A dynamic detection and analysis method for early performance of ultra-high performance concrete

Abstract

The present invention relates to the field of concrete detection technology, and in particular to a method for dynamic detection and analysis of the early performance of ultra-high performance concrete. Its technical solution includes establishing a raw material characteristic spectrum library based on multi-physical field coupling, constructing a dynamic detection array based on time series, developing a multimodal data fusion algorithm, establishing a nonlinear coupling effect decomposition model, and implementing dynamic parameter inversion. The present invention realizes dynamic and accurate detection and intelligent regulation of the early performance of ultra-high performance concrete by constructing a "four-dimensional monitoring network-multi-physical field coupling analysis-cross-scale prediction-adaptive regulation" technology chain, solving the technical bottlenecks of traditional methods in parameter coupling, cross-scale correlation and real-time regulation, significantly improving the accuracy of early performance prediction and maintenance efficiency, reducing the risk of cracking, and providing reliable technical support for the application of ultra-high performance concrete in major projects.

Description

Technical Field

[0001] The present invention relates to the technical field of concrete detection, and in particular to a method for dynamic detection and analysis of early performance of ultra-high performance concrete. Background Art

[0002] Ultra-High Performance Concrete (UHPC), as a new building material, has been increasingly used in modern construction engineering due to its excellent properties such as high strength, high durability, and high toughness. As the construction industry continues to increase its requirements for project quality and construction efficiency, accurately understanding the early performance of UHPC is crucial to ensuring project quality and optimizing construction processes. In concrete performance testing, there is a multi-parameter coupling problem:

[0003] Complex material composition: Ultra-high performance concrete (UHPC) typically consists of multiple raw materials, such as cement, silica fume, fly ash, fiber, and admixtures, with complex interactions between these components. Different ratios of these raw materials can have varying effects on the concrete's early performance, and these factors can also be coupled. For example, while fiber incorporation can improve concrete's toughness, it can also affect its fluidity and early hydration reaction, making it difficult to isolate the impact of any single factor on concrete's early performance and hindering testing and analysis.

[0004] Interrelated Performance: Ultra-high Performance Concrete's early strength, shrinkage, and impermeability are interrelated and mutually influential. For example, an increase in early strength may be accompanied by an increase in shrinkage, which in turn may affect the concrete's impermeability. During testing and analysis, it is difficult to study these properties completely independently. The relationships between multiple performance indicators must be comprehensively considered, increasing the complexity and difficulty of testing and analysis.

[0005] In summary, this application proposes a method for dynamic detection and analysis of the early performance of ultra-high performance concrete. Summary of the Invention

[0006] The purpose of the present invention is to propose a dynamic detection and analysis method for the early performance of ultra-high performance concrete to address the problem in the background technology that multi-parameter coupling increases the complexity and difficulty of detection and analysis in concrete performance detection.

[0007] In one aspect, the present application provides a method for dynamic detection and analysis of early performance of ultra-high performance concrete, comprising the following steps:

[0008] Establish a raw material feature spectrum library based on multi-physics field coupling, including three-dimensional material fingerprints of cement matrix phase, mineral admixture phase, and fiber reinforcement phase;

[0009] A dynamic detection array based on time series was constructed. A distributed fiber optic sensor array was embedded inside the concrete specimen, and a multispectral imaging device and an acoustic emission sensor array were arranged outside to form a four-dimensional monitoring network.

[0010] Develop a multimodal data fusion algorithm to align electrical impedance tomography data, hydration heat distribution data, and dielectric property data in time and space, and extract hidden layer coupling features through variational autoencoders;

[0011] A nonlinear coupling effect decomposition model was established, and tensor decomposition technology was used to separate the coupling effects of fiber orientation distribution, interface transition zone development, and hydration reaction process on early performance.

[0012] Implement dynamic parameter inversion and use an improved particle swarm optimization algorithm to simultaneously solve the hydration kinetic parameters, shrinkage stress field distribution, and microscopic pore evolution equations;

[0013] A cross-scale performance prediction model was constructed, and a mapping relationship from nanoscale CSH gel growth to macroscopic compressive strength development was established through a long short-term memory network.

[0014] Design an adaptive control strategy to dynamically adjust the curing environment parameters based on real-time performance deviation analysis, and establish a multi-objective optimization control model based on temperature, humidity, and constraints;

[0015] Deploy a network of edge computing nodes and build a distributed computing architecture at the pouring site to enable local feature extraction of monitoring data and incremental updates of cloud-based models;

[0016] Establish a full life cycle data chain, deeply associate early performance detection data with structural health monitoring data during service, and form a digital twin of performance evolution.

[0017] Optionally, the three-dimensional material fingerprint comprises:

[0018] X-ray diffraction characteristic peak area ratio of cement matrix phase, backscattered electron image grayscale distribution, nanoindentation modulus distribution;

[0019] Atomic force microscopy surface roughness parameters, Raman spectroscopy characteristic peak shift, and zeta potential-particle size distribution curve of the mineral admixture phase;

[0020] Micro-CT three-dimensional reconstruction parameters of fiber-reinforced phase, single-filament tensile stress relaxation curve, and interface bonding-slip constitutive relationship.

[0021] Optionally, the four-dimensional monitoring network includes:

[0022] A Bragg grating array of embedded fiber optic sensors monitors the evolution of internal strain and temperature fields with a spatial resolution of 2 mm;

[0023] The multispectral imaging device collects surface water evaporation and microcrack growth dynamics in the 400-2500nm band at a frequency of 10Hz;

[0024] The acoustic emission sensor array uses waveform manifold learning technology to identify fiber pullout, matrix cracking and interface debonding events.

[0025] Optionally, the multimodal data fusion algorithm includes:

[0026] A spatiotemporal registration coordinate system was established to perform voxel-level registration between the electrical impedance tomography conductivity distribution data and the hydration thermal infrared imaging data;

[0027] Develop a feature-level fusion module that uses a bidirectional gated recurrent unit weighted by an attention mechanism to process multi-source heterogeneous time series data;

[0028] Construct a physical constraint loss function and embed the mass conservation equation and energy balance equation into the neural network training process.

[0029] Optionally, the dynamic parameter inversion includes:

[0030] Establish a parameter sensitivity ranking model and determine the priority of key inversion parameters based on the Morris global sensitivity analysis method;

[0031] Design a hybrid particle swarm optimization strategy and introduce the tabu table mechanism of tabu search into the particle update process to prevent falling into local optimality;

[0032] A multi-objective fitness function is constructed to simultaneously optimize the hydration degree prediction error, shrinkage strain root mean square error, and pore fractal dimension matching.

[0033] Optionally, the cross-scale performance prediction model includes:

[0034] Nanoscale module: CSH gel growth kinetics model based on molecular dynamics simulation;

[0035] Microscale module: micromechanical constitutive model considering the characteristics of the interface transition zone;

[0036] Macroscale module: 3D damage evolution model integrating digital image correlation technology.

[0037] Optionally, the adaptive control strategy includes:

[0038] An environmental response surface model was established, and central composite design experiments were conducted to determine the influence of temperature and humidity interactions on performance development.

[0039] Develop fuzzy predictive control algorithms to dynamically adjust steam curing parameters and external constraints based on rolling horizon optimization;

[0040] Build a safety boundary early warning system to calculate the deviation between the current maintenance path and the ideal performance development trajectory in real time.

[0041] Optionally, the edge computing node network includes:

[0042] Local feature extraction unit: FPGA-based parallel computing architecture realizes real-time wavelet packet decomposition of sensor data;

[0043] Fog computing layer: deploys lightweight convolutional neural networks for key feature pattern recognition;

[0044] Cloud-based collaborative platform: uses a federated learning framework to implement privacy-preserving model updates for multi-project data.

[0045] Optionally, the full life cycle data chain construction method includes:

[0046] Establish spatiotemporal coding rules to hash the timestamps of early detection data with the spatial coordinates of the structure;

[0047] Develop a knowledge graph engine to build a multi-dimensional relationship network of material composition, process parameters, early performance, and long-term durability;

[0048] Design a transfer learning interface to migrate laboratory-scale detection models to digital twins of actual engineering structures.

[0049] Compared with the prior art, this application has at least one of the following beneficial technical effects:

[0050] This invention establishes a four-dimensional monitoring network system, breaking through the limitations of traditional single-point detection and achieving multi-scale, full-process dynamic tracking from nanoscopic to macroscopic. It also proposes a coupling effect decomposition model, successfully solving the problem of multi-parameter interaction and achieving, for the first time, the quantitative separation of the contribution of each component material to early performance.

[0051] The developed adaptive control system enables real-time matching of curing conditions with material performance development, reduces the risk of early cracking, and increases the speed of strength development. The constructed full lifecycle data chain breaks down the barriers between laboratory research and engineering application, reducing maintenance costs of structures during service life.

[0052] The edge-cloud collaborative architecture greatly improves data processing efficiency, meeting the real-time requirements of construction sites while ensuring data security.

[0053] The present invention realizes dynamic and precise detection and intelligent regulation of the early performance of ultra-high performance concrete by constructing a technical chain of "four-dimensional monitoring network-multi-physical field coupling analysis-cross-scale prediction-adaptive regulation", solving the technical bottlenecks of traditional methods in parameter coupling, cross-scale correlation and real-time regulation, significantly improving the accuracy of early performance prediction and maintenance efficiency, reducing the risk of cracking, and providing reliable technical support for the application of ultra-high performance concrete in major projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a flow chart of a method for dynamic detection and analysis of early performance of ultra-high performance concrete;

[0055] Figure 2 This is a principle block diagram of a dynamic detection and analysis system for the early performance of ultra-high performance concrete. DETAILED DESCRIPTION

[0056] The technical solution of the present invention is further described below with reference to the accompanying drawings and specific embodiments.

[0057] Example 1

[0058] like Figure 1 As shown, the present invention proposes a method for dynamic detection and analysis of early performance of ultra-high performance concrete, and each step of the process is described in detail below.

[0059] 1. Establish a raw material feature spectrum library based on multi-physics field coupling, including three-dimensional material fingerprints of cement matrix phase, mineral admixture phase, and fiber reinforcement phase. The three-dimensional material fingerprint includes:

[0060] X-ray diffraction characteristic peak area ratio of cement matrix phase, backscattered electron image grayscale distribution, nanoindentation modulus distribution;

[0061] Backscattered electron images were acquired using a field emission scanning electron microscope (FE-SEM, Hitachi SU8230). The grayscale distribution histogram was analyzed using ImageJ software to quantify the proportion of unhydrated particles. A nanoindenter (Hysitron TI950) was used to perform grid indentation tests (50×50 dot matrix, load 2 mN) to establish the elastic modulus probability density function:

[0062]

[0063] Among them, E i is the single point modulus value, σ E is the standard deviation of the distribution, N is the total number of measurement points;

[0064] Atomic force microscopy surface roughness parameters, Raman spectroscopy characteristic peak shift, and zeta potential-particle size distribution curve of the mineral admixture phase;

[0065] The fly ash surface was scanned by atomic force microscopy (AFM, Bruker Dimension Icon) to extract three-dimensional morphological parameters (Sa = 0.78 μm, Sq = 1.02 μm) where Sa is the arithmetic mean height of surface roughness; Sq is the root mean square height of surface roughness.

[0066] The zeta potential-particle size distribution curve was measured by a laser particle size analyzer (Malvern Mastersizer 3000), and the correlation equation between D50 and early hydration rate was established:

[0067]

[0068] Where k1 is the material constant, D 50 : Average particle size (median particle size) of mineral admixtures; The zeta potential of mineral admixtures characterizes the surface charge characteristics of particles;

[0069] Micro-CT three-dimensional reconstruction parameters of fiber-reinforced phase, single-filament tensile stress relaxation curve, and interface bonding-slip constitutive relationship.

[0070] Micro-CT (ZEISS Xradia 620) scans the fiber distribution and calculates the orientation tensor:

[0071]

[0072] in, is the direction cosine of the pth fiber, and N is the total number of fibers;

[0073] The stress relaxation curve was obtained by single-filament tensile test (Instron 5967) and fitted with the Prony series model:

[0074]

[0075] Among them, σ(t) is the stress value at time t, σ0 is the initial stress, g i is the weight coefficient of the i-th relaxation term, τ i is the time constant of the i-th relaxation term.

[0076] In this example, three-dimensional material fingerprinting (X-ray diffraction, nanoindentation, and micro-CT) comprehensively quantifies the properties of the cement matrix, mineral admixtures, and fiber reinforcement, providing highly accurate input data. This reduces performance fluctuations caused by raw material batch differences and improves the reliability of concrete mix design. Traditional methods rely on empirical mix design and are unable to accurately quantify the impact of raw material characteristics on performance.

[0077] Second, a time-series-based dynamic detection array was constructed. A distributed fiber optic sensor array was embedded inside the concrete specimen, and a multispectral imaging device and an acoustic emission sensor array were arranged outside to form a four-dimensional monitoring network. The four-dimensional monitoring network included:

[0078] The embedded fiber Bragg grating array monitors the internal strain and temperature field evolution with a spatial resolution of 2mm. The embedded 4×4 fiber Bragg grating array (FBG, wavelength 1528-1562nm) has a spatial resolution of 2mm and a strain measurement accuracy of ±1με.

[0079] The multispectral imaging device collects surface water evaporation and microcrack growth dynamics in the 400-2500nm band at a frequency of 10Hz;

[0080] A multispectral camera (SpecimIQ) acquires images at 10 fps in the 400-1000 nm band, and the evaporation rate is inverted using the NDWI:

[0081]

[0082] Among them, ρ 860 is the reflectivity in the 860nm band, ρ 1240 It is the reflectivity in the 1240nm band.

[0083] Infrared thermal imager (FLIR A655sc) monitored the surface temperature field with a spatial resolution of 0.5°C;

[0084] The acoustic emission sensor array uses waveform manifold learning technology to identify fiber pullout, matrix cracking and interface debonding events. Waveform manifold learning (t-SNE) is used to identify damage types:

[0085] Fiber pullout: frequency peak 80-120kHz, duration >200μs;

[0086] Matrix cracking: Peak frequency 150-250kHz, rise time <50μs. A four-dimensional monitoring network (internal fiber optic sensing + external multispectral imaging + acoustic emission monitoring) achieves full temporal and spatial coverage, capturing dynamic details of early performance evolution. Real-time monitoring of concrete internal strain, temperature fields, and surface moisture evaporation provides data support for early performance control.

[0087] 3. Develop a multimodal data fusion algorithm to align electrical impedance tomography data, hydration heat distribution data, and dielectric property data in time and space, and extract hidden layer coupling features using a variational autoencoder. The multimodal data fusion algorithm includes:

[0088] Establish a spatiotemporal registration coordinate system, and perform voxel-level registration between the conductivity distribution data of electrical impedance tomography and the hydration thermal infrared thermal imaging data; establish a voxel coordinate system (1mm 3 The ERT and infrared data are aligned using the ICP algorithm, with a registration error of <0.3mm.

[0089] Develop a feature-level fusion module that uses a bidirectional gated recurrent unit weighted by an attention mechanism to process multi-source heterogeneous time series data;

[0090] class FusionGRU(nn.Module):

[0091] def__init__(self):

[0092] super().__init__()

[0093] self.attention=MultiHeadAttention(embed_dim=128,num_hea ds=4)

[0094] self.gru=BidirectionalGRU(input_size=256,hidden_size=128)

[0095] self.phys_constraint=PhysicsLoss(alpha=0.5, beta=0.3)

[0096] def forward(self,x1,x2):

[0097] attn_out=self.attention(x1,x2)

[0098] gru_out = self.gru(attn_out)

[0099] loss=self.phys_constraint(gru_out)

[0100] return gru_out,loss

[0101] A physical constraint loss function is constructed to embed the mass conservation equation and energy balance equation into the neural network training process. Fick's second law is introduced into the loss function to constrain water diffusion:

[0102]

[0103] Where C is the water concentration, D is the diffusion coefficient, t is the time, and λ1 is the weight coefficient of the physical constraint term. The Laplace operator represents the second-order spatial derivative. Through physically constrained deep learning models (such as variational autoencoders), precise alignment and fusion of electrical impedance tomography, hydration heat distribution, and dielectric property data are achieved, improving the utilization of multi-source heterogeneous data and enhancing the reliability of early performance predictions.

[0104] 4. Establish a nonlinear coupling effect decomposition model and use tensor decomposition technology to separate the coupling effects of fiber orientation distribution, interface transition zone development, and hydration reaction process on early performance; coupling effect decomposition

[0105] (1) Tensor construction:

[0106] Construct a third-order tensor:

[0107]

[0108] in:

[0109] m = 6 (cement content, water-binder ratio, fiber content, curing temperature, humidity, and degree of restraint);

[0110] n = 5 (compressive strength, shrinkage strain, permeability, porosity, elastic modulus);

[0111] p = 4 (1h, 6h, 24h, 72h);

[0112] (2) Tucker decomposition:

[0113] T=G×1U×2V×3W

[0114] in,

[0115] T: third-order tensor, representing multi-parameter coupling effects;

[0116] G: core tensor, revealing the interaction pattern between parameters;

[0117] UV, W: factor matrix, representing the contribution of material parameters, performance indicators and time dimension respectively;

[0118] × i : Represents the product of tensors in the i-th dimension.

[0119] The core tensor G reveals parameter interaction patterns, while the factor matrix U characterizes the contributions of material parameters. Using tensor decomposition techniques, we isolate the coupled effects of fiber distribution, interfacial transition zone development, and hydration reactions, quantifying the contribution of each factor to performance. This provides a theoretical basis for material design and optimization, guiding formulation adjustments and process improvements.

[0120] 5. Implement dynamic parameter inversion, combining an improved particle swarm optimization algorithm to simultaneously solve the hydration kinetic parameters, shrinkage stress field distribution, and microscopic pore evolution equations; the dynamic parameter inversion includes:

[0121] Establish a parameter sensitivity ranking model and determine the priority of key inversion parameters based on the Morris global sensitivity analysis method;

[0122] Design a hybrid particle swarm optimization strategy, introduce the taboo table mechanism of taboo search into the particle update process to prevent falling into local optimality; (1) Hybrid optimization algorithm:

[0123] The particle update formula introduces the taboo search mechanism:

[0124]

[0125] in, The velocity of the i-th particle at the k+1 iteration; w: inertia weight; c1, c2: learning factors; r1, r2 are random numbers in the range [0, 1]; pbest i : the historical optimal solution of the i-th particle; gbest: the global optimal solution; is the position of the i-th particle in the k-th iteration; η is the weight coefficient of the taboo search term, and TS() is the taboo term to avoid repeated search of historical inferior solutions.

[0126] A multi-objective fitness function is constructed to simultaneously optimize the hydration degree prediction error, shrinkage strain root mean square error, and pore fractal dimension matching. Multi-objective fitness function:

[0127] Fitness=ω1·RMSE hyd +ω2·|Δε sh |+ω3·(1-FD match )

[0128] Among them, the weight coefficient ω i Dynamically adjusted by entropy weight method. Fitness is the fitness value, used to evaluate the quality of the inversion results; ω1, ω2, ω3 are weight coefficients, dynamically adjusted by entropy weight method; RMSE hyd : root mean square error of hydration degree prediction; Δε sh : Prediction deviation of contraction strain; FD match is the matching degree of the pore fractal dimension. Combined with an improved particle swarm optimization algorithm, it simultaneously solves the hydration kinetic parameters, shrinkage stress field, and microscopic pore evolution equations, achieving accurate inversion of complex properties. This allows for real-time prediction of early concrete performance development, providing a scientific basis for curing and regulation.

[0129] 6. Construct a cross-scale performance prediction model, using a long-short-term memory network to establish a mapping relationship from nanoscale CSH gel growth to macroscopic compressive strength development; the cross-scale performance prediction model includes:

[0130] Nanoscale module: CSH gel growth kinetics model based on molecular dynamics simulation;

[0131] Microscale module: micromechanical constitutive model considering the characteristics of the interface transition zone;

[0132] Macroscale Module: A three-dimensional damage evolution model that integrates digital image correlation technology. Using a long short-term memory (LSTM) network, it establishes a mapping relationship from nanoscale CSH gel growth to macroscopic compressive strength development, achieving seamless integration of multi-scale performance evolution. This module accurately predicts early concrete strength development and guides construction scheduling.

[0133] 7. Design an adaptive control strategy to dynamically adjust the curing environment parameters based on real-time performance deviation analysis and establish a multi-objective optimization control model of temperature, humidity, and constraints. The adaptive control strategy includes:

[0134] An environmental response surface model was established, and central composite design experiments were conducted to determine the influence of temperature and humidity interactions on performance development.

[0135] Develop fuzzy predictive control algorithms to dynamically adjust steam curing parameters and external constraints based on rolling horizon optimization;

[0136] Build a safety boundary early warning system to calculate the deviation between the current maintenance path and the ideal performance development trajectory in real time.

[0137] Based on real-time performance deviation analysis, the curing environment parameters (temperature, humidity, and constraints) are dynamically adjusted to achieve real-time matching of curing conditions with material performance development, significantly reducing the risk of early cracking and improving concrete durability.

[0138] 8. Deploy an edge computing node network and build a distributed computing architecture at the pouring site to achieve local feature extraction of monitoring data and incremental updates of cloud models; the edge computing node network includes:

[0139] Local feature extraction unit: FPGA-based parallel computing architecture realizes real-time wavelet packet decomposition of sensor data;

[0140] Fog computing layer: deploys lightweight convolutional neural networks for key feature pattern recognition;

[0141] Cloud Collaboration Platform: This platform utilizes a federated learning framework to enable privacy-preserving model updates for multi-project data. A collaborative "end-edge-cloud" architecture is built on-site to enable local processing of monitoring data and incremental updates of cloud-based models. This improves data processing efficiency, meets real-time requirements, and ensures data security and privacy.

[0142] 9. Establish a full life cycle data chain, deeply associate early performance test data with service life structural health monitoring data, and form a performance evolution digital twin. The full life cycle data chain construction method includes:

[0143] Establish spatiotemporal coding rules to hash the timestamps of early detection data with the spatial coordinates of the structure;

[0144] Develop a knowledge graph engine to build a multi-dimensional relationship network of material composition, process parameters, early performance, and long-term durability;

[0145] Design a transfer learning interface to migrate laboratory-scale detection models to digital twins of actual engineering structures.

[0146] In this embodiment, early performance test data is deeply correlated with in-service structural health monitoring data to form a performance evolution digital twin. This provides a scientific basis for structural health monitoring and maintenance, extending the service life of the project.

[0147] Example 2

[0148] like Figure 2 As shown, this embodiment provides a system for dynamic detection and analysis of early performance of ultra-high performance concrete, including:

[0149] A multi-physics field coupling monitoring module, which includes a distributed fiber optic sensor array, a multispectral imager, an acoustic emission sensor, and an electrical impedance tomography device;

[0150] Edge intelligent processing terminal equipped with programmable logic devices and embedded neural network acceleration chips;

[0151] A cloud-based collaborative analysis platform, deployed with a coupling effect decomposition model library and performance prediction digital twins;

[0152] Dynamic control actuator, including intelligent temperature control protection cover, adaptive restraint device and micro-environment adjustment system;

[0153] Data security interaction gateway, using blockchain technology to achieve tamper-proof storage and trusted sharing of detection data;

[0154] Among them, each module realizes data interaction through the industrial Internet of Things protocol, forming an "end-edge-cloud" collaborative architecture.

[0155] This embodiment provides a dynamic detection and analysis system for the early performance of ultra-high performance concrete (UHPC). Through the collaborative work of multiple modules, it realizes intelligent management of the entire process from data collection, processing, analysis to regulation. The following is a detailed expansion description of the system:

[0156] 1. Multi-physics field coupling monitoring module

[0157] Distributed fiber optic sensor array:

[0158] Using Fiber Bragg Grating (FBG) technology, it monitors the evolution of internal strain and temperature fields in concrete with a spatial resolution of 2mm. It supports multi-channel synchronous acquisition, a sampling frequency of 100Hz, a measurement accuracy of ±1με, and a temperature resolution of 0.1°C. It also has a built-in self-diagnosis function that detects the working status of the sensor in real time to ensure data reliability.

[0159] Multispectral imager: Covering the 400-2500nm band, it collects surface moisture evaporation and microcrack expansion dynamics at a frequency of 10Hz; equipped with a high-precision optical lens, a spatial resolution of 0.1mm, and supports the calculation of multispectral indices (such as NDWI); with a built-in ambient light compensation algorithm to adapt to complex lighting conditions at construction sites.

[0160] Acoustic Emission Sensor Array: Utilizes wide-bandwidth sensors (frequency range 20kHz-1MHz) to capture microscopic damage signals at a 1MHz sampling rate. Equipped with a waveform manifold learning algorithm, it automatically identifies fiber pullout, matrix cracking, and interface debonding events. Supports three-dimensional sound source positioning with a positioning accuracy of ±2mm.

[0161] Electrical impedance tomography device: uses a 16-electrode array to collect the internal conductivity distribution of concrete at a frequency of 10Hz; is equipped with an adaptive regularization algorithm to improve imaging resolution and noise resistance; supports multi-parameter fusion to achieve simultaneous monitoring of moisture distribution and pore structure.

[0162] 2. Edge Intelligent Processing Terminal

[0163] Hardware configuration: Equipped with FPGA (field programmable gate array) and embedded neural network acceleration chip (such as NVIDIA Jetson Nano); supports parallel processing of multi-channel sensor data, with a computing power of 4TOPS (trillion operations per second); built-in large-capacity storage module (maximum support 1TB) to meet long-term data caching needs.

[0164] Software functions: Real-time wavelet packet decomposition: noise reduction and feature extraction of sensor data;

[0165] Lightweight convolutional neural network: realizes key feature pattern recognition (such as crack type classification);

[0166] Local data compression: uses lossless compression algorithm to reduce data transmission bandwidth requirements.

[0167] 3. Cloud-based collaborative analysis platform

[0168] Coupling effect decomposition model library: Contains multi-parameter interaction analytical models based on tensor decomposition; supports online model training and updating to meet the needs of different engineering scenarios.

[0169] Performance prediction digital twin: Builds a cross-scale performance prediction model based on the long short-term memory network (LSTM); supports real-time performance evolution visualization, and provides prediction results for key indicators such as early strength and shrinkage strain.

[0170] Data management and analysis: Use distributed databases (such as MongoDB) to store massive monitoring data;

[0171] Provides multi-dimensional data analysis tools (such as cluster analysis and trend forecasting) and supports user-defined report generation.

[0172] 4. Dynamic control of actuators

[0173] Intelligent temperature-controlled curing cover: adopts PID control algorithm, with temperature control accuracy of ±0.5℃; supports independent temperature control in multiple zones to meet the curing needs of large-volume concrete.

[0174] Adaptive restraint device: equipped with a hydraulic servo system to achieve precise adjustment of the restraint force (adjustment range 0-10MPa); built-in stress feedback mechanism to adjust the restraint conditions in real time to prevent early cracking.

[0175] Microenvironment adjustment system: Integrates humidification, ventilation and heating functions, with a humidity control range of 30%-95% and an accuracy of ±3%; supports remote control and real-time adjustment of maintenance parameters through mobile terminals.

[0176] 5.Data security interaction gateway

[0177] Application of blockchain technology: Adopting alliance chain architecture to achieve tamper-proof storage of detection data; supporting data traceability to ensure the credibility and transparency of monitoring data.

[0178] Secure communication protocol: Adopts industrial IoT protocols (such as MQTT and OPC UA) to ensure the security of data transmission; supports data encryption and identity authentication to prevent unauthorized access.

[0179] 6. End-Edge-Cloud Collaborative Architecture

[0180] Device side (device layer): responsible for data collection and local processing, reducing cloud computing load;

[0181] Supports the function of resuming downloads after network disconnection to ensure data integrity.

[0182] Edge computing layer: It realizes data feature extraction and preliminary analysis to improve system response speed; it supports multi-terminal collaborative work to meet the needs of large-scale construction sites.

[0183] Cloud side (cloud platform): provides high-performance computing resources to support complex model training and optimization; supports multi-project data sharing and collaborative analysis to improve the overall efficiency of the system.

[0184] 7. System workflow

[0185] Data acquisition: The multi-physics field coupling monitoring module collects concrete internal strain, temperature, moisture distribution and acoustic emission signals in real time;

[0186] Edge processing: Edge intelligent processing terminals perform noise reduction, feature extraction, and preliminary analysis on data;

[0187] Cloud analysis: The cloud collaborative analysis platform performs coupling effect decomposition and performance prediction, and generates control instructions;

[0188] Dynamic control: Dynamic control actuators adjust maintenance conditions according to instructions to achieve precise control of early performance;

[0189] Data storage and sharing: The data security interaction gateway stores monitoring data in the blockchain to ensure data security and trusted sharing.

[0190] Through the collaborative work of multiple modules, this system achieves dynamic and accurate detection and intelligent control of UHPC's early performance, solving the technical bottlenecks of traditional methods in parameter coupling, cross-scale correlation and real-time control. It significantly improves the accuracy of early performance prediction and maintenance efficiency, reduces the risk of cracking, and provides reliable technical support for the application of ultra-high performance concrete in major projects.

[0191] Example 3

[0192] Based on Example 2, this embodiment further includes an intelligent diagnosis and early warning module, a system optimization and expansion module, and a system workflow optimization module.

[0193] Intelligent diagnosis and early warning modules include:

[0194] Real-time analysis of monitoring data identifies abnormal performance trends (such as early cracking and delayed strength development); provides graded warnings (mild, moderate, and severe) and targeted control recommendations. Deep reinforcement learning algorithms (such as DQN) are used to build anomaly detection models. Expert knowledge bases are integrated, combining historical data with real-time monitoring data to improve diagnostic accuracy. Multi-dimensional warning indicators (such as strain rate, temperature gradient, and acoustic emission energy) are supported. When abnormal strain rates are detected in a local area, warnings are automatically triggered and maintenance conditions are adjusted. Areas of uneven fiber distribution are identified, guiding subsequent improvements to the construction process.

[0195] The digital twin enhancement function, based on Example 2, enhances the real-time interaction and simulation capabilities of the digital twin; supports virtual reality (VR) and augmented reality (AR) visualization, providing an immersive monitoring experience. The Unity 3D engine is used to build a three-dimensional visualization model to map the internal state of concrete in real time; an integrated physics engine (such as PhysX) is used to simulate the performance evolution process under different curing conditions; and supports multi-user collaborative operations to achieve remote consultation and decision support. The internal strain field distribution of concrete can be viewed through VR equipment to identify potential cracking risk areas; AR technology is used to overlay virtual data on site to guide curing operations.

[0196] Multi-objective optimization and control strategy, based on Example 2, introduces a multi-objective optimization algorithm to achieve intelligent control of curing conditions; comprehensively considers multiple objectives such as strength development, shrinkage control, and energy consumption optimization, and adopts NSGA-II (non-dominated sorting genetic algorithm) to solve the multi-objective optimization problem;

[0197] Construct the objective function:

[0198] Fitness=ω1·f1(strength)+ω2·f2(shrinkage)+ω3·f3(energy)

[0199] Among them, f1, f2, and f3 are the objective functions of strength, shrinkage, and energy consumption respectively, ω i The weight coefficient supports dynamic weight adjustment to meet different project requirements. While ensuring strength development, it optimizes maintenance energy consumption, reduces construction costs, balances shrinkage control and strength development, and reduces the risk of premature cracking.

[0200] In this embodiment, the system optimization and expansion module specifically includes:

[0201] The multi-physics field coupling monitoring module has been expanded to include the following sensor types: ultrasonic sensor: monitors the evolution of the pore structure inside concrete, with a frequency range of 50kHz-1MHz; microwave humidity sensor: measures the moisture distribution inside concrete in real time, with an accuracy of ±0.5%.

[0202] Data fusion optimization: Introduce the Kalman filter algorithm to improve the accuracy of multi-source data fusion; develop an adaptive sampling strategy to dynamically adjust the sampling frequency according to performance evolution.

[0203] Edge intelligent processing terminal upgrades specifically include:

[0204] Hardware upgrade: Adopting the new generation of AI chips (such as Huawei Ascend 910), computing power is increased to 16TOPS;

[0205] Add 5G communication module to support high-speed data transmission and low-latency control.

[0206] Software function expansion: New anomaly detection algorithm supports localized real-time warning; integrated data encryption module ensures edge-side data security.

[0207] The cloud-based collaborative analysis platform has been enhanced to support comparative analysis of multi-project data and identify common patterns and individual differences; it also provides intelligent report generation tools to automatically generate inspection reports and regulatory recommendations.

[0208] Performance optimization: Adopt distributed computing frameworks (such as Spark) to improve the efficiency of large-scale data processing; introduce model compression technology to reduce cloud computing resource consumption.

[0209] It is worth noting that the system workflow optimization module in this embodiment specifically includes the following aspects:

[0210] Data collection and preprocessing: Added ultrasonic and microwave humidity sensor data to enrich monitoring dimensions;

[0211] Use Kalman filter algorithm to improve data quality.

[0212] Edge intelligent processing: Added anomaly detection function to achieve localized real-time warning; supports 5G high-speed transmission to improve data upload efficiency.

[0213] Cloud-based analysis and decision-making: Enhance digital twin functionality and support VR / AR visualization; use multi-objective optimization algorithms to generate optimal control strategies.

[0214] Dynamic control and feedback: Dynamically adjust maintenance conditions based on multi-objective optimization results; provide real-time feedback on control effects to optimize subsequent decisions.

[0215] This embodiment further enhances the system's intelligence and practicality by adding intelligent diagnosis and early warning modules, digital twin enhancements, and multi-objective optimization and control strategies. Deep reinforcement learning algorithms boost anomaly recognition accuracy to over 95%. VR / AR visualization is supported, providing immersive monitoring and decision support. Multi-objective optimization algorithms reduce maintenance energy consumption by 20%-30% while maintaining performance.

[0216] The above specific embodiments are merely several optional embodiments of the present invention. Based on the technical solutions of the present invention and the relevant inspirations of the above embodiments, those skilled in the art may make various alternative improvements and combinations to the above specific embodiments.

Claims

1. A method for dynamic detection and analysis of early performance of ultra-high performance concrete, characterized in that: The following steps are involved: Establish a raw material feature spectrum library based on multi-physics field coupling, including three-dimensional material fingerprints of cement matrix phase, mineral admixture phase, and fiber reinforcement phase; A dynamic detection array based on time series was constructed. A distributed fiber optic sensor array was embedded inside the concrete specimen, and a multispectral imaging device and an acoustic emission sensor array were arranged outside to form a four-dimensional monitoring network. Develop a multimodal data fusion algorithm to align electrical impedance tomography data, hydration heat distribution data, and dielectric property data in time and space, and extract hidden layer coupling features through variational autoencoders; A nonlinear coupling effect decomposition model was established, and tensor decomposition technology was used to separate the coupling effects of fiber orientation distribution, interface transition zone development, and hydration reaction process on early performance. Implement dynamic parameter inversion and use an improved particle swarm optimization algorithm to simultaneously solve the hydration kinetic parameters, shrinkage stress field distribution, and microscopic pore evolution equations; A cross-scale performance prediction model was constructed, and a mapping relationship from nanoscale CSH gel growth to macroscopic compressive strength development was established through a long short-term memory network. Design an adaptive control strategy to dynamically adjust the curing environment parameters based on real-time performance deviation analysis, and establish a multi-objective optimization control model based on temperature, humidity, and constraints; Deploy a network of edge computing nodes and build a distributed computing architecture at the pouring site to enable local feature extraction of monitoring data and incremental updates of cloud-based models; Establish a full life cycle data chain, deeply associate early performance detection data with structural health monitoring data during service, and form a digital twin of performance evolution.

2. The method for dynamic detection and analysis of early performance of ultra-high performance concrete according to claim 1, characterized in that: The three-dimensional material fingerprint comprises: X-ray diffraction characteristic peak area ratio of cement matrix phase, backscattered electron image grayscale distribution, nanoindentation modulus distribution; Atomic force microscopy surface roughness parameters, Raman spectroscopy characteristic peak shift, and zeta potential-particle size distribution curve of the mineral admixture phase; Micro-CT three-dimensional reconstruction parameters of fiber-reinforced phase, single-filament tensile stress relaxation curve, and interface bonding-slip constitutive relationship.

3. The method for dynamic detection and analysis of early performance of ultra-high performance concrete according to claim 1, characterized in that: The four-dimensional monitoring network includes: A Bragg grating array of embedded fiber optic sensors monitors the evolution of internal strain and temperature fields with a spatial resolution of 2 mm; The multispectral imaging device collects surface water evaporation and microcrack growth dynamics in the 400-2500nm band at a frequency of 10Hz; The acoustic emission sensor array uses waveform manifold learning technology to identify fiber pullout, matrix cracking and interface debonding events.

4. The method for dynamic detection and analysis of early performance of ultra-high performance concrete according to claim 1, characterized in that: The multimodal data fusion algorithm includes: A spatiotemporal registration coordinate system was established to perform voxel-level registration between the electrical impedance tomography conductivity distribution data and the hydration thermal infrared imaging data; Develop a feature-level fusion module that uses a bidirectional gated recurrent unit weighted by an attention mechanism to process multi-source heterogeneous time series data; Construct a physical constraint loss function and embed the mass conservation equation and energy balance equation into the neural network training process.

5. The method for dynamic detection and analysis of early performance of ultra-high performance concrete according to claim 1, characterized in that: The dynamic parameter inversion includes: Establish a parameter sensitivity ranking model and determine the priority of key inversion parameters based on the Morris global sensitivity analysis method; Design a hybrid particle swarm optimization strategy and introduce the tabu table mechanism of tabu search into the particle update process to prevent falling into local optimality; A multi-objective fitness function is constructed to simultaneously optimize the hydration degree prediction error, shrinkage strain root mean square error, and pore fractal dimension matching.

6. The method for dynamic detection and analysis of early performance of ultra-high performance concrete according to claim 1, characterized in that: The cross-scale performance prediction model includes: Nanoscale module: CSH gel growth kinetics model based on molecular dynamics simulation; Microscale module: micromechanical constitutive model considering the characteristics of the interface transition zone; Macroscale module: 3D damage evolution model integrating digital image correlation technology.

7. The method for dynamic detection and analysis of early performance of ultra-high performance concrete according to claim 1, characterized in that: The adaptive control strategy includes: An environmental response surface model was established, and central composite design experiments were conducted to determine the influence of temperature and humidity interactions on performance development. Develop fuzzy predictive control algorithms to dynamically adjust steam curing parameters and external constraints based on rolling horizon optimization; Build a safety boundary early warning system to calculate the deviation between the current maintenance path and the ideal performance development trajectory in real time.

8. The method for dynamic detection and analysis of early performance of ultra-high performance concrete according to claim 1, characterized in that: The edge computing node network includes: Local feature extraction unit: FPGA-based parallel computing architecture realizes real-time wavelet packet decomposition of sensor data; Fog computing layer: deploys lightweight convolutional neural networks for key feature pattern recognition; Cloud-based collaborative platform: uses a federated learning framework to implement privacy-preserving model updates for multi-project data.

9. The method for dynamic detection and analysis of early performance of ultra-high performance concrete according to claim 1, characterized in that: The full life cycle data chain construction method includes: Establish spatiotemporal coding rules to hash the timestamps of early detection data with the spatial coordinates of the structure; Develop a knowledge graph engine to build a multi-dimensional relationship network of material composition, process parameters, early performance, and long-term durability; Design a transfer learning interface to migrate laboratory-scale detection models to digital twins of actual engineering structures.

Images