Building material multi-parameter intelligent detection and analysis system

Abstract

The invention discloses a building material multi-parameter intelligent detection and analysis system, relates to the technical field of building material detection and monitoring, and is used for solving the problem of early abnormality recognition of a building material. According to the method, the dynamic change trend of the concrete material in the early maintenance stage is effectively described by constructing the multi-parameter state vector and combining the behavior evolution model, and the atypical abnormal behaviors such as early dehydration, structure segregation and reaction stagnation can be accurately recognized. Based on a non-linear classification and risk linkage mechanism, a response strategy matched with a behavior state is quickly generated, the intelligent level of detection and decision making is improved, a particularly introduced emergency triggering and short message quick response mechanism can quickly generate a minimum control instruction and directly reach a construction terminal after a high-risk state is identified, and the safety of the construction terminal is improved. Rapid intervention and action closed loop control are achieved, the recognition precision is higher, the response efficiency is higher, the anti-interference capacity is higher, and the method can be widely applied to concrete intelligent maintenance and structure early-stage quality control scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of building material detection, and more particularly to a building material multi-parameter intelligent detection and analysis system comprising double-row piles and front-support steel pipes. Background Art

[0002] As one of the most commonly used building materials, concrete's early curing phase (typically 0 to 72 hours after pouring) has a crucial impact on its ultimate structural strength, durability, and cracking risk. During this phase, concrete undergoes multiple processes, including cement hydration, temperature rise, initial strength development, and pore structure evolution. Concrete is also highly dynamic and sensitive, influenced by factors such as ambient humidity, external temperature fluctuations, and curing procedures.

[0003] Deficiencies in existing technologies: During the early curing stage of concrete materials, processes such as hydration exotherm, water migration, and initial setting are highly coupled. While existing detection systems can collect multiple parameters such as temperature, moisture content, and resistivity, they often employ separate processing and static threshold judgments. They lack unified modeling of the correlation characteristics between multiple parameters, making it difficult to truly reflect the evolution of material behavior. In particular, when abnormal conditions such as early dehydration, reaction stagnation, or segregation and sedimentation occur, the system cannot effectively identify them based on parameter linkage characteristics, posing the risk of misjudgment and omission. Furthermore, field data is susceptible to environmental disturbances, and current systems lack dynamic correction and signal debiasing mechanisms, resulting in insufficient stability and reliability in detection results. Furthermore, most detection results are output as static reports, unable to form a feedback mechanism linked to the construction process, limiting the substantive role of data in quality control. An integrated detection system with state modeling, anomaly identification, and intelligent feedback capabilities is urgently needed. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the following solution is provided to solve the problem of early abnormality identification of building materials in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A multi-parameter intelligent detection and analysis system for building materials, including a concrete data acquisition module, a material state evolution module, a state category analysis module, and an intelligent detection module, wherein the modules are connected via signals;

[0007] The concrete data acquisition module implements a synchronous acquisition mechanism for concrete material state parameters and environmental disturbance data, constructs a time-series observation set, performs spline-based interpolation reconstruction and data synchronization calibration, constructs a differential value based on the difference between similar parameters of the surface and interior, and performs disturbance response compensation to generate a complete physical data set at a unified scale;

[0008] The material state evolution module collects the temperature, moisture content, resistivity, and thermal conductivity of concrete materials, constructs a material state vector containing instantaneous values, rates of change, and spatial gradients, and determines the state evolution characteristics of the concrete material based on the state mapping function. It generates a comprehensive state mapping value at the current moment and describes the behavior trend of the concrete material.

[0009] The state category analysis module divides the material state vector sequence into fixed windows, constructs a labeled behavior sample set, uses a nonlinear classification mapping model to extract dynamic evolution characteristics and outputs behavior categories and confidence scores. Based on risk mapping rules, it determines the material state and outputs the corresponding risk level and intervention recommendations.

[0010] The intelligent detection module triggers the emergency channel based on the identified high-risk status and confidence score, matches the preset response template to generate a minimum control instruction package and sends it to the construction terminal, automatically executes the response action and returns feedback information, and enters the high-frequency monitoring state after execution.

[0011] In a preferred embodiment, a synchronous acquisition mechanism of concrete material state parameters and environmental disturbance data is implemented, and a time series observation set is constructed. Interpolation reconstruction based on spline functions and data synchronization calibration are performed. The specific steps are as follows:

[0012] Synchronous data collection for concrete materials. Concrete materials include two categories: material state data and external environmental disturbance data. Material state data includes temperature, moisture content, resistivity, and thermal conductivity. Environmental disturbance data includes outdoor air temperature, wind speed, relative humidity, sunlight intensity, and surface radiation flux. All sensors perform periodic sampling according to a unified clock to construct a time series observation set.

[0013] Use spline functions to construct local fitters for interpolation.

[0014] In a preferred embodiment, a difference quantity is constructed based on the difference between the surface and internal parameters of the same type and disturbance response compensation is performed to generate a complete physical data set at a unified scale. The specific steps include:

[0015] At each synchronous time point, select the same parameters of the surface and interior of the material and record them separately, and calculate the difference between the two. The difference is obtained by subtracting the internal response from the surface response;

[0016] The external environmental parameter set is introduced as the disturbance enhancement factor to modulate the difference and perform dynamic compensation for environmental interference.

[0017] After interpolation and disturbance response compensation are completed, all physical parameters are mapped to a unified reference benchmark to form a complete physical data set.

[0018] In a preferred embodiment, the material state data of the concrete material is re-collected to construct a material state vector including instantaneous value, change rate and spatial gradient. The specific steps are as follows:

[0019] Establish first-order derivative expressions for the time series of temperature, moisture content, resistivity, and thermal conductivity as the rate of change of the material physical parameters in the time dimension;

[0020] Determine the spatial response difference based on the inhomogeneity of the spatial state inside the material;

[0021] The material physical state, change rate and spatial response difference are integrated into a unified data expression structure and used as the material state vector.

[0022] In a preferred embodiment, the state evolution characteristics of the concrete material are determined based on the state mapping function, and a comprehensive state mapping value at the current moment is generated to describe the behavior trend of the concrete material, including the following steps:

[0023] Construct a state mapping function to establish the transition relationship between states and simulate the evolution of material behavior;

[0024] Collect the state vector sequence of concrete material, and calculate the direction consistency index and change rate factor within the time window for each state vector;

[0025] Through the changes in various state parameters over historical time, the potential signal of the material transitioning from a healthy to an abnormal state is extracted, the corresponding weight of the time point is obtained using the time decay weight function, and the state change is weighted.

[0026] The state vectors of each historical time are mapped and superimposed according to the corresponding weights to obtain the comprehensive state mapping value at the current moment.

[0027] In a preferred embodiment, the material state vector sequence is divided into fixed windows to construct a labeled behavior sample set. The specific steps are as follows:

[0028] The state vector sequence is organized in a sliding window manner. In the continuous state vector sequence, it is divided into time windows of fixed length. Each window represents the state change trajectory of the material at several consecutive time points. The state change trajectory includes the value of the physical parameter and the rate of change of the physical parameter, spatial gradient and response trend characteristics.

[0029] Define a behavior classification system to determine the physical behavior state based on the physical response mechanism of concrete. The behavior categories include: normal hydration, early dehydration, delayed reaction, and segregation settlement.

[0030] Assign a label to each behavior category, and pair the state vector sequence in the sliding window with the corresponding label to form a behavior sample set for supervised learning.

[0031] In a preferred embodiment, a nonlinear classification mapping model is used to extract dynamic evolution features and output behavior categories and confidence scores. The specific steps are as follows:

[0032] Construct a nonlinear classification mapping model, which extracts features and recognizes patterns from the input state vector sequence fragments and outputs the prediction results of the corresponding behavior category;

[0033] The nonlinear classification mapping model design uses a time-aware structure, including a bidirectional recurrent neural network with a gating mechanism, or a graph neural network with state vectors as nodes and state transitions as edges.

[0034] The output of the nonlinear classification mapping model includes a behavior category prediction result and a corresponding confidence score. The confidence score indicates the degree of credibility of the nonlinear classification mapping model for the current judgment.

[0035] In a preferred embodiment, the material status is determined according to the risk mapping rules, and the corresponding risk level and intervention suggestions are output. The specific steps are as follows:

[0036] When the nonlinear classification mapping model completes the identification of the current material state, the material state risk level is determined based on the identification results;

[0037] Set up a risk mapping rule table to map each behavior category to a risk level, including low risk, medium risk, and high risk levels;

[0038] Determine the material status according to the risk mapping rule table, and output the corresponding risk level and intervention suggestions based on the judgment results.

[0039] In a preferred embodiment, the emergency channel is triggered based on the identified high-risk status and confidence score, and a minimum control instruction packet is generated by matching the preset response template and sent to the construction terminal. The response action is automatically executed and feedback information is returned. After execution, the high-frequency monitoring state is entered. The specific steps are as follows:

[0040] Continuously perform behavioral identification on the material state vector sequence. When the risk level is identified as a high-risk state level, determine whether the triggering conditions for entering the emergency state are met.

[0041] When the triggering conditions of the emergency state are met, the emergency channel is activated;

[0042] Match the minimum control instruction package corresponding to the current risk type from the predefined high-risk scenario response template library;

[0043] Once the command is generated, it is immediately broadcast to all response control nodes at the construction site;

[0044] After receiving the control instructions in the instruction package, the control node at the construction site automatically activates the specified response action within the specified time window without manual approval, and enters the high-frequency monitoring state after execution.

[0045] In a preferred embodiment, the behavior of the material state vector sequence is continuously judged. When the risk level is identified as a high-risk state level, it is determined whether the triggering conditions for entering the emergency state are met. The specific steps are as follows:

[0046] The first logical condition for triggering the emergency state is that the currently identified behavior type belongs to a clearly marked structural dangerous behavior. The second logical condition is that the confidence score of the recognition result exceeds the set high confidence threshold. The third logical condition is that the behavior classification results are consistent in consecutive time periods.

[0047] When at least one logical condition is met, the emergency response channel is activated, the regular policy evaluation is skipped, and the lightweight response logic is directly called.

[0048] The technical effects and advantages of the multi-parameter intelligent detection and analysis system for building materials of the present invention are as follows:

[0049] The present invention is aimed at the early curing stage of concrete, and realizes the synchronous collection, dynamic modeling and behavior recognition of multi-source parameters such as temperature, moisture content, resistivity, thermal conductivity, etc., breaking through the technical bottlenecks of existing detection systems such as parameter isolation, response lag and subjective judgment. By constructing a state vector and evolution trajectory model, the system can accurately characterize the internal behavior state of the material and realize real-time identification of atypical anomalies such as early dehydration, reaction delay, segregation and settlement. At the same time, combining nonlinear classification networks with risk-driven strategy scheduling methods, the system can link the construction control terminal to output differentiated response strategies, and has emergency triggering and short message rapid response mechanisms. In emergency situations, control instructions can be issued in seconds and intervention actions can be executed in minutes, significantly improving response efficiency and curing reliability. The overall system has the advantages of strong data fusion capabilities, high recognition accuracy, fast closed-loop response, and adaptability to multiple engineering scenarios. It has wide application value in the fields of intelligent construction and concrete quality monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a structural diagram of a multi-parameter intelligent detection and analysis system for building materials according to the present invention. DETAILED DESCRIPTION

[0051] 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.

[0052] In order to achieve the above objectives, Figure 1 The present invention provides a structural diagram of a multi-parameter intelligent detection and analysis system for building materials, which specifically includes a concrete data acquisition module, a material state evolution module, a state category analysis module, and an intelligent detection module. The modules are connected by signals.

[0053] The concrete data acquisition module implements a synchronous acquisition mechanism for concrete material state parameters and environmental disturbance data, constructs a time-series observation set, performs spline-based interpolation reconstruction and data synchronization calibration, constructs a differential value based on the difference between similar parameters of the surface and interior, and performs disturbance response compensation to generate a complete physical data set at a unified scale;

[0054] The material state evolution module collects the temperature, moisture content, resistivity, and thermal conductivity of concrete materials, constructs a material state vector containing instantaneous values, rates of change, and spatial gradients, and determines the state evolution characteristics of the concrete material based on the state mapping function. It generates a comprehensive state mapping value at the current moment and describes the behavior trend of the concrete material.

[0055] The state category analysis module divides the material state vector sequence into fixed windows, constructs a labeled behavior sample set, uses a nonlinear classification mapping model to extract dynamic evolution characteristics and outputs behavior categories and confidence scores. Based on risk mapping rules, it determines the material state and outputs the corresponding risk level and intervention recommendations.

[0056] The intelligent detection module triggers the emergency channel based on the identified high-risk status and confidence score, matches the preset response template to generate a minimum control instruction package and sends it to the construction terminal, automatically executes the response action and returns feedback information, and enters the high-frequency monitoring state after execution.

[0057] Step 1: Collect and fusion model multi-source environmental data:

[0058] During the early curing stage of concrete materials, the processes such as hydration reaction, heat release, pore change and water migration inside the concrete materials are highly dependent on environmental conditions. Therefore, the multi-source physical parameters of the concrete materials are first collected synchronously, covering two categories: the material state of the concrete itself and the external environmental disturbance. The material state data includes temperature, moisture content, resistivity and thermal conductivity, while the environmental disturbance data includes outside air temperature, wind speed, relative humidity, sunlight intensity and surface radiation flux. All sensors are periodically sampled according to a unified clock to construct a complete time series observation set;

[0059] After the original data is collected, the time consistency of the data of each channel must be determined first. Interpolation reconstruction is performed on the channel data with inconsistent sampling periods or missing records. This process uses a local fitter constructed with a spline function to fit the current missing point with the valid observation values ​​of the adjacent time points before and after. The interpolation process maintains the continuity of the data and the consistency of the physical trend. This interpolation is not only applicable to internal material parameters (such as thermal conductivity or resistivity), but also to environmental sensor channels such as wind speed and humidity.

[0060] For example, in early concrete curing monitoring, the resistivity sensor experienced signal loss during a certain period. Specifically, at time points t1 = 3 and t2 = 7, effective values ​​R1 = 9.8 and R2 = 12.1 were observed, respectively, but the value at the intermediate time point t = 5 was missing.

[0061] To reconstruct the resistivity at the missing point t=5, a local spline fitting function can be constructed. Using the data at points t1 and t2 as endpoint conditions, the following interpolation structure is established using a quadratic spline. Within this interpolation structure, a local function with second-order derivative continuity is constructed so that the derivative and curvature of the fitting function within the time interval are consistent with the physical evolution trend. For example, the expression: R(t)=a(t-t1) 2 +b(t-t1)+R1, where the coefficients a and b can be solved by boundary values ​​and smoothing conditions to ensure that the interpolation point transitions naturally with the previous and next points in terms of physical change trends;

[0062] After solving, the interpolation value reconstructed at t = 5 is: R (5) = 10.9;

[0063] After the basic data are unified, in order to identify the intensity of the impact of external disturbances on the material state signal, it is necessary to construct a differential response model between the surface and interior of the material. This model takes the surface-interior difference as the core variable and introduces external environmental parameters as disturbance enhancement factors to simulate the modulation ability of wind cooling effect, radiation heating or evaporative heat dissipation on surface parameters.

[0064] The specific method is as follows: First, at each synchronous time point t, select the same parameters (such as temperature or resistivity) of the surface and interior of the material and record them as X s (t) and X d (t), calculate the difference between the two, ΔX(t), which is expressed as the surface response minus the internal response, reflecting the instantaneous gradient change in the vertical direction;

[0065] To evaluate the modulation effect of environmental disturbances on this difference, a set of external environmental parameters, including variables such as wind speed, relative humidity, and solar radiation intensity, are introduced as disturbance enhancement factors. Taking into account the non-equivalence of each disturbance factor, they are normalized and then a disturbance contribution factor function is introduced. This disturbance contribution factor function achieves dynamic regulation of the disturbance intensity by mapping the sensitivity of each disturbance factor to changes in surface signals. Taking wind speed as an example, the constructed disturbance enhancement function is as follows;

[0066] The wind speed value is input into an increasing nonlinear mapping structure (such as an exponential ramp function), so that high wind speeds correspond to larger disturbance weights. The mapping result is used as the modulation coefficient. Other disturbance factors such as humidity and irradiance intensity are similarly obtained to obtain corresponding modulation coefficients. Subsequently, the system linearly combines all modulation coefficients and constructs a disturbance enhancement factor, which is used to weight the original difference value to form an enhanced difference response. The difference enhancement response value is equal to the original difference value multiplied by the disturbance enhancement factor.

[0067] The disturbance enhancement factor is designed to be a positive value greater than 1. The larger the value, the more significant the impact of the current environment on the material surface.

[0068] For example, in the temperature channel, the difference between the temperature value on the material surface and the deep temperature value recorded by the internal sensor is calculated, and the rate of change of the surface temperature in a short period of time, as well as the wind speed, outside air temperature and sunshine intensity at this time, are combined as a set of input features to estimate the intensity of the disturbance received by the temperature channel at this moment.

[0069] Quantifying the strength of the disturbance response involves constructing a disturbance response function that uses the temperature difference as the primary input, taking into account its first-order temporal direction and rate of change, introducing wind speed or outside air temperature as a synergistic factor, and modeling the disturbance amplification trend with a double exponential structure. This function has two main parts: one is used to express the amplified response of the surface signal relative to the internal signal, and its response intensity increases exponentially with the absolute value of the temperature difference, reflecting the increased sensitivity of the disturbance under high temperature difference conditions; the other part sets the activation threshold based on the actual change of the disturbance source (such as a sudden change in wind speed) to determine whether the disturbance is sufficient to affect the measurement signal of the material itself;

[0070] For example, a disturbance response function expression is: RD(t) = A exp(β1 |ΔT(t)|) [1 + exp(-β2 (W(t) - θ))] -1 , where RD(t) represents the total disturbance response intensity at time t, which is used to measure the modulation amplitude of surface environmental changes on the observed values ​​of internal material parameters; ΔT(t) is the temperature difference between the surface and internal sensors, reflecting the thermal difference between the external exposed area of ​​the structure and the internal protective layer; the coefficient β1 controls the temperature difference amplification rate; the larger the value, the more sensitive the system is to temperature changes; W(t) is the difference between the wind speed and the wind speed threshold θ; the coefficient β2 controls the steepness of the response; the larger the value, the more aggressive the system is to wind speed changes; exp represents an exponential function with the natural constant e as the base.

[0071] Once the disturbance response function is modeled, the estimated disturbance intensity can be incorporated into the material's parameters as a correction term, completing the dynamic compensation process for environmental disturbances. For example, if a significant disturbance is detected in a temperature channel at the current moment, during the correction phase, this disturbance response is added as an incremental term to the deep-layer temperature value, thereby constructing a corrected temperature with a stable reference baseline. Similarly, moisture content can be identified and corrected for disturbances by coupling the surface-to-internal relationship with the evaporation rate model.

[0072] After the above-mentioned interpolation completion and disturbance response compensation are completed, the system maps all physical parameters to a unified reference benchmark, forming a complete physical data set with continuous time series and consistent scale, providing a standardized input basis for subsequent state vector construction and behavior recognition.

[0073] Step 2: Multi-parameter state vector construction and evolution modeling. The system obtains time series data for multiple physical parameters, including temperature, moisture content, resistivity, and thermal conductivity. To accurately model the behavior of concrete materials, these dispersed physical quantities need to be mapped into a unified state representation structure. This means constructing a multidimensional state vector that can describe the current degree of internal material reaction, structural evolution trends, and potential signs of abnormalities. The specific steps are as follows:

[0074] The material state data of concrete materials are re-collected. In order to obtain the response speed (rate of change) of each material physical parameter of the material state data in the time dimension, it is necessary to establish a first-order derivative expression for the time series of temperature, moisture content, resistivity and thermal conductivity. This derivative is not directly obtained by continuous differentiation, but is approximated based on the rate of change between two adjacent sampling points. Specifically, the numerical difference between the current sampling moment and the previous moment is selected as the parameter change, and the division operation is performed with the corresponding time interval as the denominator, and finally the increment form within unit time is formed. For example, if the current temperature is 36.2℃, the previous sampling point is 35.8℃, and the interval between the two points is 15 minutes, then the temperature change rate is expressed as an increase of 0.0267℃ per minute. Similar methods are used simultaneously for moisture, resistivity and thermal conductivity, and then the change rate vector corresponding to each moment is constructed to reflect the dynamic trend characteristics of the current behavior of the material;

[0075] Capturing the inhomogeneity of the material's internal spatial state, such as the difference in response between the surface and deep layers, is used to indicate abnormal evolution within the structural space. Taking temperature and moisture as typical indicators, to identify the uneven diffusion of heat and moisture within the material's internal space, it is necessary to construct a physical difference term between the surface and internal sensors to express the gradient strength in the vertical direction. The specific operation is as follows:

[0076] The temperature or moisture content readings at the surface and deep layers at the same time are extracted separately, and the difference between the two is calculated to obtain the surface-depth difference. If the difference shows a continuous increase or sudden change, it may reflect the existence of non-uniform hydration, evaporation concentration, or structural segregation within the material. For example, if the surface temperature is continuously higher than the internal temperature and the difference between the two increases, it may indicate the risk of surface cracking; if moisture accumulates significantly at the bottom, it may indicate segregation. On this basis, the system further calculates the gradient difference change between adjacent moments. If the continuous change rate exceeds the safety threshold set by the system, it is considered that a structural response jump has occurred, and the result will be marked in the state vector.

[0077] Based on the above processing, the system integrates the collected material physical states (such as temperature and resistivity), rates of change (such as temperature rise rate and conductivity decay rate), and spatial response differences (such as thermal gradient and water potential gradient) into a unified data representation structure, namely the material state vector. This material state vector is composed of a fixed number of elements, each of which represents a type of physical response mechanism, enabling the system to model complex multi-source responses on a unified scale. Each vector represents the material state within a time segment, and the entire state vector sequence constitutes a trajectory chain of the material behavior evolution over time.

[0078] To simulate the evolution of material behavior, the system introduces a state mapping function to establish transition relationships between states. In this process, the state at any moment is considered to be determined not only by the current physical quantity but also by the previous state. Therefore, the current state vector is used as input, and its direction and rate of change are combined to predict the behavioral state vector at the next moment using nonlinear mapping rules. This state mapping function can integrate multiple response models, such as the temperature-conductivity linkage model and the moisture-heat diffusion hysteresis model, to simulate the state migration process of materials in various behavioral paths, such as continuous hydration, rapid dehydration, or uncontrolled curing.

[0079] The state mapping function is used to characterize the evolution trend of the material over a continuous period of time and to respond to the current state with a score. Its core is to extract the potential signal of the material transitioning from a healthy to an abnormal state by integrating the changes in various state parameters over a period of time. The specific implementation method is as follows:

[0080] First, the state vector sequence of the concrete material {S(t1), S(t2), ..., S(t τ )}, where each state vector S(t τ ) is included at time t τ The standardized values ​​of indicators such as temperature, resistivity, and moisture content are calculated. For each state vector, its directional consistency index and change rate factor within the window are calculated to reflect whether the indicator has a trend of continuously deviating from the normal trajectory;

[0081] The directional consistency index is used to evaluate whether a physical parameter presents a continuous unidirectional change trend in the time series, while the change rate factor is used to quantify the average speed or fluctuation intensity of the change;

[0082] For the directional consistency index, the system first extracts the adjacent change directions of a physical parameter (such as temperature or resistivity) within a time window, that is, comparing the increase and decrease relationship between each pair of consecutive time points. "Increase" is recorded as +1, "decrease" is recorded as -1, and "no significant change" is recorded as 0. In the entire window, the ratio of the sum of all directional values ​​to the total number of times is calculated. If the result is close to +1 or -1, it means that the change direction of the parameter within the window is highly consistent (continuous increase or decrease); if it is close to 0, it means that the change direction frequently reverses and lacks a stable trend.

[0083] For the rate of change factor, the system calculates the instantaneous rate by dividing the difference between the values ​​of each two adjacent time points by the time interval. The system then averages the absolute values ​​of all instantaneous rates within the window to obtain the average rate of change of the parameter within the window. A large rate value indicates that the parameter has undergone drastic changes during that period, possibly corresponding to an active reaction or sudden change. A small value indicates that the parameter changes more slowly.

[0084] The state mapping function is used to characterize the evolution trend of the material in a continuous time period and to respond to the current state. Its core is to extract the potential signal of the material transitioning from a healthy to an abnormal state by integrating the changes in various state parameters over a period of time. The specific implementation steps are as follows: To enhance the response sensitivity to recent evolutionary trends, a time-attenuated weight function is introduced to weight the state changes. Let the total number of historical time periods be τ, the current time be τ, and the corresponding weight of each time point be defined as: ω i =exp(-λ·(τ-i)), where, ω i represents the weight of the i-th time segment, and λ is the weight decay constant, which controls the decay rate of the time effect and is determined according to actual needs;

[0085] Map and superimpose each historical state vector according to the corresponding weight to obtain the comprehensive state mapping value YS(t) at the current moment, which is expressed as: The function f(·) is used to extract non-stationary change factors in the state vector, such as the sign transition rate or instantaneous amplitude slope of a certain indicator, and thus reflect potential signs of instability.

[0086] The comprehensive state mapping value reflects the dynamic evolution of a material from a healthy state to an abnormal state by performing a time-weighted aggregation of state vectors over a historical period. Compared to the state value at a single point in time, the comprehensive state mapping value can express non-stationary characteristics (such as mutations, hysteresis, and reversals) along the continuous evolution path, thereby supporting early identification and early warning of material behavior trends.

[0087] Step 3: Build a material behavior classification and anomaly recognition model:

[0088] After obtaining the multi-parameter state vector sequence of the material state, it is necessary to determine whether the concrete is currently in a typical behavior state, such as normal hydration, early abnormality, water segregation, or reaction stagnation. Because these behaviors have complex temporal evolution characteristics, relying solely on physical quantities at a single time point or their threshold comparisons cannot effectively determine them. Therefore, it is necessary to construct a behavior recognition model based on temporal state evolution characteristics.

[0089] Extract the structural patterns of state changes and map them to a predefined set of material behavior categories to achieve classification recognition and anomaly determination;

[0090] The state vector sequence is organized and processed in a sliding window manner. The specific method is to divide the continuous state vector sequence into time windows of fixed length. Each window represents the state change trajectory of the material at several consecutive time points. The trajectory not only contains the value of the physical parameter, but also includes complex features such as its change rate, spatial gradient and response trend. Therefore, each window can be regarded as a material behavior segment, representing the state evolution process at a certain stage. By sliding the window and repeatedly extracting the material state segment sequence, a training sample set is constructed, so that the subsequent model can learn the dynamic pattern of material behavior based on local time series features;

[0091] A behavioral classification system was defined. Based on the physical response mechanisms of concrete during the early curing phase, several behavioral states with clear physical meanings can be summarized, including but not limited to the following: The first category is normal hydration, characterized by a steady increase in temperature over time, a slow decrease in resistivity, and a constant moisture content; the second category is early dehydration, characterized by a rapid increase in surface temperature and a significant decrease in moisture content, often accompanied by abnormal resistivity fluctuations; the third category is delayed reaction, characterized by a lack of significant hydration heat release and slow resistivity changes, indicating that the cement reaction has not fully initiated; and the fourth category is segregation and settlement, typically characterized by high bottom moisture, minimal temperature fluctuations, and a reverse water potential gradient at the surface. Each behavior category is assigned a clear label, and the state sequence within the sliding window is paired with the corresponding label to form a behavioral sample set for supervised learning.

[0092] Construct a nonlinear classification mapping model, the goal of which is to perform feature extraction and pattern recognition on the input state sequence fragments, and output the prediction results of the corresponding behavior category. The model design adopts a structure with time-series perception capabilities, such as a bidirectional recurrent neural network with a gating mechanism, or a graph neural network with state vectors as nodes and state transitions as edges. The training goal of the model is to automatically form the pattern boundaries implicit in the state sequence by learning the correspondence between multiple historical behavior sequences and their labels. Unlike traditional threshold-based methods, this model does not set fixed discrimination rules, but automatically learns classification boundaries from data through nonlinear classification mapping functions. The model output includes a behavior category prediction result and a corresponding confidence score. The latter indicates the degree of credibility of the model's current judgment, which is used as a reference for subsequent risk level assessment;

[0093] For example, in early-stage concrete curing monitoring, the system continuously collects multidimensional state parameters such as temperature, resistivity, and moisture content, dividing them into state sequence segments within a specific time window (every 5 minutes). Suppose that within a certain period, sensor data show a gradual increase in temperature, a continuous increase in resistivity, and a continuous decrease in moisture content from an initial 12.8% to 7.5%. Traditional threshold-based judgment methods may struggle to respond promptly and accurately to this gradual evolutionary process. This is because while each indicator approaches its limit, it may not reach the set threshold simultaneously, potentially leading to misjudgments or underreporting. Therefore, a nonlinear classification model with time-series awareness, such as a bidirectional gated recurrent neural network structure or a graph neural network model based on graph state evolution, is introduced. This model automatically learns the dynamic characteristic patterns exhibited by each behavior category by training on historical multi-behavior state samples. In this example, the model identifies that the rising resistivity and rapid decrease in moisture content constitute the characteristic boundary of early dehydration behavior, and thus outputs a category prediction of early dehydration with a confidence score of 0.93, indicating a high degree of reliability.

[0094] The basic logic of this nonlinear classification mapping function is as follows: For any state vector sequence segment, the model first feeds it into a feature extraction unit, which identifies information such as local rising and falling trends, lag relationships, and cross-reaction effects based on the changing characteristics of the time dimension. It is then mapped into an embedding space through a multi-layer state transformation network. In this space, the boundaries between behavioral categories are self-organized by the training data and do not rely on any predefined parameter structure or weighted averaging mechanism. Finally, the classification head determines the most likely behavior label based on the position of the embedded vector.

[0095] For example, a sample expression for a categorical mapping function is: Where X = [s1, s2, ..., s t ,...,s T ] represents a state sequence segment of length T, each is the multi-parameter state vector at time t; F(X) represents the processing output of the feature extraction unit for the entire sequence, which can be generated by structures such as bidirectional GRU and graph temporal network to capture dynamic information such as trends, coupling, and delays; QR(·) represents the embedding function, which converts the original sequence features into a high-dimensional embedding vector. The category boundaries in this space are nonlinearly distributed and data-driven; C is the set of all possible behavior categories (such as normal, active hydration, early dehydration, segregation and sedimentation, etc.); W c are the parameters of the linear classification head corresponding to category c, and the input is the embedding vector, which can actually be replaced by MLP or attention head; σ(·) represents the Softmax activation function, which is used to normalize the scores of each category to a probability distribution; The most likely behavior category label output by the final model.

[0096] After the nonlinear classification mapping model completes the identification of the current material state, it is necessary to determine the material state risk level based on the identification results. Set up a risk mapping rule table to correspond each behavior category to a risk level: for example, normal hydration is classified as a low-risk state level and no intervention is required; early dehydration or delayed reaction is a medium-risk state level, prompting construction personnel to strengthen maintenance or re-evaluate construction parameters; and segregation settlement or multi-point abnormal overlap will be determined as a high-risk state level, triggering a system alarm and suggesting the suspension of the construction process in that area. In addition, the system also introduces a confidence weight correction mechanism. When the confidence of the model's recognition of a certain type of anomaly exceeds the specified threshold for multiple consecutive times, even if the behavior category belongs to the medium-risk state level, it will be actively upgraded to a high-risk state level to prevent missed judgments;

[0097] It should be noted that confidence is directly generated from the model's probabilistic prediction results at the output layer. The confidence weight correction mechanism aims to improve the system's response stability to abnormal behavior identification results and the sensitivity of risk assessment. When the model's judgment confidence for a specific behavior category consistently exceeds a set threshold (e.g., 0.9) across multiple consecutive time windows, the system will determine that the behavior not only has a high degree of classification certainty but also exhibits temporal persistence. Therefore, even if the behavior category is originally classified as medium risk in the risk mapping rules, the system will proactively raise its risk level to high risk, triggering a more stringent response strategy. This mechanism introduces a temporal confidence weight factor, assigning an exponentially decaying weight to each confidence judgment. It then integrates the confidence values ​​across multiple consecutive time points to assess the trend and strength of the behavior judgment. When the cumulative confidence weight score exceeds the system's preset response threshold, it is considered that there are significant and credible signs of continued evolution of the abnormal state. The system then initiates a risk escalation operation to prevent missed judgments or delayed responses caused by short-term fluctuations, signal delays, or occasional model errors.

[0098] The output results will include: the current material behavior category, the time point when the behavior change occurs, the model confidence, the corresponding risk level, and the recommended intervention strategy. All results will be pushed to the construction control end or structural health management platform in real time, and can be bound to the BIM model to achieve spatial visualization of abnormal areas.

[0099] In the early curing stage of concrete materials, due to the complex hydration reaction process and drastic changes in the external environment, there are potential abnormal risks such as local shrinkage, segregation or non-uniform evaporation in the structure. Although conventional response strategies can achieve layer-by-layer judgment and scheduling control, there is a response delay problem in extreme cases.

[0100] Step 4: Conduct intelligent linkage feedback and construction response control, formulate differentiated response strategies based on different risk scenarios, and link the construction system to effectively improve the quality control efficiency of the early concrete curing stage. The specific process is as follows:

[0101] The system will continuously perform behavioral identification on the material state vector sequence. Once it identifies that the risk level belongs to the high-risk state level, the system will immediately determine whether the triggering conditions for entering the emergency state are met. The triggering condition is set as a multiple logic joint judgment. The first logical condition is that the currently identified behavior type belongs to a clearly marked structural dangerous behavior, such as early segregation, rapid dehydration or crack precursors; the second logical condition is that the confidence score of the identification result exceeds the set high confidence threshold, indicating that the system has a high degree of certainty in the judgment at the algorithm level; the third logical condition is that the behavior classification results in multiple consecutive time periods are highly consistent, indicating that the abnormal behavior is not sporadic noise, but a manifestation of the stable evolution of the behavior trend. As long as any of the above conditions is met, the system will immediately activate the emergency channel, skip the conventional strategy evaluation, and directly call the lightweight response logic;

[0102] After entering the emergency channel, the system no longer calls the policy optimization function for complex evaluation. Instead, it matches the minimum control instruction package corresponding to the current risk type from a predefined library of response templates for different scenarios. This instruction package has strict field control and only contains four basic fields: abnormal area identification, response action code, instruction validity period, and emergency identification. Through compression design, the system can control each emergency instruction to an extremely small byte length, thereby adapting to low-bandwidth or high-interference communication environments, ensuring that instructions can be issued quickly and stably under harsh working conditions. Once the instruction package is generated, it is immediately broadcast by the system's core control module to all response control nodes on the construction site.

[0103] After receiving the control instructions from the instruction package, the control terminal at the construction site will automatically activate the specified response action within the specified time window without manual approval. For example, if the system receives an abnormality of rapid surface dehydration, it will directly trigger the sprinkler system to start; if the risk of cement segregation is detected, the system will issue a command to suspend pouring or request to switch the mixing time and rate;

[0104] After the command is executed, the on-site control node will immediately transmit a status confirmation message to the system's main control unit, including key parameters such as reception status, execution start time, and action feedback value. If the system does not receive confirmation feedback within the specified time window, it will automatically activate the secondary broadcast mechanism and resend the emergency command using redundant channels until the command is implemented.

[0105] After the emergency response is completed, the system enters a high-frequency monitoring state, collecting multi-parameter states of the material at an increased sampling frequency over a short period of time, recording the trajectory of changes in material behavior after the response. A behavioral response offset model can also be established within the system to compare the evolutionary differences between the continuous states before and after the response, and determine whether the emergency action has achieved the goal of suppressing anomalies. The behavioral response offset model compares multiple change patterns based on physical change trends, such as the reversal of the resistivity change rate, the emergence of a temperature rise platform, and the alleviation of water potential gradients.

[0106] If the system determines that the material status has returned to the normal behavior range, the emergency response is marked as effective; if the status still does not improve significantly, it will immediately enter the regular response process and formulate the next remedial strategy through the full parameter optimization path.

[0107] In summary, the simultaneous acquisition, dynamic modeling, and behavioral recognition of multiple parameters, including temperature, moisture content, resistivity, and thermal conductivity, during the early curing stage of concrete address technical bottlenecks such as parameter isolation, delayed response, and subjective judgment in existing detection systems. By constructing a state vector and evolution trajectory model, the system accurately characterizes the internal behavior of the material, enabling real-time identification of atypical anomalies such as early dehydration, delayed reaction, and segregation settlement. Furthermore, by combining a nonlinear classification network with a risk-driven strategy scheduling approach, the system can link construction control terminals to output differentiated response strategies. It also features an emergency trigger and short message rapid response mechanism. In emergencies, control commands can be issued in seconds and intervention actions can be executed in minutes, significantly improving response efficiency and curing reliability. The overall system boasts strong data fusion capabilities, high recognition accuracy, fast closed-loop response, and adaptability to a wide range of engineering scenarios, promising broad application value in the fields of intelligent construction and concrete quality monitoring.

[0108] The above formulas are all dimensionless and numerical calculations. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0109] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0110] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0111] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0112] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0113] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A multi-parameter intelligent detection and analysis system for building materials, characterized by: It includes concrete data acquisition module, material state evolution module, state category analysis module and intelligent detection module, and each module is connected through signals; The concrete data acquisition module implements a synchronous acquisition mechanism for concrete material state parameters and environmental disturbance data, constructs a time-series observation set, performs spline-based interpolation reconstruction and data synchronization calibration, constructs a differential value based on the difference between similar parameters of the surface and interior, and performs disturbance response compensation to generate a complete physical data set at a unified scale; The material state evolution module collects the temperature, moisture content, resistivity, and thermal conductivity of concrete materials, constructs a material state vector containing instantaneous values, rates of change, and spatial gradients, and determines the state evolution characteristics of the concrete material based on the state mapping function. It generates a comprehensive state mapping value at the current moment and describes the behavior trend of the concrete material. The state category analysis module divides the material state vector sequence into fixed windows, constructs a labeled behavior sample set, uses a nonlinear classification mapping model to extract dynamic evolution characteristics and outputs behavior categories and confidence scores. Based on risk mapping rules, it determines the material state and outputs the corresponding risk level and intervention recommendations. The intelligent detection module triggers the emergency channel based on the identified high-risk status and confidence score, matches the preset response template to generate a minimum control instruction package and sends it to the construction terminal, automatically executes the response action and returns feedback information, and enters the adjustment monitoring state after execution.

2. A multi-parameter intelligent detection and analysis system for building materials according to claim 1, characterized in that: A synchronous acquisition mechanism for concrete material state parameters and environmental disturbance data is established, and a time series observation set is constructed. Interpolation reconstruction based on spline functions and data synchronization calibration are performed. The specific steps are as follows: Synchronous data collection for concrete materials. Concrete materials include two categories: material state data and external environmental disturbance data. Material state data includes temperature, moisture content, resistivity, and thermal conductivity. Environmental disturbance data includes outdoor air temperature, wind speed, relative humidity, sunlight intensity, and surface radiation flux. All sensors perform periodic sampling according to a unified clock to construct a time series observation set. Use spline functions to construct local fitters for interpolation.

3. The multi-parameter intelligent detection and analysis system for building materials according to claim 2, characterized in that: Based on the difference between the surface and internal parameters of the same type, a difference quantity is constructed and disturbance response compensation is performed to generate a complete physical data set at a unified scale. The specific steps include: At each synchronous time point, select the same parameters of the surface and interior of the material and record them separately, and calculate the difference between the two. The difference is obtained by subtracting the internal response from the surface response; The external environmental parameter set is introduced as the disturbance enhancement factor to modulate the difference and perform dynamic compensation for environmental interference. After interpolation and disturbance response compensation are completed, all physical parameters are mapped to a unified reference benchmark to form a complete physical data set.

4. The multi-parameter intelligent detection and analysis system for building materials according to claim 3, characterized in that: The material state data of concrete materials are re-collected to construct a material state vector containing instantaneous value, rate of change, and spatial gradient. The specific steps are as follows: Establish first-order derivative expressions for the time series of temperature, moisture content, resistivity, and thermal conductivity as the rate of change of the material physical parameters in the time dimension; Determine the spatial response difference based on the inhomogeneity of the spatial state inside the material; The material physical state, change rate and spatial response difference are integrated into a unified data expression structure as the material state vector.

5. The multi-parameter intelligent detection and analysis system for building materials according to claim 4, characterized in that: Based on the state mapping function, the state evolution characteristics of the concrete material are determined, and the comprehensive state mapping value at the current moment is generated to describe the behavior trend of the concrete material, including the following steps: Construct a state mapping function to establish the transition relationship between states and simulate the evolution of material behavior; Collect the state vector sequence of concrete material, and calculate the direction consistency index and change rate factor within the time window for each state vector; Through the changes in various state parameters over historical time, the potential signal of the material transitioning from a healthy to an abnormal state is extracted, the corresponding weight of the time point is obtained using the time decay weight function, and the state change is weighted. The state vectors of each historical time are mapped and superimposed according to the corresponding weights to obtain the comprehensive state mapping value at the current moment.

6. The multi-parameter intelligent detection and analysis system for building materials according to claim 5, characterized in that: Divide the material state vector sequence into fixed windows and construct a labeled behavior sample set. The specific steps are as follows: The state vector sequence is organized in a sliding window manner. In the continuous state vector sequence, it is divided into time windows of fixed length. Each window represents the state change trajectory of the material at several consecutive time points. The state change trajectory includes the value of the physical parameter and the rate of change of the physical parameter, spatial gradient and response trend characteristics. Define a behavior classification system to determine the physical behavior state based on the physical response mechanism of concrete. The behavior categories include: normal hydration, early dehydration, delayed reaction, and segregation settlement. Assign a label to each behavior category, and pair the state vector sequence in the sliding window with the corresponding label to form a behavior sample set for supervised learning.

7. The multi-parameter intelligent detection and analysis system for building materials according to claim 6, characterized in that: Use a nonlinear classification mapping model to extract dynamic evolution features and output behavior categories and confidence scores. The specific steps are as follows: Construct a nonlinear classification mapping model, which extracts features and recognizes patterns from the input state vector sequence fragments and outputs the prediction results of the corresponding behavior category; The nonlinear classification mapping model design uses a time-aware structure, including a bidirectional recurrent neural network with a gating mechanism, or a graph neural network with state vectors as nodes and state transitions as edges. The output of the nonlinear classification mapping model includes a behavior category prediction result and a corresponding confidence score. The confidence score indicates the degree of credibility of the nonlinear classification mapping model for the current judgment.

8. The multi-parameter intelligent detection and analysis system for building materials according to claim 7, characterized in that: According to the risk mapping rules, determine the material status and output the corresponding risk level and intervention suggestions. The specific steps are as follows: When the nonlinear classification mapping model completes the identification of the current material state, the material state risk level is determined based on the identification results; Set up a risk mapping rule table to map each behavior category to a risk level, including low risk, medium risk, and high risk levels; Determine the material status according to the risk mapping rule table, and output the corresponding risk level and intervention suggestions based on the judgment results.

9. The multi-parameter intelligent detection and analysis system for building materials according to claim 8, characterized in that: The emergency channel is triggered based on the identified high-risk status and confidence score. The preset response template is matched to generate a minimum control instruction package and sent to the construction terminal. The response action is automatically executed and feedback information is returned. After execution, the system enters the adjustment monitoring state. The specific steps are as follows: Continuously perform behavioral identification on the material state vector sequence. When the risk level is identified as a high-risk state level, determine whether the triggering conditions for entering the emergency state are met. When the triggering conditions of the emergency state are met, the emergency channel is activated; Match the minimum control instruction package corresponding to the current risk type from the predefined response template library; Once the command is generated, it is immediately broadcast to all response control nodes at the construction site; After receiving the control instructions in the instruction package, the control node at the construction site automatically activates the specified response action within the specified time window without manual approval, and enters the adjustment monitoring state after execution.

10. The multi-parameter intelligent detection and analysis system for building materials according to claim 9, characterized in that: Continuously perform behavioral discrimination on the material state vector sequence. When the risk level is identified as a high-risk state level, determine whether the triggering conditions for entering the emergency state are met. The specific steps are as follows: The first logical condition for triggering the emergency state is that the currently identified behavior type belongs to a clearly marked structural dangerous behavior. The second logical condition is that the confidence score of the recognition result exceeds the set high confidence threshold. The third logical condition is that the behavior classification results are consistent in consecutive time periods. When at least one logical condition is met, the emergency response channel is activated.

Images