Engineering quality safety risk intelligent assessment method and system

Abstract

The invention discloses an engineering quality safety risk intelligent assessment method and system, relates to the technical field of engineering quality safety assessment, adopts multiple types of sensors to collect data, breaks through the limitation that the traditional structure monitoring only depends on a single index to judge the risk, and realizes cross-parameter and cross-component structure state comprehensive perception. The problem that the structure risk is commonly evolved by multiple factors is solved, and the overall recognition precision is improved. A sliding window and a periodic variable coefficient calculation mechanism are introduced, and the system can dynamically extract fluctuation characteristics of structure response and identify different evolution forms of short-time fluctuation, medium-term creep and long-term mutation. And the defects of insufficient abnormal precursor identification capability and serious hysteresis response in the traditional method are overcome. According to the method, an abnormal mutation index AEI and an evolution acceleration factor EAF are constructed, the formation process and the development path of the risk are identified no longer only on the basis of whether a certain moment value exceeds a threshold value or not, and the risk process perception capability is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering quality and safety assessment, and in particular to an intelligent assessment method and system for engineering quality and safety risks. Background Art

[0002] Engineering quality and safety assessment, a key application area that integrates civil engineering, structural engineering, and intelligent monitoring technologies, has garnered increasing attention in recent years as the complexity of large-scale infrastructure construction increases. Within this macro-sector, structural health monitoring technology has become a crucial tool for ensuring project safety throughout its lifecycle. Its core task is to dynamically monitor critical structural components using multi-source sensors to promptly identify potential degradation and abnormal evolution trends.

[0003] Currently, the identification of abnormal risks in engineering structural health monitoring mostly relies on threshold judgment or isolated anomaly detection based on single sensor data. Conventional methods ignore the coupling and temporal evolution of multiple parameters in structural behavior. This single-point monitoring mechanism has significant drawbacks: First, structural anomalies are often gradual processes driven by the coupling of multiple parameters, and a single point anomaly cannot accurately reflect the overall trend. Second, traditional static threshold methods cannot capture the evolutionary paths of gradual and sudden risks, and are prone to monitoring blind spots known as "silence before mutation."

[0004] The fundamental reason for these shortcomings is that traditional monitoring methods focus more on identifying instantaneous outliers and lack a comprehensive understanding of the spatial and temporal dimensions of structural response time series. In actual engineering applications, parameters such as localized differential foundation settlement, superstructure tilt, and stress fluctuations often do not undergo synchronous, sudden changes, but instead exhibit a hidden evolutionary pattern of "first slow change, then sudden change." Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a method and system for intelligent assessment of engineering quality and safety risks, which solves the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent assessment system for engineering quality and safety risks, including a time series data acquisition and preprocessing module, a data variation analysis module, an abnormal mutation index fusion module, a driving factor identification module, a warning threshold dynamic determination module, and a trend evolution report generation module;

[0007] The time series data acquisition and preprocessing module collects engineering quality data through sensors, fits it into the original data set W, and performs preprocessing to obtain the quality data set ZW;

[0008] The data variation analysis module calculates the local variation degree of the data in the quality data set ZW within a fixed period T and obtains the coefficient of variation Ψ;

[0009] The abnormal mutation index fusion module fuses the obtained coefficient of variation Ψ to construct the abnormal mutation index AEI;

[0010] The driving factor identification module identifies the acceleration of the abnormal mutation index AEI, determines whether there is a mutation driving point in the time series of the abnormal mutation index AEI, obtains the evolution acceleration factor EAF, and identifies potential hidden dangers;

[0011] The early warning threshold dynamic judgment module jointly judges the abnormal mutation index AEI and evolution acceleration factor EAF based on the adaptive dynamic threshold of historical statistics to obtain the risk level;

[0012] The trend evolution report generation module generates an execution strategy based on the acquired abnormal mutation index AEI, evolution acceleration factor EAF and risk level.

[0013] Preferably, the time series data acquisition and preprocessing module includes a parameter acquisition and fitting unit and a data preprocessing unit;

[0014] The parameter acquisition and fitting unit collects engineering quality data through sensors deployed at the engineering structure location, including foundation secondary settlement increment Adz, crack width micro-drift value Acw, stress rebound frequency deviation value Are and component inclination gradient Asp, and fits them into the original data set W;

[0015] The secondary settlement increment Adz of the foundation is acquired through multi-point displacement meters and fiber optic displacement sensors:

[0016] Specifically, it is obtained by the difference between the foundation secondary settlement increment Adz at time t and the foundation secondary settlement increment Adz at time t-1;

[0017] The crack width micro-drift value Acw is acquired through the crack meter and the resistive crack width sensor:

[0018] The crack width micro-drift value Acw is obtained by the following formula:

[0019] ;

[0020] Where Ct represents the real-time measurement value of the crack width at time t, and Cref represents the reference width of the crack during the initial stable period;

[0021] The stress rebound frequency deviation value Are is acquired through the stress wave frequency sensor and the high-frequency structural response sensor:

[0022] The stress rebound frequency deviation value Are is obtained by the following formula:

[0023] ;

[0024] Where fr(t) represents the instantaneous resonance frequency in the rebound stage after unloading at time t, and fs(t) represents the background frequency in the stable stage at time t;

[0025] The component inclination gradient Asp is acquired through MEMS inclination sensor and fiber optic inclinometer:

[0026] The component inclination gradient Asp is obtained by the following formula:

[0027] ;

[0028] Wherein, JD(t) represents the inclination angle of the component at time t, JD(t-1) represents the inclination angle of the component at time t-1, and Δt represents the time interval;

[0029] The data preprocessing unit cleans and normalizes the original data set W to obtain the quality data set ZW;

[0030] Cleaning is done by identifying outliers in the original dataset W using adjacent transition determination and local linear fitting residual thresholding;

[0031] For any data Wo in the original data set W, construct a local linear fit and obtain the fitting value Wofit(t);

[0032] Among them, Wo represents the o-th data in the original data set W;

[0033] The fitted value Wofit(t) is obtained by the following formula:

[0034] Wofit(t)=at+b;

[0035] Where Wofit(t) represents the fitted value, specifically the fitted value estimated by the linear fitting function for the o-th item in the original data set W at time t, a represents the fitted slope term, and b represents the fitted intercept term;

[0036] Calculate the fitting residual E based on the fitting value Wofit(t) and determine whether it is abnormal;

[0037] The fitting residual E is obtained by the following formula:

[0038] E(t)=Wo(t)-Wofit(t);

[0039] Where E(t) represents the fitting residual at time t, and Wo(t) represents the o-th data item in the original data set W at time t;

[0040] The judgment formula is as follows:

[0041] ;

[0042] Where Ejump represents the jump tolerance factor used to adjust the judgment sensitivity, which is usually set between 1.5 and 2.0. The larger the value, the more conservative it is. Δ(t-1, t+1) represents the difference between adjacent points.

[0043] If the residual of the current point is larger than the average jump between adjacent points, it means that the point may be a "jump outlier" and needs to be cleared;

[0044] Normalization processing is performed by using the maximum and minimum normalization method to process the data in the original data set W to obtain the quality data set ZW;

[0045] ;

[0046] Where ZWo represents the o-th data item in the quality dataset ZW, Wo represents the o-th data item in the original dataset W, minWo represents the valley value of the o-th data item in the original dataset W, and maxWo represents the peak value of the o-th data item in the original dataset W.

[0047] Preferably, the data variation analysis module extracts the local variation intensity of each parameter in the quality data set ZW to obtain the coefficient of variation Ψ, and the acquisition steps are as follows:

[0048] S1. Set a fixed period T and construct a multi-scale nested window: T1=T, T2=2T, T3=0.5T;

[0049] The setting of multi-scale nested windows (T1, T2, T3) is not arbitrary, but is intended to extract variation features from multiple angles in time series data and improve the sensitivity and adaptability to the evolution of different structural anomalies;

[0050] The multi-scale nested window design (T1, T2, T3) is an asymmetric dynamic ratio construction method used to integrate the differences between long-term stable trends and short-term severe disturbances. This allows for a comprehensive perception of mutation precursors, anomaly intensity, and evolutionary trends through time series statistics without relying on additional hardware.

[0051] T1 is used as the main period to calculate the coefficient of variation Ψ on the basic time scale o (T1) to capture the local risk change characteristics; the fixed period T here is usually determined according to structural characteristics or monitoring frequency, such as 1 day, 1 week and 1 month.

[0052] T2=2T is used to construct a stable trend benchmark for the variation enhancement factor (VAF). Doubling the period can smooth mutations and capture trend stability. When constructing the variation enhancement factor (VAF), T2 provides a relatively smooth low-frequency trend.

[0053] This helps identify the difference between sudden changes and normal trends, and improves the ability to capture slow-changing hidden dangers.

[0054] T3 = 0.5T is used to enhance sensitivity to abnormal spike fluctuations; by shortening the period, it enhances the response to sudden small-scale anomalies; when compared with T2, it can form a contrast gradient of mutation speed, constituting a molecule of mutation enhancement factor;

[0055] Suitable for identifying "precursory micro-changes" and "local disturbances".

[0056] S2. For the oth data item ZWo in the quality data set ZW, within a fixed period T1, calculate and obtain the sliding variation coefficient Ψo;

[0057] The sliding variation coefficient Ψo is obtained by the following formula:

[0058] ;

[0059] Where Ψo(T1) represents the sliding coefficient of variation of the o-th data item in the quality data set ZW within a fixed period T1, σZWo(T1) represents the standard deviation of the o-th data item in the quality data set ZW within a fixed period T1, μZWo(T1) represents the mean of the o-th data item in the quality data set ZW within a fixed period T1, and eo represents a non-zero constant;

[0060] S3, by introducing the variation enhancement factor VAF, the sensitivity of different trend states is improved;

[0061] The variation enhancement factor VAF is obtained by the following formula:

[0062] ;

[0063] Where ln represents the logarithmic function, Ψo(T2) represents the sliding coefficient of variation of the o-th data item in the quality data set ZW within a fixed period T2, and Ψo(T3) represents the sliding coefficient of variation of the o-th data item in the quality data set ZW within a fixed period T3;

[0064] S4, obtaining the coefficient of variation Ψ;

[0065] Ψ=Ψo(T1)*VAF;

[0066] Where Ψ represents the coefficient of variation, including the coefficient of variation of foundation secondary settlement increment ΨAdz, the coefficient of variation of crack width micro-drift value ΨAcw, the coefficient of variation of stress rebound frequency deviation ΨAre, and the coefficient of variation of component inclination gradient rate ΨAsp.

[0067] Preferably, the abnormal mutation index fusion module performs nonlinear structural aggregation on the obtained foundation secondary settlement increment variation coefficient ΨAdz, crack width micro-drift value variation coefficient ΨAcw, stress rebound frequency deviation value variation coefficient ΨAre and component inclination gradient variation coefficient ΨAsp to obtain the abnormal mutation index AEI;

[0068] The abnormal mutation index AEI is obtained by the following formula:

[0069] ;

[0070] Where ln represents the exponential function, A represents the nonlinear enhancement factor (the value of the nonlinear enhancement factor A is 1.3), and arctan represents the inverse tangent function;

[0071] The abnormal mutation index (AEI) formula maps various types of abnormal risks into a single, measurable mutation index (AEI) by performing nonlinear transformation and unified aggregation on four structural variation coefficients with different physical meanings.

[0072] Process various parameters through nonlinear functions to unify the risk response scale; each function selects different transformation logic and risk adjustment curve according to parameter characteristics;

[0073] The abnormal mutation index (AEI) is no longer obtained by conventional linear superposition. Instead, different mapping functions are designed based on the engineering physics behavior mechanism of each parameter. Each transformation takes into account the goals of "early identification" and "late mutation regulation".

[0074] ΨAdz 2 It indicates that the foundation settlement fluctuates and tends to accumulate slowly. As the value increases, the risk increases exponentially.

[0075] It indicates that the crack opening and closing is unstable. Low values ​​have early warning significance, while high values ​​are suppressed to highlight early fluctuations.

[0076] Ln (1 + ΨAre) represents the stress rebound frequency deviation. High-frequency disturbances are often accompanied by local microcracks.

[0077] arctanΨAsp represents the gradient of the tilt angle, which controls the saturation response at high tilt angles;

[0078] Compare the obtained abnormal mutation index AEI with the preset abnormal mutation threshold Tyc to obtain the mutation status;

[0079] The mutation status is obtained by matching:

[0080] When the abnormal mutation index AEI is less than the abnormal mutation threshold Tyc*0.5, it indicates that the mutation status is stable and the monitoring is normal;

[0081] When the abnormal mutation threshold Tyc*0.5≤abnormal mutation index AEI<abnormal mutation threshold Tyc, it indicates a reversible abnormality, local intensified monitoring, and trend analysis is initiated;

[0082] When the abnormal mutation threshold Tyc≤abnormal mutation index AEI<abnormal mutation threshold Tyc*1.5, it indicates a precursor of synergistic mutation and issues an early warning;

[0083] When the abnormal mutation threshold Tyc*1.5≤abnormal mutation index AEI, it indicates a high-risk state of multi-source structural mutation and the clinical inspection mechanism is activated.

[0084] Preferably, the driving factor identification module includes an acceleration identification unit and a mutation driving point determination unit;

[0085] The acceleration recognition unit constructs the evolution acceleration a by extracting the second-order change characteristics of the abnormal mutation index AEI;

[0086] The evolution acceleration a is obtained by the following formula:

[0087] ;

[0088] Wherein, AEI(t) represents the abnormal mutation index at time t, AEI(t+1) represents the abnormal mutation index at time t+1, AEI(t-1) represents the abnormal mutation index at time t-1, and Δt represents the time interval;

[0089] By introducing the risk growth inertia coefficient GX, the obtained evolution acceleration a is corrected to obtain the evolution acceleration factor EAF;

[0090] The risk growth inertia coefficient GX is obtained by the following formula:

[0091] ;

[0092] Wherein, AEI(t-3) represents the abnormal mutation index at time t-3, ae represents a positive number, and V(t) represents the first-order derivative of the abnormal mutation index AEI; it is obtained by the ratio of the difference between the abnormal mutation index AEI(t) at time t and the abnormal mutation index AEI(t-1) at time t-1 to the time interval Δt;

[0093] The evolution acceleration factor EAF is obtained by multiplying the risk growth inertia coefficient GX and the evolution acceleration a.

[0094] Preferably, the mutation driving point determination unit calculates and obtains the evolution threshold Teaf according to the evolution fluctuation of the evolution acceleration factor EAF;

[0095] The evolution threshold Teaf is obtained by the following formula:

[0096] ;

[0097] Where μEAF represents the mean of the evolution acceleration factor, σEAF represents the standard deviation of the evolution acceleration factor, and af represents the sensitivity coefficient;

[0098] Compare the evolution acceleration factor EAF with the evolution threshold Teaf to determine the risk status;

[0099] The risk status is obtained in the following ways:

[0100] When the evolution acceleration factor EAF ≤ the evolution threshold Teaf, it indicates a normal stable state;

[0101] When the evolution acceleration factor EAF is greater than the evolution threshold Teaf, it indicates a risk state, hidden dangers appear, and an early warning is issued.

[0102] Preferably, the early warning threshold dynamic determination module includes a joint index construction unit and a risk level classification unit;

[0103] The joint index construction unit jointly constructs the acquired evolution acceleration factor EAF and abnormal mutation index AEI to obtain the joint index ARIA;

[0104] The combined indicator ARIA is obtained by the following formula:

[0105] ;

[0106] Where r1 represents the risk intensity adjustment factor, and r2 represents the risk acceleration adjustment factor.

[0107] Preferably, the risk level classification unit extracts all values ​​of the joint indicator ARIA from the historical period to form a set Har;

[0108] The threshold value THr is obtained by using the mean method and standard deviation method to extract the threshold value of the set Har;

[0109] The index threshold THr was obtained by the sum of the mean of the combined index ARIA and 0.5 times the standard deviation of the combined index ARIA;

[0110] Compare the combined indicator ARIA with the indicator threshold THr to obtain the risk level;

[0111] The risk level is obtained by matching:

[0112] When the combined indicator ARIA is less than 0.5*indicator threshold THr, it indicates the first risk level and safe state;

[0113] When 0.5*indicator threshold THr≤combined indicator ARIA<indicator threshold THr, it indicates the second risk level and a controllable abnormality occurs;

[0114] When the indicator threshold THr ≤ combined indicator ARIA < 1.5*indicator threshold THr, it indicates the third risk level, and a precursor to mutation occurs;

[0115] When 1.5*index threshold THr≤combined indicator ARIA, it indicates the fourth risk level, an abnormally high-risk state.

[0116] Preferably, the trend evolution report generation module performs trend identification on the time series composed of the abnormal mutation index AEI and the evolution acceleration factor EAF, and constructs a time series curve graph of the abnormal mutation index AEI and an acceleration curve graph of the evolution acceleration factor EAF;

[0117] Abnormal mutation index AEI time series curve: Draw the abnormal mutation index AEI curve according to the timestamp sequence, and mark the peak point, inflection point and risk level segmented color band in the graph;

[0118] The EAF acceleration curve includes the length of the continuous rising section of the EAF and the time interval from the EAF peak to the mutation point, reflecting the response hysteresis.

[0119] The length of the continuous rising section of the evolution acceleration factor EAF represents the duration of the accelerated evolution, in hours, and is used to evaluate the accelerated preparation period before the outbreak of structural risks;

[0120] The time interval from the peak value of the evolution acceleration factor EAF to the mutation point represents the time difference from the maximum acceleration to the extreme value of AEI, reflecting the response hysteresis;

[0121] Generate an execution strategy based on the abnormal mutation index AEI time series curve and evolution acceleration factor EAF acceleration curve, combined with the risk level;

[0122] The execution strategy is matched and obtained in the following ways:

[0123] When at the first risk level, the trend of abnormal mutation index AEI is stable, and the trend of evolution acceleration factor EAF is stable, indicating a stable state. The execution strategy is: normal inspection and periodic sampling;

[0124] When in the second risk level, the abnormal mutation index AEI trend fluctuates slightly, and the evolution acceleration factor EAF trend occasionally rises slightly, indicating that there is a disturbance and no continuous evolution. The execution strategy is to increase the monitoring frequency and start the local trend tracking algorithm;

[0125] When it is at the third risk level, the abnormal mutation index AEI has a significant upward trend, and the evolution acceleration factor EAF has a continuous upward trend, indicating that cumulative instability is forming. The implementation strategy is: preparation of local maintenance plans, structural re-inspection and assessment of whether the function is affected;

[0126] When in the fourth risk level, the abnormal mutation index AEI is close to the extreme value, and the evolution acceleration factor EAF trend peak has appeared, indicating critical evolution and imminent mutation. The execution strategy is: structural unloading, personnel evacuation and suspension of work.

[0127] An intelligent assessment method for engineering quality safety risk includes the following steps:

[0128] Step 1: The time series data acquisition and preprocessing module collects engineering quality data through sensors, fits it into the original data set W, and performs preprocessing to obtain the quality data set ZW;

[0129] Step 2: The data variation analysis module calculates the local variation degree of the data in the quality data set ZW within a fixed period T and obtains the coefficient of variation Ψ;

[0130] Step 3: The abnormal mutation index fusion module fuses the obtained coefficient of variation Ψ to construct the abnormal mutation index AEI;

[0131] Step 4: The driving factor identification module identifies the acceleration of the abnormal mutation index AEI, determines whether there is a mutation driving point in the time series of the abnormal mutation index AEI, obtains the evolution acceleration factor EAF, and identifies potential hidden dangers;

[0132] Step 5: The early warning threshold dynamic judgment module jointly judges the abnormal mutation index AEI and the evolution acceleration factor EAF based on the adaptive dynamic threshold of historical statistics to obtain the risk level;

[0133] Step 6: The trend evolution report generation module generates an execution strategy based on the acquired abnormal mutation index AEI, evolution acceleration factor EAF and risk level.

[0134] The present invention provides a method and system for intelligent assessment of engineering quality and safety risks, which has the following beneficial effects:

[0135] (1) When the system is running, it uses multiple types of sensors to collect data, breaking the limitation of relying on a single indicator to judge risk in traditional structural monitoring, and achieving comprehensive perception of structural status across parameters and components. It solves the problem that structural risks often evolve from multiple factors, and improves the overall recognition accuracy. By introducing a sliding window and a periodic coefficient of variation calculation mechanism, the system can dynamically extract the fluctuation characteristics of structural response and identify different evolutionary forms such as "short-term fluctuation", "medium-term creep" and "long-term mutation". It makes up for the shortcomings of traditional methods in terms of insufficient recognition ability of abnormal precursors and serious lag response.

[0136] The Abnormal Mutation Index (AEI) and Evolution Acceleration Factor (EAF) have been constructed. These systems no longer rely solely on "whether a value at a specific moment exceeds a threshold" but instead identify the formation and development path of risks, enhancing risk "process perception." By building dynamic warning thresholds based on historical data from stable periods, the system can adapt personalized judgment criteria to different structures and environments, avoiding false positives or false negatives based on fixed thresholds.

[0137] (2) Data preprocessing not only performs conventional missing / noise cleaning, but also introduces the "adjacent jump judgment" and "local linear fitting residual threshold" methods to identify trend constraints on outliers. Compared with the static upper and lower limit elimination method, this method is more suitable for capturing the starting point of sudden changes in time series anomalies, improving the system's resistance to interference data and its ability to retain potential risk signals. Through the maximum and minimum normalization method, various engineering parameters are stretched and compressed, so that data from different physical dimensions can be processed uniformly, providing consistency support for the basic data structure for subsequent variation analysis and trend fusion, and solving the problem of cross-dimensional data fusion.

[0138] By collecting micro-variable structural parameters rather than static indicators, we can detect structural degradation signals earlier, intervene in potential hazards in advance, and effectively shorten the risk identification lag period. A cleaning algorithm eliminates sudden data jumps in structural response, effectively improving data stability and the credibility of analysis results, and preventing "false alarms" or "false missed alarms."

[0139] (3) The construction of the abnormal mutation index AEI is no longer the conventional weighted synthesis or simple averaging, but a deep coupling modeling of the coefficient of variation is performed by combining nonlinear functions such as exponential function and inverse tangent function, which strengthens the response degree of the abnormal co-evolution of each dimension, solves the traditional problem of "signal weakening after fusion of multi-source indicators", and significantly improves the system's ability to analyze the abnormal evolution of high-order structures.

[0140] By comparing the abnormal mutation index (AEI) with preset abnormal mutation thresholds and classifying it into four mutation states, a continuous closed-loop control from "monitoring" to "response" is achieved. This mechanism not only identifies the anomaly itself but also categorizes its development direction and controllability, enhancing the system's dynamic adaptability and response accuracy.

[0141] (4) By calculating the current threshold based on the dynamic distribution characteristics of the evolution acceleration factor (EAF), a risk state discrimination mechanism adaptive to the structure is implemented. The system can dynamically adjust the sensitivity according to different structures and different monitoring cycles to adapt to the risk judgment needs in different engineering scenarios. The system is no longer limited to judging whether it is abnormal, but further locates the starting point of abnormal growth, that is, the "starting node of risk evolution." This capability enables the system to identify the "driving point" before reaching a full mutation state, achieving accurate early warning and preemptive intervention, significantly improving the system's depth in identifying mutation precursors. BRIEF DESCRIPTION OF THE DRAWINGS

[0142] Figure 1 This is a flowchart diagram of an intelligent assessment system for engineering quality and safety risks according to the present invention;

[0143] Figure 2 This is a schematic diagram of the steps of an intelligent assessment method for engineering quality and safety risks according to the present invention;

[0144] Figure 3 A schematic diagram of the process of obtaining the combined index of the present invention;

[0145] Figure 4 It is a bar graph of abnormal mutation index of the present invention;

[0146] Figure 5 is a time series trend diagram of the abnormal mutation index of the present invention;

[0147] Figure 6 It is a time series curve diagram of the evolution acceleration factor of the present invention. DETAILED DESCRIPTION

[0148] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0149] Example 1: The present invention provides an intelligent assessment system for engineering quality and safety risks. Figures 1 to 6, including time series data acquisition and preprocessing module, data variation analysis module, abnormal mutation index fusion module, driving factor identification module, early warning threshold dynamic judgment module and trend evolution report generation module;

[0150] The time series data acquisition and preprocessing module collects engineering quality data through sensors, fits it into the original data set W, and performs preprocessing to obtain the quality data set ZW;

[0151] The data variation analysis module calculates the local variation degree of the data in the quality data set ZW within a fixed period T and obtains the coefficient of variation Ψ;

[0152] The abnormal mutation index fusion module fuses the obtained coefficient of variation Ψ to construct the abnormal mutation index AEI;

[0153] The driving factor identification module identifies the acceleration of the abnormal mutation index AEI, determines whether there is a mutation driving point in the time series of the abnormal mutation index AEI, obtains the evolution acceleration factor EAF, and identifies potential hidden dangers;

[0154] The early warning threshold dynamic judgment module jointly judges the abnormal mutation index AEI and evolution acceleration factor EAF based on the adaptive dynamic threshold of historical statistics to obtain the risk level;

[0155] The trend evolution report generation module generates an execution strategy based on the acquired abnormal mutation index AEI, evolution acceleration factor EAF and risk level.

[0156] In this embodiment, data collection using multiple sensors overcomes the limitations of traditional structural monitoring, which relies solely on a single indicator to determine risk, and achieves comprehensive perception of structural status across parameters and components. This addresses the problem that structural risks often evolve from multiple factors, improving overall identification accuracy. By introducing a sliding window and periodic coefficient of variation calculation mechanism, the system can dynamically extract the fluctuation characteristics of structural responses and identify the different evolutionary forms of "short-term fluctuations," "medium-term creep," and "long-term mutations." This overcomes the shortcomings of traditional methods in their lack of ability to identify abnormal precursors and their severe lag in response.

[0157] The Abnormal Mutation Index (AEI) and Evolution Acceleration Factor (EAF) have been constructed. These systems no longer rely solely on "whether a value at a specific moment exceeds a threshold" but instead identify the formation and development path of risks, enhancing risk "process perception." By building dynamic warning thresholds based on historical data from stable periods, the system can adapt personalized judgment criteria to different structures and environments, avoiding false positives or false negatives based on fixed thresholds.

[0158] Leveraging multi-dimensional parameter fusion and trend evolution mechanisms, this system can proactively identify "progressive hazards" or "precursors of sudden changes" that are difficult for traditional systems to detect. The dual-metric mechanism of the Abnormal Change Index (AEI) and the Evolution Acceleration Factor (EAF) balances risk intensity and development speed, enabling the system to naturally suppress false alarms, under-reported alarms, and delayed responses. The dynamic threshold algorithm is tailored to project specific needs, independent of manual settings, and is applicable to a variety of structures, including bridges, tunnels, and high-rise buildings.

[0159] Example 2: This example is explained in Example 1. Figure 1 and Figure 3 ,Specifically: the time series data acquisition and preprocessing module includes a ,parameter acquisition and fitting unit and a data preprocessing unit;

[0160] The parameter acquisition and fitting unit collects engineering quality data through sensors deployed at the engineering structure location, including foundation secondary settlement increment Adz, crack width micro-drift value Acw, stress rebound frequency deviation value Are and component inclination gradient Asp, and fits them into the original data set W;

[0161] The foundation secondary settlement increment Adz is acquired through multi-point displacement meter and optical fiber displacement sensor;

[0162] The crack width micro-drift value Acw is acquired through the crack meter and the resistive crack width sensor;

[0163] The stress rebound frequency deviation value Are is acquired through the stress wave frequency sensor and the high-frequency structural response sensor;

[0164] The component inclination gradient Asp is acquired through MEMS inclination sensor and fiber optic inclinometer;

[0165] The data preprocessing unit cleans and normalizes the original data set W to obtain the quality data set ZW;

[0166] Cleaning is done by identifying outliers in the original dataset W using adjacent transition determination and local linear fitting residual thresholding;

[0167] Normalization processing is performed by using the maximum and minimum normalization method to process the data in the original data set W to obtain the quality data set ZW;

[0168] ;

[0169] Where ZWo represents the o-th data item in the quality dataset ZW, Wo represents the o-th data item in the original dataset W, minWo represents the valley value of the o-th data item in the original dataset W, and maxWo represents the peak value of the o-th data item in the original dataset W.

[0170] This embodiment introduces four types of specific quantitative parameters of structural behavior: foundation secondary settlement increment Adz, crack width micro-drift value Acw, stress rebound frequency deviation value Are, and component inclination gradient Asp. Compared with the traditional structural monitoring method that focuses on a single value, these parameters have stronger dynamic response characteristics and gradual change recognition capabilities, and can capture the implicit evolution signs of "transition from stability" earlier. The parameters are collected through common sensor types in current engineering monitoring, such as fiber optic sensors, multi-point displacement meters, MEMS inclination sensors, etc., and can be upgraded in situ without the need for new hardware equipment, thereby ensuring the deployability and economy of the system in engineering, and meeting the core innovation requirement of "no need to add new sensors".

[0171] Data preprocessing not only performs conventional missing / noise cleaning, but also introduces the "adjacent jump judgment" and "local linear fitting residual threshold" methods to identify trend constraints on outliers. Compared with the static upper and lower limit elimination method, this method is more suitable for capturing the starting point of sudden changes in time series anomalies, improving the system's immunity to interference data and its ability to retain potential risk signals. The maximum and minimum normalization method is used to stretch and compress the intervals of various engineering parameters, so that data from different physical dimensions can be processed uniformly, providing consistency support for the basic data structure for subsequent variation analysis and trend fusion, and solving the problem of difficult cross-dimensional data fusion.

[0172] By collecting micro-variable structural parameters rather than static indicators, we can detect structural degradation signals earlier, intervene in potential hazards in advance, and effectively shorten the risk identification lag period. A cleaning algorithm eliminates sudden data jumps in structural response, effectively improving data stability and the credibility of analysis results, and preventing "false alarms" or "false missed alarms."

[0173] Example 3, this example is explained in Example 2, please refer to Figure 1 and Figure 4 ,Specifically: the data variation analysis module extracts the local variation intensity of ,each parameter in the quality data set ZW and obtains the coefficient of variation Ψ. ,The acquisition steps are as follows;

[0174] S1. Set a fixed period T and construct a multi-scale nested window: T1=T, T2=2T, T3=0.5T;

[0175] S2. For the oth data item ZWo in the quality data set ZW, within a fixed period T1, calculate and obtain the sliding variation coefficient Ψo;

[0176] The sliding variation coefficient Ψo is obtained by the following formula:

[0177] ;

[0178] Where Ψo(T1) represents the sliding coefficient of variation of the o-th data item in the quality data set ZW within a fixed period T1, σZWo(T1) represents the standard deviation of the o-th data item in the quality data set ZW within a fixed period T1, μZWo(T1) represents the mean of the o-th data item in the quality data set ZW within a fixed period T1, and eo represents a non-zero constant;

[0179] S3, by introducing the variation enhancement factor VAF, the sensitivity of different trend states is improved;

[0180] The variation enhancement factor VAF is obtained by the following formula:

[0181] ;

[0182] Where ln represents the logarithmic function, Ψo(T2) represents the sliding coefficient of variation of the o-th data item in the quality data set ZW within a fixed period T2, and Ψo(T3) represents the sliding coefficient of variation of the o-th data item in the quality data set ZW within a fixed period T3;

[0183] S4, obtaining the coefficient of variation Ψ;

[0184] Ψ=Ψo(T1)*VAF;

[0185] Where Ψ represents the coefficient of variation, including the coefficient of variation of foundation secondary settlement increment ΨAdz, the coefficient of variation of crack width micro-drift value ΨAcw, the coefficient of variation of stress rebound frequency deviation ΨAre, and the coefficient of variation of component inclination gradient rate ΨAsp.

[0186] The abnormal mutation index fusion module performs nonlinear structural aggregation on the acquired foundation secondary settlement increment variation coefficient ΨAdz, crack width micro-drift value variation coefficient ΨAcw, stress rebound frequency deviation value variation coefficient ΨAre and component inclination gradient variation coefficient ΨAsp to obtain the abnormal mutation index AEI, as shown in Table 1.

[0187] The abnormal mutation index AEI is obtained by the following formula:

[0188] ;

[0189] Where ln represents the exponential function, A represents the nonlinear enhancement factor, and arctan represents the inverse tangent function;

[0190] Specific examples:

[0191] Table 1: Abnormal mutation index AEI calculation table;

[0192] Group 1, Group 2, Group 3, Group 4, and Group 5 represent five groups of collected data.

[0193] Compare the obtained abnormal mutation index AEI with the preset abnormal mutation threshold Tyc to obtain the mutation status;

[0194] The mutation status is obtained by matching:

[0195] When the abnormal mutation index AEI is less than the abnormal mutation threshold Tyc*0.5, it indicates that the mutation status is stable and the monitoring is normal;

[0196] When the abnormal mutation threshold Tyc*0.5≤abnormal mutation index AEI<abnormal mutation threshold Tyc, it indicates a reversible abnormality, local intensified monitoring, and trend analysis is initiated;

[0197] When the abnormal mutation threshold Tyc≤abnormal mutation index AEI<abnormal mutation threshold Tyc*1.5, it indicates a precursor of synergistic mutation and issues an early warning;

[0198] When the abnormal mutation threshold Tyc*1.5≤abnormal mutation index AEI, it indicates a high-risk state of multi-source structural mutation and the clinical inspection mechanism is activated.

[0199] In this embodiment, by setting a fixed period T and constructing a multi-scale nested window to perform local variation analysis, a significant breakthrough is made in the problem of "fixed period / single scale judgment leading to loss of trend details" in traditional methods. This method enables the system to identify long-period fluctuation trends at the macro scale, capture local disturbances at the micro scale, and achieve multi-dimensional dynamic analysis of abnormal evolution. Traditional structural monitoring often evaluates the amplitude of fluctuations based on the standard deviation under a fixed threshold, and easily ignores "gradual mutation" signals. This system introduces a "variation enhancement factor VAF" by enhancing the local sliding variability, dynamically amplifying the trend differences between parameters, so that even if the abnormal state does not form a drastic numerical jump, it can be captured and enhanced by the system in the early stage.

[0200] The construction of the abnormal mutation index AEI is no longer a conventional weighted synthesis or simple averaging, but a deep coupling modeling of the coefficient of variation combined with nonlinear functions such as exponential functions and inverse tangent functions, which strengthens the response degree of the anomaly co-evolution in various dimensions, solves the traditional problem of "signal weakening after fusion of multi-source indicators", and significantly improves the system's ability to analyze the abnormal evolution of high-order structures.

[0201] By comparing the abnormal mutation index (AEI) with preset abnormal mutation thresholds and classifying it into four mutation states, a continuous closed-loop control from "monitoring" to "response" is achieved. This mechanism not only identifies the anomaly itself but also categorizes its development direction and controllability, enhancing the system's dynamic adaptability and response accuracy.

[0202] Multi-scale variation analysis and index fusion mechanisms enable the system to respond to subtle deviations from trend fluctuations before risks materialize, effectively reducing the probability of sudden accidents and enhancing the proactive safety management capabilities of structural operations. Rather than relying on fixed thresholds and single time windows, the system constructs a dynamic threshold and state determination framework based on the multidimensional characteristics of time series. This allows for flexible adaptation to different operating conditions, structural types, and changing environmental conditions, enhancing the system's versatility and scalability.

[0203] The four-level mutation status can support managers to initiate monitoring enhancement, trend tracking, early repair or emergency reinforcement strategies as needed, to achieve accurate resource scheduling and management cost optimization.

[0204] Example 4: This example is explained in Example 3. Figure 1 and Figure 6 ,Specifically: the driving factor identification module includes an acceleration ,identification unit and a mutation driving point determination unit;

[0205] The acceleration recognition unit constructs the evolution acceleration a by extracting the second-order change characteristics of the abnormal mutation index AEI;

[0206] The evolution acceleration a is obtained by the following formula:

[0207] ;

[0208] Wherein, AEI(t) represents the abnormal mutation index at time t, AEI(t+1) represents the abnormal mutation index at time t+1, AEI(t-1) represents the abnormal mutation index at time t-1, and Δt represents the time interval;

[0209] By introducing the risk growth inertia coefficient GX, the obtained evolution acceleration a is corrected to obtain the evolution acceleration factor EAF;

[0210] The risk growth inertia coefficient GX is obtained by the following formula:

[0211] ;

[0212] Wherein, AEI(t-3) represents the abnormal mutation index at time t-3, ae represents a positive number, and V(t) represents the first-order derivative of the abnormal mutation index AEI;

[0213] The evolution acceleration factor EAF is obtained by multiplying the risk growth inertia coefficient GX and the evolution acceleration a.

[0214] The mutation driving point determination unit calculates the evolution threshold Teaf according to the evolution fluctuation of the evolution acceleration factor EAF;

[0215] The evolution threshold Teaf is obtained by the following formula:

[0216] ;

[0217] Where μEAF represents the mean of the evolution acceleration factor, σEAF represents the standard deviation of the evolution acceleration factor, and af represents the sensitivity coefficient;

[0218] Compare the evolution acceleration factor EAF with the evolution threshold Teaf to determine the risk status;

[0219] The risk status is obtained in the following ways:

[0220] When the evolution acceleration factor EAF ≤ the evolution threshold Teaf, it indicates a normal stable state;

[0221] When the evolution acceleration factor EAF is greater than the evolution threshold Teaf, it indicates a risk state, hidden dangers appear, and an early warning is issued.

[0222] This embodiment introduces a dynamic indicator, "evolution acceleration a," into structural health risk identification. This expands traditional judgment methods based solely on current risk status or historical averages to a trend identification logic centered on "second-order change characteristics in time series." This mechanism enables the system to identify the direction and rate of change of rapidly increasing risk trends, effectively overcoming the drawback of static thresholds, which are often slow to respond to sudden evolutionary trends.

[0223] The present invention designs the "risk growth inertia coefficient" as a "weighted correction term" for the rising trend of structural risk, fully considering the cumulative growth trend and development inertia of the abnormal mutation index. This enables the system to capture the current rate of change and trace the past risk accumulation process when determining potential mutation driving points, thereby enhancing the historical depth and foresight of the judgment.

[0224] By calculating the current threshold based on the dynamic distribution characteristics of the Evolution Acceleration Factor (EAF), a structurally adaptive risk state identification mechanism is implemented. The system can dynamically adjust its sensitivity based on different structures and monitoring cycles to meet the risk assessment needs of different engineering scenarios. The system is no longer limited to determining whether an anomaly exists, but further locates the starting point of anomaly growth, the "starting point of risk evolution." This capability enables the system to identify "driving points" before a full mutation state is reached, enabling precise early warning and proactive intervention, significantly enhancing the system's depth in identifying mutation precursors.

[0225] The identification method based on evolutionary acceleration enables the system to be sensitive to risk states that "change quickly but in small amounts", especially in micro-change behaviors such as early crack expansion, accumulation of foundation deformation, and stress micro-drift. It can accurately identify potential trends and achieve "one step forward" in risk identification.

[0226] The dynamic judgment mechanism integrates current changes, historical inertia, and fluctuation range, enabling the system to determine not only whether an anomaly exists but also whether the anomaly will persist and intensify, effectively reducing misjudgments and missed detections and enhancing the credibility of the early warning mechanism. Utilizing an adaptive evolutionary threshold mechanism, the system dynamically sets sensitivity levels based on the evolving risk characteristics of the monitored object. This approach is applicable to a variety of project types, including bridges, tunnels, and high-rise buildings, enhancing the system's portability and universality for deployment.

[0227] The system not only outputs the current risk level, but also identifies in advance the "direction of risk evolution in the next few steps". It can provide quantitative support and scientific suggestions for management behaviors such as structural reinforcement, inspection frequency adjustment, and construction plan optimization, and build an intelligent decision-making closed loop based on data evolution.

[0228] Example 5: This example is explained in Example 4. Please refer to Figure 5 and Figure 6 ,Specifically: the early warning threshold dynamic determination module includes a ,joint index building unit and a risk level classification unit;

[0229] The joint index construction unit jointly constructs the acquired evolution acceleration factor EAF and abnormal mutation index AEI to obtain the joint index ARIA;

[0230] The combined indicator ARIA is obtained by the following formula:

[0231] ;

[0232] Where r1 represents the risk intensity adjustment factor, and r2 represents the risk acceleration adjustment factor.

[0233] The risk level division unit extracts all values ​​of the joint indicator ARIA from the historical period to form a set Har;

[0234] The threshold value THr is obtained by using the mean method and standard deviation method to extract the threshold value of the set Har;

[0235] The index threshold THr was obtained by the sum of the mean of the combined index ARIA and 0.5 times the standard deviation of the combined index ARIA;

[0236] Compare the combined indicator ARIA with the indicator threshold THr to obtain the risk level;

[0237] The risk level is obtained by matching:

[0238] When the combined indicator ARIA is less than 0.5*indicator threshold THr, it indicates the first risk level and safe state;

[0239] When 0.5*indicator threshold THr≤combined indicator ARIA<indicator threshold THr, it indicates the second risk level and a controllable abnormality occurs;

[0240] When the indicator threshold THr ≤ combined indicator ARIA < 1.5*indicator threshold THr, it indicates the third risk level, and a precursor to mutation occurs;

[0241] When 1.5*index threshold THr≤combined indicator ARIA, it indicates the fourth risk level, an abnormally high-risk state.

[0242] The trend evolution report generation module identifies the trend of the time series composed of the abnormal mutation index AEI and the evolution acceleration factor EAF, and constructs the abnormal mutation index AEI time series curve and the evolution acceleration factor EAF acceleration curve;

[0243] Abnormal mutation index AEI time series curve: Draw the abnormal mutation index AEI curve according to the timestamp sequence, and mark the peak point, inflection point and risk level segmented color band in the graph;

[0244] The acceleration curve of the evolution acceleration factor EAF includes the length of the continuous rising section of the evolution acceleration factor EAF and the time interval from the peak value of the evolution acceleration factor EAF to the mutation point;

[0245] Generate an execution strategy based on the abnormal mutation index AEI time series curve and evolution acceleration factor EAF acceleration curve, combined with the risk level;

[0246] The execution strategy is matched and obtained in the following ways:

[0247] When it is at the first risk level, it means the status is stable and the execution strategy is: normal inspection and periodic sampling;

[0248] When it is at the second risk level, it means that there is a disturbance but it has not evolved continuously. The execution strategy is to increase the monitoring frequency and start the local trend tracking algorithm.

[0249] When it is at the third risk level, it means that cumulative instability is forming, and the implementation strategy is: preparation of local maintenance plan, structural re-inspection and assessment of whether it affects the function;

[0250] When it is at the fourth risk level, it indicates critical evolution and imminent mutation, and the implementation strategy is: structural unloading, personnel evacuation and suspension of work.

[0251] This implementation no longer treats the abnormal mutation index (AEI) and evolution acceleration factor (EAF) in isolation, but instead constructs a unified joint risk indicator (ARIA) to reflect the comprehensive performance of risk status in terms of both intensity and rate of change. This mechanism effectively overcomes the narrow risk expression and information fragmentation issues in traditional systems, making system judgments more holistic, trend-based, and developmental.

[0252] By automatically extracting assessment thresholds based on the statistical distribution of ARIA over historical monitoring periods, the system categorizes risk levels into four levels. This enables dynamic, self-calibrated, and scenario-adaptive risk grading, overcoming the issues of "failure in different structural scenarios" or "over-warning / delayed alarms" often encountered with fixed threshold models. For the first time, the system establishes a direct mapping logic between risk identification, risk level assessment, and engineering disposal operations, clearly defining the corresponding treatment plan for each risk level (such as initiating trend algorithms, structural unloading, and personnel evacuation). This addresses the long-standing issue of "alarm-only, no-response" in structural monitoring, establishing an intelligent, interconnected closed loop of "perception-identification-response."

[0253] By analyzing the "length of the continuous rising segment" and the "time distance from the peak to the mutation point" in the evolution acceleration factor EAF curve, the system constructs a quantitative assessment of the rhythm of abnormal acceleration evolution, effectively capturing difficult-to-identify non-dramatic change patterns such as "slow-speed sudden changes" or "gentle cumulative anomalies", and greatly improving the system's adaptability to different types of abnormal evolution paths.

[0254] By combining multi-dimensional features to establish a composite judgment model, the system can effectively identify borderline risk states—those with high intensity but slow evolution, or those with low intensity but extremely high acceleration—to avoid missed or delayed risk assessments. Each level of risk is mapped to an execution strategy, enabling managers to implement differentiated resource allocation, inspection frequency adjustments, or construction pacing controls based on risk levels, achieving refined risk response management and optimal resource allocation.

[0255] Example 6, a method for intelligent assessment of engineering quality and safety risks, please refer to Figure 2 , specifically: including the following steps:

[0256] Step 1: The time series data acquisition and preprocessing module collects engineering quality data through sensors, fits it into the original data set W, and performs preprocessing to obtain the quality data set ZW;

[0257] Step 2: The data variation analysis module calculates the local variation degree of the data in the quality data set ZW within a fixed period T and obtains the coefficient of variation Ψ;

[0258] Step 3: The abnormal mutation index fusion module fuses the obtained coefficient of variation Ψ to construct the abnormal mutation index AEI;

[0259] Step 4: The driving factor identification module identifies the acceleration of the abnormal mutation index AEI, determines whether there is a mutation driving point in the time series of the abnormal mutation index AEI, obtains the evolution acceleration factor EAF, and identifies potential hidden dangers;

[0260] Step 5: The early warning threshold dynamic judgment module jointly judges the abnormal mutation index AEI and the evolution acceleration factor EAF based on the adaptive dynamic threshold of historical statistics to obtain the risk level;

[0261] Step 6: The trend evolution report generation module generates an execution strategy based on the acquired abnormal mutation index AEI, evolution acceleration factor EAF and risk level.

[0262] In this embodiment, this method, through six sequential steps, spans raw data collection, identifying abnormal trends, risk classification, and implementing policy output. This approach bridges the entire structural safety assessment process, from "bottom-level perception" to "top-level decision-making." Unlike traditional monitoring systems that "monitor without judgment" or "judgment without control," this solution establishes a closed-loop logic encompassing "perception—fusion—identification—warning—response," enabling truly intelligent structural risk assessment.

[0263] This system integrates the temporal evolution characteristics of multi-dimensional structural state parameters and constructs a trend-sensitive composite indicator through variation analysis and anomaly fusion. This solves the problems of "fragmented judgment of various anomaly characteristics" and "static threshold limitations" in traditional methods, comprehensively improving the systematicity and sensitivity of risk identification. By identifying the acceleration of the abnormal mutation index and determining the driving factors behind it, the system can provide early insights into the source location and timing of the accelerated accumulation of structural risks, effectively addressing the gaps in existing methods in identifying the "precursor stage of mutation" and enabling proactive identification of engineering hazards.

[0264] By building the joint indicator ARIA based on historical data and combining it with dynamic statistical thresholds to stratify risk levels, the assessment system can be adaptively evolved in different structural environments and states, breaking the problems of "generalization failure" and "poor scenario adaptability" in the traditional fixed warning threshold mode.

[0265] Through its mechanism for identifying abnormal trends and locating driving points, the system can identify the evolutionary tendencies of problems before they develop into overt disasters, enabling early detection and response to potential hazards and enhancing the proactive and stable nature of project management. This method can effectively identify sudden, gradual, cyclical, or cumulative risk evolution paths, significantly improving the system's ability to discern the evolutionary patterns of various types of structural instability and expanding its adaptability to complex structural systems. By pairing different intervention strategies with different risk levels, the monitoring system can determine not only whether an abnormality exists but also how to respond, providing a quantitative basis for resource allocation, inspection rhythm adjustments, and reinforcement plan implementation.

[0266] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent assessment system for engineering quality and safety risks, characterized by: It includes time series data acquisition and preprocessing module, data variation analysis module, abnormal mutation index fusion module, driving factor identification module, early warning threshold dynamic determination module and trend evolution report generation module; The time series data acquisition and preprocessing module collects engineering quality data through sensors, fits it into the original data set W, and performs preprocessing to obtain the quality data set ZW; The data variation analysis module calculates the local variation degree of the data in the quality data set ZW within a fixed period T and obtains the coefficient of variation Ψ; The abnormal mutation index fusion module fuses the obtained coefficient of variation Ψ to construct the abnormal mutation index AEI; The driving factor identification module identifies the acceleration of the abnormal mutation index AEI, determines whether there is a mutation driving point in the time series of the abnormal mutation index AEI, obtains the evolution acceleration factor EAF, and identifies potential hidden dangers; The early warning threshold dynamic judgment module jointly judges the abnormal mutation index AEI and evolution acceleration factor EAF based on the adaptive dynamic threshold of historical statistics to obtain the risk level; The trend evolution report generation module generates an execution strategy based on the acquired abnormal mutation index AEI, evolution acceleration factor EAF and risk level.

2. The intelligent engineering quality and safety risk assessment system according to claim 1, characterized in that: The time series data acquisition and preprocessing module includes a parameter acquisition and fitting unit and a data preprocessing unit; The parameter acquisition and fitting unit collects engineering quality data through sensors deployed at the engineering structure location, including foundation secondary settlement increment Adz, crack width micro-drift value Acw, stress rebound frequency deviation value Are and component inclination gradient Asp, and fits them into the original data set W; The secondary settlement increment Adz of the foundation is acquired through multi-point displacement meters and fiber optic displacement sensors: The crack width micro-drift value Acw is acquired through the crack meter and the resistive crack width sensor: The stress rebound frequency deviation value Are is acquired through the stress wave frequency sensor and the high-frequency structural response sensor: The component inclination gradient Asp is acquired through MEMS inclination sensor and fiber optic inclinometer: The data preprocessing unit cleans and normalizes the original data set W to obtain the quality data set ZW; Cleaning is done by identifying outliers in the original dataset W using adjacent transition determination and local linear fitting residual thresholding; Normalization processing is performed by using the maximum and minimum normalization method to process the data in the original data set W to obtain the quality data set ZW; ； Where ZWo represents the o-th data item in the quality dataset ZW, Wo represents the o-th data item in the original dataset W, minWo represents the valley value of the o-th data item in the original dataset W, and maxWo represents the peak value of the o-th data item in the original dataset W.

3. The intelligent engineering quality and safety risk assessment system according to claim 2 is characterized by: The data variation analysis module extracts the local variation intensity of each parameter in the quality data set ZW and obtains the coefficient of variation Ψ, including the coefficient of variation of the foundation secondary settlement increment ΨAdz, the coefficient of variation of the crack width micro-drift value ΨAcw, the coefficient of variation of the stress rebound frequency deviation value ΨAre, and the coefficient of variation of the component inclination gradient rate ΨAsp. The acquisition steps are as follows; S1. Set a fixed period T and construct a multi-scale nested window: T1=T, T2=2T, T3=0.5T; S2. For the oth data item ZWo in the quality data set ZW, within a fixed period T1, calculate and obtain the sliding variation coefficient Ψo; The sliding variation coefficient Ψo is obtained by the following formula: ； Where Ψo(T1) represents the sliding coefficient of variation of the o-th data item in the quality data set ZW within a fixed period T1, σZWo(T1) represents the standard deviation of the o-th data item in the quality data set ZW within a fixed period T1, μZWo(T1) represents the mean of the o-th data item in the quality data set ZW within a fixed period T1, and eo represents a non-zero constant; S3, by introducing the variation enhancement factor VAF, the sensitivity of different trend states is improved; The variation enhancement factor VAF is obtained by the following formula: ； Where ln represents the logarithmic function, Ψo(T2) represents the sliding coefficient of variation of the o-th data item in the quality data set ZW within a fixed period T2, and Ψo(T3) represents the sliding coefficient of variation of the o-th data item in the quality data set ZW within a fixed period T3; S4, obtaining the coefficient of variation Ψ; Ψ=Ψo(T1)*VAF; Where Ψ represents the coefficient of variation.

4. The intelligent engineering quality and safety risk assessment system according to claim 3 is characterized by: The abnormal mutation index fusion module performs nonlinear structural aggregation on the acquired foundation secondary settlement increment variation coefficient ΨAdz, crack width micro-drift value variation coefficient ΨAcw, stress rebound frequency deviation value variation coefficient ΨAre and component inclination gradient variation coefficient ΨAsp to obtain the abnormal mutation index AEI; The abnormal mutation index AEI is obtained by the following formula: ； Where ln represents the exponential function, A represents the nonlinear enhancement factor, and arctan represents the inverse tangent function; Compare the obtained abnormal mutation index AEI with the preset abnormal mutation threshold Tyc to obtain the mutation status; The mutation status is obtained by matching: When the abnormal mutation index AEI is less than the abnormal mutation threshold Tyc*0.5, it indicates that the mutation status is stable and the monitoring is normal; When the abnormal mutation threshold Tyc*0.5≤abnormal mutation index AEI<abnormal mutation threshold Tyc, it indicates a reversible abnormality, local intensified monitoring, and trend analysis is initiated; When the abnormal mutation threshold Tyc≤abnormal mutation index AEI<abnormal mutation threshold Tyc*1.5, it indicates a precursor of synergistic mutation and issues an early warning; When the abnormal mutation threshold Tyc*1.5≤abnormal mutation index AEI, it indicates a high-risk state of multi-source structural mutation and the clinical inspection mechanism is activated.

5. The intelligent engineering quality safety risk assessment system according to claim 4 is characterized by: The driving factor identification module includes an acceleration identification unit and a mutation driving point determination unit; The acceleration recognition unit constructs the evolution acceleration a by extracting the second-order change characteristics of the abnormal mutation index AEI; The evolution acceleration a is obtained by the following formula: ； Wherein, AEI(t) represents the abnormal mutation index at time t, AEI(t+1) represents the abnormal mutation index at time t+1, AEI(t-1) represents the abnormal mutation index at time t-1, and Δt represents the time interval; By introducing the risk growth inertia coefficient GX, the obtained evolution acceleration a is corrected to obtain the evolution acceleration factor EAF; The risk growth inertia coefficient GX is obtained by the following formula: ； Wherein, AEI(t-3) represents the abnormal mutation index at time t-3, ae represents a positive number, and V(t) represents the first-order derivative of the abnormal mutation index AEI; The evolution acceleration factor EAF is obtained by multiplying the risk growth inertia coefficient GX and the evolution acceleration a.

6. The intelligent engineering quality safety risk assessment system according to claim 5, characterized in that: The mutation driving point determination unit calculates the evolution threshold Teaf according to the evolution fluctuation of the evolution acceleration factor EAF; The evolution threshold Teaf is obtained by the following formula: ； Where μEAF represents the mean of the evolution acceleration factor, σEAF represents the standard deviation of the evolution acceleration factor, and af represents the sensitivity coefficient; Compare the evolution acceleration factor EAF with the evolution threshold Teaf to determine the risk status; The risk status is obtained in the following ways: When the evolution acceleration factor EAF ≤ the evolution threshold Teaf, it indicates a normal stable state; When the evolution acceleration factor EAF is greater than the evolution threshold Teaf, it indicates a risk state, hidden dangers appear, and an early warning is issued.

7. The intelligent engineering quality and safety risk assessment system according to claim 6, characterized in that: The early warning threshold dynamic determination module includes a joint index construction unit and a risk level classification unit; The joint index construction unit jointly constructs the acquired evolution acceleration factor EAF and abnormal mutation index AEI to obtain the joint index ARIA; The combined indicator ARIA is obtained by the following formula: ； Where r1 represents the risk intensity adjustment factor, and r2 represents the risk acceleration adjustment factor.

8. The intelligent engineering quality safety risk assessment system according to claim 7, characterized in that: The risk level division unit extracts all values ​​of the joint indicator ARIA from the historical period to form a set Har; The threshold value THr is obtained by using the mean method and standard deviation method to extract the threshold value of the set Har; The index threshold THr was obtained by the sum of the mean of the combined index ARIA and 0.5 times the standard deviation of the combined index ARIA; Compare the combined indicator ARIA with the indicator threshold THr to obtain the risk level; The risk level is obtained by matching: When the combined indicator ARIA is less than 0.5*indicator threshold THr, it indicates the first risk level and safe state; When 0.5*indicator threshold THr≤combined indicator ARIA<indicator threshold THr, it indicates the second risk level and a controllable abnormality occurs; When the indicator threshold THr ≤ combined indicator ARIA < 1.5*indicator threshold THr, it indicates the third risk level, and a precursor to mutation occurs; When 1.5*index threshold THr≤combined indicator ARIA, it indicates the fourth risk level, an abnormally high-risk state.

9. The intelligent engineering quality safety risk assessment system according to claim 8, characterized in that: The trend evolution report generation module identifies the trend of the time series composed of the abnormal mutation index AEI and the evolution acceleration factor EAF, and constructs the abnormal mutation index AEI time series curve and the evolution acceleration factor EAF acceleration curve; Abnormal mutation index AEI time series curve: Draw the abnormal mutation index AEI curve according to the timestamp sequence, and mark the peak point, inflection point and risk level segmented color band in the graph; The acceleration curve of the evolution acceleration factor EAF includes the length of the continuous rising section of the evolution acceleration factor EAF and the time interval from the peak value of the evolution acceleration factor EAF to the mutation point; Generate an execution strategy based on the abnormal mutation index AEI time series curve and evolution acceleration factor EAF acceleration curve, combined with the risk level; The execution strategy is matched and obtained in the following ways: When it is at the first risk level, it means the status is stable and the execution strategy is: normal inspection and periodic sampling; When it is at the second risk level, it means that there is a disturbance but it has not evolved continuously. The execution strategy is to increase the monitoring frequency and start the local trend tracking algorithm; When it is at the third risk level, it means that cumulative instability is forming, and the implementation strategy is: preparation of local maintenance plan, structural re-inspection and assessment of whether it affects the function; When it is at the fourth risk level, it indicates critical evolution and imminent mutation, and the implementation strategy is: structural unloading, personnel evacuation and suspension of work.

10. A method for intelligent assessment of engineering quality and safety risks, applied to an intelligent assessment system for engineering quality and safety risks according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: The time series data acquisition and preprocessing module collects engineering quality data through sensors, fits it into the original data set W, and performs preprocessing to obtain the quality data set ZW; Step 2: The data variation analysis module calculates the local variation degree of the data in the quality data set ZW within a fixed period T and obtains the coefficient of variation Ψ; Step 3: The abnormal mutation index fusion module fuses the obtained coefficient of variation Ψ to construct the abnormal mutation index AEI; Step 4: The driving factor identification module identifies the acceleration of the abnormal mutation index AEI, determines whether there is a mutation driving point in the time series of the abnormal mutation index AEI, obtains the evolution acceleration factor EAF, and identifies potential hidden dangers; Step 5: The early warning threshold dynamic judgment module performs a joint judgment on the abnormal mutation index AEI and the evolution acceleration factor EAF based on the adaptive dynamic threshold of historical statistics to obtain the risk level; Step 6: The trend evolution report generation module generates an execution strategy based on the acquired abnormal mutation index AEI, evolution acceleration factor EAF and risk level.

Images