is coded as [1,0], and prefabricated assembly "TP" is coded as [0,1].

[0124] Normalize numerical parameters (such as concrete strength and reinforcement ratio) and map them to the [0,1] interval using the Min-Max normalization method;

[0125] The historical sample data comes from 20 similar bridges built in the region in the past 10 years, including inspection reports at various stages, construction logs and geographic information archives.

[0126] Step 2: Calculate the decay rate and build a multi-output regression model;

[0127] Calculation of effect parameter attenuation rate: The bridge service life (design life of 100 years) is divided into five stages: T1 (0 years, when it is completed), T2 (20 years, regular inspection 1), T3 (40 years, regular inspection 2), T4 (60 years, regular inspection 3), T5 (80 years, regular inspection 4), and T6 (100 years, the end of service);

[0128] Calculation of stage duration weight: For example, if the first stage lasts 20 years, accounting for 20% of the total period, the weight w1=0.2.

[0129] For the concrete compressive strength E1, the first stage attenuation rate D1,1=(E1(T1)-E1(T2)) / E1(T1), and the comprehensive attenuation rate D1=Σ(Dn,1・wn) (n=1-5);

[0130] Multi-output regression model architecture:

[0131] Input layer: concatenates 15-dimensional construction parameters and 3-dimensional spatial features to form an 18-dimensional input vector X;

[0132] Hidden layer: A fully connected layer with 256 neurons is used, with the activation function ReLU. The spatial feature weight matrix Ws is set to give higher weights to features such as high altitude (h>100m) and humid and rainy areas (based on the climate zones associated with UTM coordinates).

[0133] Output layer: 4 independent output heads, corresponding to the prediction of 4 effect parameters.

[0134] The loss function adopts WMSE, and the weights β1=0.4 (compressive strength), β2=0.3 (rebar corrosion), β3=0.2 (pavement deviation), and β4=0.1 (dynamic frequency) are set according to the engineering focus to highlight the importance of structural safety parameters.

[0135] The generalization ability of the model was evaluated using 5-fold cross-validation. The training data contained 16 bridge samples and the test data contained 4 bridges.

[0136] Step 3: Improve the genetic algorithm to optimize the parameter combination;

[0137] Optimization function and constraints: objective function minF=0.4D1+0.3D2+0.2D3+0.1D4;

[0138] Constraints: Lower limits of effect parameters: E1 ≥ 30 MPa, E2 ≤ 5 μm / year, E3 ≤ 5 mm, E4 ≥ 10 Hz;

[0139] Construction parameter range: such as concrete strength C1∈[C-25,C-40], prestressing tension control stress C6∈[1200MPa,1400MPa];

[0140] Algorithm implementation: The fitness function uses 1 / (F+ε) as the fitness (ε is a very small positive number to avoid the denominator being zero), transforming the minimization problem into a maximization problem;

[0141] Genetic operation: Crossover: Simulated binary crossover (SBX) is used, and the crossover weight is set to 0.9 for highly sensitive parameters such as main material properties and prestressing parameters, and 0.6 for other parameters;

[0142] Mutation: Polynomial mutation, with a mutation probability of 1 / 4 (number of effect parameters), and automatically reducing the mutation amplitude for parameters close to the upper limit of the intensity level (for example, when C1=C-30, the step length is reduced from 0.5 to 0.2);

[0143] Elite retention: The top 20% optimal solutions in each generation are retained and directly enter the next generation;

[0144] Termination condition: the mean change rate of the objective function is less than 1% for 10 consecutive generations or 50 generations have been iterated;

[0145] Step 4: Visualization platform construction and early warning application;

[0146] A 3D bridge model is built based on Revit, integrating parametric components (main beams, piers, and pile foundations). It is connected to the construction management system through the IFC standard, synchronizing construction parameters (such as concrete pouring time, curing temperature and humidity) in real time. The digital twin has:

[0147] 3D visual browsing: supports hierarchical display of construction progress and quality parameters;

[0148] Historical data tracing: you can query the material batch and test report of any component;

[0149] Place sensors at key locations:

[0150] Strain sensors (monitoring E1) and acceleration sensors (monitoring E4) are arranged in the middle of the main beam span; corrosion electrodes are built into the concrete protective layer of the pier (monitoring E2); and thickness sensors are buried in the bridge deck pavement layer (monitoring E3).

[0151] The LSTM model is used to predict the effect parameters for the next 10 years. The input is the time series data of the past 5 years (including influencing factors such as ambient temperature, humidity, and traffic load), and the model parameters are updated every quarter.

[0152] When the single error between the multi-output regression model prediction value and the LSTM time series prediction value exceeds 10% (for example, the difference between the E1 prediction value is >3MPa) or the cumulative error of multiple items is >15%, the platform triggers an alert and pushes a list of abnormal parameters to the on-site engineer.

[0153] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A visual supervision method for engineering quality applied to municipal construction, characterized by: The method comprises the following steps: Step 1: Construct construction parameter sets and effect parameter sets; obtain historical bridge construction sample data, process geographic features to form spatial feature vectors; complete parameter encoding and normalization; Step 2: Calculate the attenuation rate of the effect parameters, build a multi-output regression model, combine the construction parameters and spatial features as input, and train the model using the WMSE loss function; Step 3: Construct an optimization function with the attenuation rate as the target, and use the improved genetic algorithm to solve the optimal parameter combination in combination with the constraints; Step 4: Build a visualization platform, construct a digital twin based on BIM, predict effect parameters based on sensors and LSTM models, compare the prediction values ​​of different models, and issue an early warning when the threshold is exceeded; In step 1, construct the construction parameter set, expressed as: [C1, C2, ..., C n ]; among them, C1, C2, …, C n Represent the 1st, 2nd,…, nth construction parameters respectively; Construct the effect parameter set, expressed as: [E1,E2,…,E m ]; where E1, E2, …, E m Represent the 1st, 2nd,…,mth effect parameters respectively; Obtain historical bridge construction samples and corresponding parameter data, effect parameter data corresponding to different stages, and geographical features; Geographical features include latitude and longitude (φ, λ) and altitude h; Encode parameters using a preset unified format; For categorical parameters, one-hot encoding is performed, and for numerical parameters, normalization is performed; Convert the longitude and latitude (φ, λ) to UTM projection coordinates (x, y) to form a spatial feature vector S = (x, y, h); In step 2, the bridge service life is divided into N stages, and the time for obtaining effect parameter data is expressed as: [T1, T2, …, T N+1 ]; For the effect parameter E i , define the attenuation index of each stage: For the nth stage, calculate the decay rate D n,i =(E n,i -E (n+1),i ) / E n,i ; Where i∈{1,2,…,m}; n∈{1,2,…,N}; D n,i Indicates the effect parameter E of the nth stage i The decay rate of E n,i Indicates the effect parameter E at the nth moment i Value; E (n+1),i Indicates the effect parameter E at the (n+1)th moment i value; Calculate the comprehensive decay rate: For the nth stage, the decay rate weight is the ratio of stage length to duration, w n =(T n+1 -T n ) / (T N+1 -T1); Comprehensive attenuation rate D i =Σ n=1 N D n,i w n ; A multi-output regression model is constructed to achieve nonlinear mapping from construction parameters, spatial features to effect parameters. The model architecture is divided into three layers: Integrate the input layer features, concatenate the construction parameters and the spatial feature vector to form the input layer feature X with a dimension of (n+3); Use the fully connected layer to extract the nonlinear relationship between parameters, and the activation function is ReLU; Set the spatial feature weight matrix W s , calculate the spatial influence coefficient; For m effect parameters, construct independent output heads; each output head corresponds to the prediction of one effect parameter; The historical bridge sample is represented as (X k ,E k ), where X k is the input layer feature of the kth sample; E k =[E k1 ,E k2 ,…,E km ] is the corresponding effect parameter value; Divide the training set and test set, and use K-fold cross validation to evaluate the generalization ability The loss function uses weighted multi-output mean square error WMSE: Loss=Σ i=1 m β i (1 / N)Σ k=1 N (E ki '-E ki ) 2 ; Among them, β i Indicates the effect parameter E i Weight; N represents the number of batch samples; E ki ' represents the kth sample effect parameter E i Predicted value; E ki Indicates the kth sample effect parameter E i Actual value.

2. The method for visually monitoring engineering quality applied to municipal construction according to claim 1 is characterized in that: In step 3, under the given spatial features [x*, y*, h*], the optimization function is constructed: minF=Σ i=1 m γ i D i ; Among them, γ i Indicates the effect parameter E i corresponding weights; Set constraints: E i ≥E i,min Among them, E i,min Indicates the effect parameter E set i The minimum value to be achieved; p j ∈[p j,min ,p j,max ]; where j∈{1,2,…,n}; p i represents the jth construction parameter; p j,min ,p j,max Represent the construction parameters p j The preset minimum and maximum values ​​of Based on the optimization function and constraints, the improved genetic algorithm GA is used to solve the optimal combination of construction parameter values: The inverse of the objective function F is used as fitness; a penalty term is added to individuals that do not meet the specification constraints; simulated binary crossover (SBX) is used, and the crossover probability is set; polynomial mutation is used with a mutation probability of 1 / m; before each generation, the A1% optimal solution is preset and directly enters the next generation; Set termination conditions: the mean change rate of the objective function for a preset number of B1 generations is less than the preset threshold A2%, or the number of iterations reaches a preset number of B2 generations; The optimal combination of construction parameter values ​​is obtained, and the corresponding effect parameter prediction values ​​are obtained based on the multi-output regression model.

3. The method for visually monitoring engineering quality applied to municipal construction according to claim 2 is characterized in that: In step 4, the bridge is constructed based on the optimal combination of construction parameter values, and a digital twin of the bridge is built based on BIM technology to achieve real-time synchronous display of the physical bridge and the virtual model; Sensors are deployed to obtain real-time data on bridge construction effect parameters, and the values ​​of each effect parameter are predicted based on the LSTM time series model. Data is continuously collected to update the time series model to obtain time-series-based effect parameter predictions. Compare the corresponding effect parameter prediction values ​​based on the multi-output regression model with the effect parameter prediction values ​​based on time series; when one or more errors between the two are greater than the set threshold, an early warning is issued and the specific effect parameters are displayed for the administrator's reference.

4. A system for visualizing the quality of construction projects applied to municipal construction, applied to the method for visualizing the quality of construction projects applied to municipal construction as claimed in any one of claims 1 to 3, characterized in that: The system includes a data preprocessing module, a model building module, a parameter optimization module and a monitoring and early warning module; The data preprocessing module is used to construct a construction parameter set and an effect parameter set; obtain historical bridge construction sample data, process geographic features to form a spatial feature vector; and complete parameter encoding and normalization. The model construction module is used to calculate the effect parameter attenuation rate, build a multi-output regression model, combine the construction parameters and spatial features as input, and train the model using the WMSE loss function. The parameter optimization module is used to construct an optimization function with the attenuation rate as the target, and use an improved genetic algorithm to solve the optimal parameter combination in combination with the constraints. The monitoring and early warning module is used to build a visualization platform, construct a digital twin based on BIM, predict effect parameters based on sensors and LSTM models, compare prediction values ​​of different models, and issue an early warning when the threshold is exceeded.

5. The visual supervision system for engineering quality applied to municipal construction according to claim 4 is characterized by: The data preprocessing module includes a parameter encoding unit and a space conversion unit; The parameter encoding unit is used to encode the construction parameter set and the effect parameter set; the space conversion unit is used to convert the geographic features into UTM coordinates.

6. The visual supervision system for engineering quality applied to municipal construction according to claim 5 is characterized by: The model building module includes a decay calculation unit, a model training unit and a verification and evaluation unit; The attenuation calculation unit is used to calculate the stage attenuation rate and the weighted comprehensive attenuation rate; the model training unit is used to construct a multi-output regression model; the verification and evaluation unit is used to evaluate the model based on K-fold cross-validation and the test set verifies the generalization ability.

7. The visual supervision system for engineering quality applied to municipal construction according to claim 6 is characterized by: The parameter optimization module includes a constraint construction unit and a genetic algorithm unit; The constraint construction unit is used to set parameter boundary constraints and effect parameter threshold constraints; the genetic algorithm unit is used to use improved GA to solve the optimal parameter combination.

8. The visual supervision system for engineering quality applied to municipal construction according to claim 7 is characterized by: The monitoring and early warning module includes a BIM modeling unit, a time series prediction unit, a model comparison unit and an early warning trigger unit; The BIM modeling unit is used to construct a digital twin of the bridge based on the optimal parameter combination to achieve visualization; the time series prediction unit is used to obtain real-time data through sensors, and the LSTM model predicts the time series changes of effect parameters; the model comparison unit is used to compare the predicted value of the multi-output regression model with the predicted value of the LSTM; the early warning trigger unit is used to trigger an early warning when the error in the comparison is greater than a preset threshold, and locate abnormal parameters.

Images