A humidity optimization control method for the curing process of foamed concrete components

Abstract

This invention discloses a method for controlling the curing humidity of foamed concrete, comprising the following steps: Step S1: data acquisition, data preprocessing, and model initialization; Step S2: a) preliminary adjustment of the process noise covariance matrix Q in Step S1 based on the statistical characteristics of the humidity prediction residual; b) further adjustment of the process noise covariance matrix Q in Step S1 based on the spatial gradient of the humidity prediction residual; Step S3: humidity estimation and curing humidity control based on adaptive Kalman filtering. This invention relates to the field of data processing technology. The beneficial effect of this invention is that the humidity control method can be combined with control strategies such as model predictive control, utilizing humidity prediction models and optimization algorithms to achieve optimized control of the spraying area, spraying volume, and spraying time, thereby improving the curing quality of foamed concrete, reducing the risk of cracking, and ensuring its performance and durability.

Description

Technical Field

[0001] This invention relates to the field of concrete humidity control, and in particular to a method for controlling the humidity during the curing of foamed concrete. Background Technology

[0002] The performance of concrete is closely related to its curing process. Sufficient moisture is essential for the cement hydration reaction during concrete hardening, and is crucial for the development of concrete strength and durability. Foamed concrete, due to its porous structure, is even more sensitive to humidity changes. Too low humidity during curing can lead to rapid water loss, causing plastic shrinkage cracking, reduced strength, and decreased durability; too high humidity can cause the accumulation of hydration products on the surface, affecting appearance quality and even delaying the hardening process. Precise control of humidity during the curing process of foamed concrete is of great significance for ensuring its engineering quality.

[0003] Traditional concrete curing methods, such as watering and covering, rely heavily on experience and manual operation, making precise humidity control difficult and prone to causing localized over-wetting or dryness. In recent years, with the development of sensor and automatic control technologies, some automatic spray curing systems based on sensor feedback have emerged. These systems monitor the humidity of the concrete surface in real time and automatically control the start and stop of the spraying device, thus improving the accuracy and efficiency of humidity control to a certain extent.

[0004] To more accurately describe the spatiotemporal evolution of the humidity field inside concrete, some studies have applied Fick's second law to model the humidity diffusion process in concrete. Humidity diffusion models based on Fick's second law can reflect the changes in humidity over time and space. Meanwhile, to estimate the humidity distribution inside concrete using limited sensor measurements, state estimation algorithms such as the Kalman filter have also been introduced into concrete humidity estimation. The Kalman filter is a recursive estimation algorithm that can fuse model predictions and sensor measurements to provide an optimal estimate of the system state. Summary of the Invention

[0005] The purpose of this invention is to solve the above problems by designing a method for controlling the curing humidity of foamed concrete.

[0006] The technical solution of the present invention to achieve the above objectives is a method for controlling the curing humidity of foamed concrete, comprising the following steps:

[0007] Step S1: Data acquisition, data preprocessing, and model initialization;

[0008] Step S2:

[0009] a. Based on the statistical characteristics of the humidity prediction residuals, the process noise covariance matrix in step S1 during model initialization is analyzed. Make initial adjustments;

[0010] b. Based on the spatial gradient of the humidity prediction residual, the process noise covariance matrix of the model initialization in step S1 is... Further adjustments will be made;

[0011] step Humidity estimation and maintenance humidity control based on adaptive Kalman filtering.

[0012] Furthermore, in step S1, humidity data is acquired, ambient temperature and humidity are measured by humidity sensors arranged inside and on the surface of the foamed concrete component, and recorded simultaneously over time to form time series data.

[0013] Data preprocessing:

[0014] The collected data is processed as follows: outliers are detected and obviously erroneous data is removed, and suspicious data is corrected; the moving average filtering method is used to smooth the data and reduce the impact of random noise.

[0015] The following operations are performed during model initialization:

[0016] Define state vector The foamed concrete component is divided into a three-dimensional mesh, and the state vector is... Includes the humidity value for each grid node:

[0017] ;

[0018] in, For the first line, number Column, No. The humidity value of the grid nodes in the layer. , , They are respectively The number of grid nodes in the direction;

[0019] Construct the state transition matrix Based on the three-dimensional humidity diffusion equation, the state transition matrix is ​​obtained by discretization using the finite difference method. Considering the humidity-dependent diffusion coefficient Optionally, the effect of heat release during hydration can be considered, and the boundary conditions can be adjusted accordingly. After correction, using the central difference scheme, the discretized form of the three-dimensional humidity diffusion equation is as follows:

[0020] ;

[0021] in, : Time of the first Humidity values ​​of each grid node; Time step; : respectively Grid spacing in the direction; ;

[0022] The state transition matrix can be obtained by combining the discretized equations of all grid nodes. It is a sparse matrix whose non-zero elements are determined by the above difference scheme and boundary conditions;

[0023] Constructing the observation matrix Mapping state vectors to sensor measurements. The number of rows equals the number of sensors, the number of columns equals the number of state variables, and each row corresponds to one sensor;

[0024] Initial state estimate Use a humidity sensor to measure the humidity distribution of the component at the initial moment, as an estimate of the initial state;

[0025] Initialize the state estimation covariance matrix Setting it to a large diagonal matrix indicates higher uncertainty in the initial estimate; the diagonal elements can be set as the variance of the initial humidity measurements.

[0026] Initialization process noise covariance matrix : Set as a diagonal matrix, where the diagonal elements represent the initial variance of each state variable. The diagonal elements are usually set to a large value;

[0027] Initialize the measurement noise covariance matrix Set according to the accuracy specifications of the humidity sensor. It is also a diagonal matrix, with the diagonal elements representing the measurement noise variance of each sensor.

[0028] Furthermore, in step S2, a) based on the statistical characteristics of the humidity prediction residual, the process noise covariance matrix in step S1 is... The specific steps for making initial adjustments are as follows:

[0029] Based on the statistical properties of humidity prediction residuals The matrix undergoes initial adjustments to determine the humidity prediction residual, which is the difference between sensor measurements and model predictions. The non-Gaussianity of the residual sequence is quantified by calculating its kurtosis and skewness. Kurtosis reflects the peak of the distribution, and skewness reflects its asymmetry. Based on kurtosis and skewness, a non-Gaussianity index is constructed, referred to as: and according to Size dynamic adjustment The matrix is ​​used to calculate the humidity prediction residual, which is the difference between the sensor measurement and the model prediction.

[0030] ;

[0031] in, for Humidity prediction residual vector at time step; for A vector of humidity sensor measurements at any given time; The observation matrix; for Time based The predicted state vector at time step;

[0032] For each sensor Calculate its residual sequence kurtosis within the sliding window and skewness The sliding window size is That is, considering the past The residuals at each time step, using a sliding window: The sliding window is used to calculate the local statistical properties of the residual sequence, namely kurtosis and skewness. The formulas for calculating kurtosis and skewness are as follows:

[0033] ;

[0034] ;

[0035] in, For sensors exist Kurtosis of the residual sequence at time step; For sensors exist Skewness of the residual sequence at time step; For sensors exist Humidity prediction residual at any given time; The mean of the residuals within the sliding window;

[0036] Construct a non-Gaussianity index based on kurtosis and skewness. :

[0037] ;

[0038] in, for The non-Gaussianity index at time; Number of sensors; and The first Kurtosis and skewness weights for each sensor;

[0039] Constructing optimization factors:

[0040] ;

[0041] in, for The optimization factor at any given time; An adaptive forgetting factor;

[0042] The update formula is:

[0043] ;

[0044] in, Forgetting factor coefficient ; for The initial value can be set to or ;

[0045] Update process noise covariance matrix :

[0046] ;

[0047] in, for The process noise covariance matrix updated at each time step; This is the baseline value for the process noise covariance matrix.

[0048] Furthermore, in step S2, b, based on the spatial gradient of the humidity prediction residual, the process noise covariance matrix in step S1 is... Further adjustments will be made, and the specific steps are as follows:

[0049] In the steps Based on the statistical properties of the humidity prediction residuals, the covariance matrix of kurtosis and skewness against process noise is given. Preliminary adjustments were made to accommodate the non-Gaussian nature of the process noise in the steps. Based on the spatial gradient information of the humidity prediction residual, the process noise covariance matrix is ​​considered. Adjustments are made, and a local anomaly index is constructed based on the spatial gradient of the humidity prediction residual, referred to as: ; Calculate the spatial gradient of the humidity prediction residual. In the three-dimensional case, the sensors are distributed in a three-dimensional grid pattern inside the foamed concrete component. Therefore, for the sensors... ,in (The number of sensors in the three directions, respectively), which in The residual gradient at time step 1 can be expressed as:

[0050] ;

[0051] ;

[0052] ;

[0053] ;

[0054] in, For sensors exist Time residual Gradient in direction; For sensors exist Time residual Gradient in direction; For sensors exist Time residual Gradient in direction; For sensors exist The magnitude of the residual gradient at time step; For sensors exist Humidity prediction residual at any given time; and The sensors are respectively in and Spacing in the direction;

[0055] The expected residual gradient is calculated using the following formula:

[0056] ;

[0057] in, For sensors exist The expected residual gradient magnitude at time step [time]. Represents the expected residual; According to Predicting the state at any time The calculated humidity diffusion coefficient; Calculated based on the predicted state The magnitude of the humidity gradient at any given time is calculated similarly to that of the residual gradient, except that the residual is used in the calculation. Replace with predicted humidity value :

[0058] ;

[0059] ;

[0060] ;

[0061] ;

[0062] Based on the actual residual gradient and the expected residual gradient, a local anomaly index is constructed, referred to as... :

[0063] ;

[0064] in, For sensors exist Local abnormal indicators at any given time; It is a very small positive number;

[0065] Constructing optimization factors :

[0066] ;

[0067] in, for The optimization factor at any given time; The number of effective sensors participating in the calculation; It is an adjustable parameter;

[0068] Update process noise covariance matrix :

[0069] ;

[0070] in, for Time goes through steps Adjusted process noise covariance matrix; for Time goes through steps Adjusted process noise covariance matrix.

[0071] Furthermore, in step S3, based on the humidity estimation results obtained in the aforementioned steps, humidity control is performed during the curing process of foamed concrete. At each time step, the optimal estimate of the internal humidity field of the concrete at the current moment is obtained using an optimized adaptive Kalman filter algorithm to form a humidity estimate. Based on this humidity estimate and combined with the preset humidity control target, it is determined whether spraying is required. The surface humidity estimate determines whether to start the spraying system. When the surface humidity is lower than the set lower limit, spraying is triggered.

[0072] Furthermore, the estimated internal humidity and the calculated humidity gradient information are used to assist in decision-making. When the internal humidity is generally low or there are areas with excessively large local humidity gradients, even if the surface humidity meets the requirements, the spraying strategy can be adjusted appropriately to achieve more precise and forward-looking humidity control.

[0073] A humidity optimization control method for the curing process of foamed concrete components, based on the technical solution of this invention, is provided. This humidity control method can be combined with model predictive control (MRC). By combining control strategies such as humidity prediction models and optimization algorithms, this invention achieves optimized control of the spraying area, spray volume, and spraying time. In this way, the invention effectively solves problems in existing technologies such as insufficient humidity control precision, difficulty in simultaneously addressing surface and internal humidity requirements, and inability to effectively handle local anomalies. This improves the curing quality of foamed concrete, reduces the risk of cracking, and ensures its performance and durability. Attached Figure Description

[0074] Figure 1 This is a schematic diagram of a humidity optimization control method for the curing process of foamed concrete components as described in this invention. Detailed Implementation

[0075] The present invention will now be described in detail with reference to the accompanying drawings, such as... Figure 1 As shown, a method for controlling the curing humidity of foamed concrete includes the following steps: Data acquisition, preprocessing, and model initialization.

[0076] First, data acquisition, data preprocessing, and model initialization are performed.

[0077] For data acquisition, humidity data is obtained through humidity sensors placed inside and on the surface of the foamed concrete components. Sensor types can include resistive, capacitive, and fiber optic sensors. Internal sensors are arranged in a regular grid (three-dimensional) pattern; surface sensors are evenly distributed. The number and spacing of sensors are determined based on the component's size and shape. In addition, ambient temperature and humidity are simultaneously measured for subsequent model calculations and boundary condition determination. Sensor measurements (humidity and temperature) are recorded along with timestamps to form time-series data.

[0078] In terms of data preprocessing, the collected data are processed as follows: outliers are detected and obviously erroneous data is removed, and suspicious data is corrected; the moving average filtering method is used to smooth the data and reduce the impact of random noise; if the sampling times of different sensors or data acquisition systems are not synchronized, time synchronization processing is required to align the data to a unified time reference.

[0079] Regarding model initialization, this invention performs the following operations:

[0080] Define state vector The foamed concrete component is divided into a three-dimensional mesh, and the state vector is... Includes the humidity value for each grid node:

[0081] ;

[0082] in, For the first line, number Column, No. The humidity value of the grid nodes in the layer. , , They are respectively The number of grid nodes in the direction.

[0083] Construct the state transition matrix Based on the three-dimensional humidity diffusion equation, the finite difference method is used ( Discretize the state to obtain the state transition matrix. Considering the humidity-dependent diffusion coefficient Optionally, the effect of heat release from hydration can be considered, and the boundary conditions (absolute humidity, given humidity, convection) can be adjusted accordingly. Make corrections.

[0084] Taking the central difference scheme as an example, the discretized form of the three-dimensional humidity diffusion equation is as follows:

[0085] ;

[0086] in:

[0087] : Time of the first The humidity value of each grid node.

[0088] Time step.

[0089] : respectively Grid spacing in the direction.

[0090] The same applies to other directions.

[0091] The state transition matrix can be obtained by combining the discretized equations of all grid nodes. . It is a sparse matrix whose non-zero elements are determined by the difference scheme and boundary conditions described above.

[0092] Constructing the observation matrix Mapping state vectors to sensor measurements. The number of rows equals the number of sensors, and the number of columns equals the number of state variables (total number of grid nodes). Each row corresponds to one sensor, and the element in that row corresponding to the grid node where that sensor is located is... Other elements are .

[0093] Initial state estimate A humidity sensor is used to measure the humidity distribution of the component at the initial moment, which is used as an estimate of the initial state.

[0094] Initialize the state estimation covariance matrix Setting this to a large diagonal matrix indicates higher uncertainty in the initial estimate. The diagonal elements can be set as the variance of the initial humidity measurements, or empirically.

[0095] Initialization process noise covariance matrix This is set as a diagonal matrix, where the diagonal elements represent the initial variance of each state variable. Due to a lack of prior knowledge, The diagonal elements are usually set to a large value.

[0096] Initialize the measurement noise covariance matrix : Set according to the accuracy specifications of the humidity sensor. It is also a diagonal matrix, with the diagonal elements representing the measurement noise variance of each sensor.

[0097] It should be noted that the model and initialization method used are based on existing technologies, including... The three-dimensional humidity diffusion equation described by the second law, the finite difference method ( Numerical discretization methods and Kalman filter state estimation algorithms. Specifically, this invention is based on... The second law establishes a humidity diffusion model, and the diffusion equation is discretized using a central difference scheme to obtain the state transition matrix. A Kalman filter algorithm is used for humidity state estimation, and the state estimate is updated based on sensor measurements. The initialization parameters of the Kalman filter algorithm, including the initial state estimate, the state estimate covariance matrix, the process noise covariance matrix, and the measurement noise covariance matrix, are all set according to conventional methods.

[0098] step :

[0099] a. Based on the statistical properties of humidity prediction residuals, the process noise covariance matrix is ​​analyzed. Preliminary adjustments were made.

[0100] During the curing of foamed concrete, the spatiotemporal evolution of the humidity field is influenced by a variety of factors. Besides the main humidity diffusion process, there are many complex factors that are difficult to model accurately. These complex factors cause the actual process noise to deviate from the Gaussian white noise distribution assumed by the traditional Kalman filter algorithm, resulting in a decrease in the accuracy of the filter estimation.

[0101] First, the heterogeneity of foamed concrete is a significant factor contributing to the complexity of process noise. The internal pore structure of foamed concrete (such as pore size, porosity, and connectivity) is not completely uniform but exhibits a certain degree of random distribution. This microstructural heterogeneity leads to differences in humidity diffusion capacity at different locations, making it difficult for macroscopic models to accurately describe humidity changes in each tiny region, thus generating prediction biases. These biases are attributed to process noise in Kalman filtering.

[0102] Secondly, the complexity of cement hydration significantly impacts the characteristics of process noise. Cement hydration is a complex physicochemical process involving multiple mineral components and reaction stages, and its rate is nonlinearly influenced by factors such as temperature, humidity, cement mineral composition, and admixtures. The hydration reaction not only consumes water and alters the humidity field but also causes changes in pore structure, further affecting humidity diffusion. Existing models typically employ simplified hydration models, which struggle to fully and accurately describe the dynamic impact of the hydration process on the humidity field, thus complicating process noise.

[0103] Furthermore, the complexity and uncertainty of boundary conditions are also significant sources of process noise. Moisture exchange between the concrete surface and its surrounding environment is influenced by various factors, including wind speed, ambient temperature and humidity, and surface cover. These factors are often difficult to measure or predict accurately and may change over time, making it difficult to precisely determine boundary conditions. This uncertainty affects the humidity field inside the concrete through the boundary, thereby increasing process noise.

[0104] Due to the combined effects of the aforementioned complex factors, actual process noise often does not meet the assumptions of traditional Kalman filtering. Specifically, process noise no longer follows a Gaussian distribution but exhibits non-Gaussianity, meaning the probability distribution is skewed (skewness is not zero) or has a longer tail (kurtosis deviates from 3); the statistical characteristics of process noise (such as variance) change over time, exhibiting time-varying behavior; and due to material heterogeneity and spatial differences in hydration reactions, humidity changes in adjacent areas are correlated, causing process noise to exhibit spatial correlation rather than independence.

[0105] In traditional Kalman filtering, the process noise covariance matrix Typically, a fixed diagonal matrix is ​​used, which fails to reflect the non-Gaussianity, time-varying nature, and spatial correlation of process noise. This leads to a decrease in the estimation accuracy of Kalman filtering and may even cause filter divergence. To address this issue, a method based on the statistical properties of humidity prediction residuals is proposed. The matrix undergoes initial adjustments. The humidity prediction residual, which is the difference between the sensor measurement and the model prediction, contains information that the model did not capture and can reflect the characteristics of process noise.

[0106] Specifically, the degree of non-Gaussianity is quantified by calculating the kurtosis and skewness of the residual sequence. Kurtosis and skewness are commonly used indicators in statistics to measure the degree to which a data distribution deviates from a normal distribution. Kurtosis reflects the peak of the distribution, while skewness reflects the asymmetry of the distribution. Based on kurtosis and skewness, a non-Gaussianity index is constructed (…). ), and according to Size dynamic adjustment Matrix: When A larger value indicates stronger non-Gaussianity of process noise, and increasing it will lead to further problems. Matrix to reduce dependence on sensor measurements; when When the value is small, it indicates that the process noise is close to a Gaussian distribution, and reducing it... The matrix relies more heavily on sensor measurements. In this way, The matrix can adaptively reflect the non-Gaussian nature of process noise, thereby improving the estimation accuracy of Kalman filtering.

[0107] First, calculate the humidity prediction residual, which is the difference between the sensor measurement and the model prediction:

[0108] ;

[0109] in:

[0110] for Humidity prediction residual vector at time (dimension: , (This represents the number of sensors). It indicates the difference between the model's predicted values ​​and the actual sensor measurements. The larger the residual, the less accurate the model's prediction.

[0111] for Humidity sensor measurement vector at time (dimension: It is the humidity value measured directly by a humidity sensor placed in the foamed concrete component.

[0112] The observation matrix (dimension 1) , (This refers to the number of state variables). It describes the relationship between state variables (i.e., the humidity at various points inside the concrete) and sensor measurements. Each row corresponds to a sensor, and the elements of the state variable in that row that correspond to the position of that sensor are: Other elements are .

[0113] for Time based The predicted state vector at time step (dimension 1) It is predicted based on the humidity diffusion model (state transition equation). The humidity value at various points inside the concrete at any given time.

[0114] Subsequently, for each sensor Calculate its residual sequence kurtosis within the sliding window and skewness The sliding window size is (i.e., considering the past) (residuals at each time step).

[0115] Sliding window: A sliding window is used to calculate the local statistical properties (kurtosis and skewness) of the residual sequence, and its size... The choice requires weighing: If the value is too small, the statistical results are easily affected by random noise; If the value is too large, it will be difficult to reflect the time-varying nature of the noise characteristics. Determined based on experience or experimentation.

[0116] The formulas for calculating kurtosis and skewness are as follows:

[0117] ;

[0118] ;

[0119] in:

[0120] For sensors exist The kurtosis of the residual sequence at time points. Kurtosis is a measure of the peakedness of a data distribution. For a normal distribution (Gaussian distribution), the kurtosis is... If the kurtosis is greater than If the kurtosis is less than 0, it indicates that the distribution is more kurtotic than the normal distribution; if the kurtosis is less than 0, it indicates that the distribution is more kurtotic than the normal distribution. If , it means that the distribution is flatter than the normal distribution.

[0121] For sensors exist The skewness of the residual sequence at time step 1. Skewness is a measure of the symmetry of the data distribution. For a normal distribution, the skewness is 1 / 2. If the skewness is greater than If the skewness is less than 0, it indicates that the distribution is right-skewed (long tail on the right); if the skewness is less than 0. If , it indicates that the distribution is left-skewed.

[0122] For sensors exist Humidity prediction residual at any given time.

[0123] This represents the mean residual within the sliding window. It indicates the sensor's... in the past The average residual within each time step.

[0124] Construct a non-Gaussianity index based on kurtosis and skewness. :

[0125] ;

[0126] in:

[0127] for The non-Gaussianity index (scalar) at time points; it comprehensively reflects the degree of non-Gaussianity of all sensor residual sequences. The larger the value, the stronger the overall non-Gaussianity; Number of sensors; and The first The kurtosis and skewness weights of each sensor are set according to the importance or reliability of different sensors. If a sensor is more susceptible to interference, or its measurements have a smaller impact on the overall humidity estimate, its weight can be set to a smaller value. Generally, unless there is a specific reason, the weights of all sensors should be set to a fixed value. .

[0128] Furthermore, construct optimization factors.

[0129] ;

[0130] in:

[0131] for The optimization factor (scalar) at time. It is based on the non-Gaussianity index. Adjusting the size of the process noise covariance matrix .

[0132] This is an adaptive forgetting factor. It is used for smoothing. The changes make the algorithm focus more on recent data, thereby reducing the influence of historical data. The update formula is:

[0133] ;

[0134] in, Forgetting factor coefficient . The closer The greater the influence of historical data, the better. The more gradual the change; The closer The smaller the influence of historical data, the better. More sensitive to changes in current data. You can take a close one The value, such as wait.

[0135] for The initial value can be set to or .

[0136] Finally, update the noise covariance matrix. :

[0137] ;

[0138] in:

[0139] for The process noise covariance matrix updated at each time step. It is an important parameter in the Kalman filter algorithm, reflecting the uncertainty of the model's predictions. The larger the value, the less reliable the model predictions are, and the more the Kalman filter will rely on sensor measurements. The smaller the value, the more reliable the model prediction, and the more the Kalman filter will rely on the model prediction.

[0140] This is the baseline (initial) value for the process noise covariance matrix. It is typically set based on prior knowledge of the model's uncertainties. Without prior knowledge, it will... Set as a diagonal matrix, where the diagonal elements represent the initial variance of each state variable.

[0141] Explanation of the function of each part of the formula:

[0142] By dynamically adjusting the process noise covariance matrix This improves the accuracy of Kalman filtering in estimating the humidity of foamed concrete. The above formula is designed precisely to achieve this goal. First, the humidity prediction residual... The calculation of residuals is fundamental to the entire optimization strategy. Residuals reflect the discrepancies between model predictions and actual observations, primarily stemming from process noise and various complex factors not considered in the model. Therefore, residual analysis can reveal the characteristics of process noise, providing insights for adjusting... The matrix provides the basis.

[0143] Secondly, kurtosis and skewness The introduction of kurtosis is to quantify the non-Gaussianity of the residual sequence. Traditional Kalman filtering assumes that process noise follows a Gaussian distribution, but in practical applications, due to factors such as material heterogeneity and hydration reactions, process noise often exhibits non-Gaussianity. Kurtosis and skewness are commonly used indicators in statistics to measure the degree to which data distribution deviates from a Gaussian distribution. By calculating the kurtosis and skewness of the residual sequence, we can determine whether the process noise is non-Gaussian and the strength of this non-Gaussianity.

[0144] Furthermore, a non-Gaussianity index was constructed. By combining the kurtosis and skewness information from all sensors, a comprehensive non-Gaussianity measure is obtained. This approach avoids the influence of outliers from individual sensors on the overall judgment, improving the stability of the evaluation. The larger the value, the stronger the non-Gaussianity of the process noise and the lower the reliability of the model prediction.

[0145] Based on non-Gaussianity index Optimization factors were designed. . Its function is based on Size dynamic adjustment Matrix. When A larger value indicates stronger non-Gaussianity of the process noise; therefore, increasing the value further is recommended. , thereby increasing This causes Kalman filtering to rely more on model predictions and less on sensor measurements that may contain outliers. When the value is small, it indicates that the process noise is close to a Gaussian distribution. close to , Approaching its benchmark value The Kalman filter operates in a conventional manner.

[0146] In order to The changes are smoother, avoiding issues caused by abnormal data at individual moments. The dramatic fluctuations in the matrix necessitate the introduction of an adaptive forgetting factor. . Able to The historical values ​​are weighted and averaged to make the current value... Not only affected by the current moment The influence is also affected by past moments. The impact of data changes is considered, with data closer to the current time having a greater weight. This improves the algorithm's stability and makes it focus more on recent data, thus better adapting to the time-varying characteristics of process noise.

[0147] Finally, through To adjust The matrix optimizes the Kalman filter algorithm. The adjusted matrix... The matrix can adaptively reflect the non-Gaussian nature of process noise, enabling Kalman filtering to maintain high estimation accuracy when facing complex, non-Gaussian process noise.

[0148] b. Based on the spatial gradient of the humidity prediction residual, the process noise covariance matrix... Further adjustments will be made.

[0149] In the steps The statistical properties (kurtosis and skewness) of the humidity prediction residuals are used to evaluate the process noise covariance matrix. Preliminary adjustments were made to accommodate the non-Gaussian nature of the process noise. However, the steps... The adjustments primarily focus on the temporal characteristics of the residual sequence, neglecting the spatial information of the residuals. During the curing process of foamed concrete, due to the heterogeneity of the material and spatial differences in hydration reactions and boundary conditions, the humidity field often exhibits complex spatial distribution characteristics, with significant humidity gradients existing between adjacent areas. This spatial gradient information is crucial for accurately estimating the humidity field.

[0150] Traditional Kalman filtering assumes that process noise is spatially independent, meaning that process noise at different locations is uncorrelated. This assumption often fails when the humidity field has a significant spatial gradient. For example, if a region experiences abnormal humidity changes due to material defects or localized dryness, this anomaly will affect not only the humidity estimate for that region but also the humidity estimates for its surrounding areas. Ignoring this spatial correlation will compromise the accuracy of Kalman filtering.

[0151] To solve this problem, the present invention provides steps Based on this, further consideration is given to the spatial gradient information of the humidity prediction residual on the process noise covariance matrix. Adjustments are made. Specifically, if the humidity prediction residual gradient for a certain region is significantly greater than the expected gradient calculated based on the humidity diffusion model, it indicates the presence of local anomalies (e.g., material defects, localized drying, abnormal hydration reactions, etc.) that the model has not captured. These local anomalies cause process noise to exhibit spatial correlation, and their intensity is related to the magnitude of the residual gradient.

[0152] Therefore, based on the spatial gradient of the humidity prediction residual, a local anomaly index is constructed ( ). The larger the value, the more the humidity change in the region deviates from the model's expectations, and the more significant the local anomalies. Then, according to... Size, further adjustment Matrix. When A larger value indicates a strong local abnormality that may further increase. The matrix makes the Kalman filter rely more on the spatial smoothing effect of the model predictions, reducing the trust in sensor measurements for that area, thus minimizing the impact of local anomalies on the overall humidity estimate. When A smaller value indicates that the humidity change in the area is consistent with the model's expectations. The matrix remains unchanged, or is slightly reduced. matrix.

[0153] In this way, The matrix can not only reflect the non-Gaussian nature of process noise (step) Furthermore, it can reflect the spatial correlation of process noise, thus more comprehensively adapting to the complex characteristics of the humidity field of foamed concrete, and further improving the estimation accuracy and robustness of Kalman filtering.

[0154] First, the spatial gradient of the humidity prediction residual is calculated. In the three-dimensional case, the sensors are distributed in a three-dimensional mesh pattern inside the foamed concrete component. Therefore, for the sensors... ,in (The number of sensors in the three directions, respectively), which in The residual gradient at time step 1 can be expressed as:

[0155] ;

[0156] ;

[0157] ;

[0158] ;

[0159] in:

[0160] For sensors exist Time residual Gradient in the direction.

[0161] For sensors exist Time residual Gradient in the direction.

[0162] For sensors exist Time residual Gradient in the direction.

[0163] For sensors exist The magnitude (scalar) of the residual gradient at time step.

[0164] For sensors exist Humidity prediction residual at any given time.

[0165] and The sensors are respectively in and Spacing in the direction.

[0166] Subsequently, the expected residual gradient is calculated. The expected residual gradient is calculated based on the humidity diffusion model and reflects the expected value of the residual gradient in the absence of local anomalies. This invention uses the following formula to calculate the expected residual gradient:

[0167] ;

[0168] in:

[0169] For sensors exist The expected residual gradient magnitude (scalar) at time step 1. This represents the expected residual.

[0170] According to Predicting the state at any time The calculated humidity diffusion coefficient.

[0171] Calculated based on the predicted state The magnitude of the humidity gradient over time. Its calculation method is similar to that of the residual gradient, except that the residual is... Replace with predicted humidity value :

[0172] ;

[0173] ;

[0174] ;

[0175] ;

[0176] Based on the actual residual gradient and the expected residual gradient, a Local Anomaly Index (LAI) is constructed:

[0177] ;

[0178] in:

[0179] For sensors exist The local anomaly index (scalar) at time step. It reflects the degree of deviation between the actual residual gradient and the expected residual gradient. The larger the value, the more the humidity change in the region deviates from the model's expectations, and the more significant the local anomalies.

[0180] It is a very small positive number to prevent the denominator from being zero; in this embodiment of the invention, it is set to... .

[0181] Furthermore, construct optimization factors. :

[0182] ;

[0183] in:

[0184] for The optimization factor at any given time. It is based on local anomaly indicators. Adjusting the size of the process noise covariance matrix .

[0185] The number of valid sensors involved in the calculation.

[0186] It is an adjustable parameter used to control right The extent of the impact. The larger the value, the more sensitive it is to local anomalies.

[0187] Finally, update the noise covariance matrix. :

[0188] ;

[0189] in:

[0190] for Time goes through steps Adjusted process noise covariance matrix.

[0191] for Time goes through steps Adjusted process noise covariance matrix.

[0192] The function and design intent of each part of the formula:

[0193] Based on the spatial gradient information of the humidity prediction residuals, the process noise covariance matrix is ​​further adjusted. To adapt to potential local anomalies in the humidity field of foamed concrete and improve the estimation accuracy of Kalman filtering, the gradient of the humidity prediction residual in three spatial directions is calculated first. and its modulus This study identifies local anomalies in the humidity field. These anomalies are caused by factors such as material defects, localized drying, and uneven hydration reactions, leading to deviations from the model's predictions in actual humidity changes. To distinguish between normal humidity changes and those caused by local anomalies, an expected residual gradient is introduced. It is calculated based on the humidity diffusion model and reflects the expected value of the residual gradient in the absence of local anomalies.

[0194] Furthermore, local anomaly indicators were constructed. It is used to quantify the degree of deviation between the actual residual gradient and the expected residual gradient. A larger value indicates a greater deviation and a more significant local anomaly. Based on the local anomaly index, this invention designs an optimization factor. Its function is based on Size dynamic adjustment Matrix. When A larger value indicates a strong local anomaly; in this case, reducing the value is appropriate. , thereby increasing This makes the Kalman filter rely more on the spatial smoothing effect of the model predictions, reducing the trust in sensor measurements in that area, and thus minimizing the impact of local anomalies on the overall humidity estimate; when When the value is relatively low, it indicates that the humidity change in the area is in line with the model's expectations. close to , The matrix remains essentially unchanged, and the Kalman filter is performed in the conventional manner. Parameters control right The extent of the impact The larger the value, the more sensitive it is to local anomalies. In this way, the optimization factor is... It can effectively identify and process local anomalies in humidity fields, further improving the Kalman filter's adaptability to complex humidity fields.

[0195] step Humidity estimation and maintenance humidity control based on adaptive Kalman filtering.

[0196] Based on the humidity estimation results obtained from the preceding steps, humidity control is performed during the curing process of foamed concrete. At each time step, an optimized adaptive Kalman filter algorithm (as described in step S2) is used to obtain the optimal estimate of the internal humidity field of the concrete at the current moment. Based on this humidity estimate, combined with preset humidity control targets (such as surface humidity range, maximum allowable humidity gradient, etc.), it is determined whether spraying is necessary. Specifically, the decision to activate the spraying system is mainly based on the surface humidity estimate; spraying is triggered when the surface humidity is below a set lower limit. Simultaneously, this invention also references the internal humidity estimate and calculated humidity gradient information to assist in decision-making. For example, when the internal humidity is generally low or there are areas with excessively large local humidity gradients, even if the surface humidity meets the requirements, the spraying strategy can be appropriately adjusted to achieve more refined and proactive humidity control.

[0197] Furthermore, the proposed humidity control method can be combined with model predictive control (MRC). By combining control strategies such as humidity prediction models and optimization algorithms, the spraying area, spray volume, and spraying time can be optimized. This approach effectively addresses issues in existing technologies, such as insufficient humidity control precision, difficulty in balancing surface and internal humidity requirements, and inability to effectively handle local anomalies. This improves the curing quality of foamed concrete, reduces the risk of cracking, and ensures its performance and durability.

[0198] It should be noted that: This invention is based on the steps The optimized Kalman filter algorithm is used to estimate the internal humidity field of foamed concrete components in real time, and to control the curing humidity based on the estimation results. The specific process is as follows: First, set the initial time step. and according to the steps Obtain the initial state estimate Initial state estimation covariance matrix The baseline value of the process noise covariance matrix and measurement noise covariance matrix Then, for each time step Perform the following operations in a loop:

[0199] 1. Data Acquisition: Obtain the current time. Sensor measurement values (Humidity measurement) and ambient temperature data.

[0200] 2. Kalman filter prediction step: Based on the state estimate from the previous time step. Using the state transition equation, predict the state at the current time step: , ,in, It is based on The state transition matrix is ​​calculated from the humidity and temperature data at each moment.

[0201] 3. Calculate the humidity-dependent diffusion coefficient. Based on the predicted humidity at the current moment and steps Collected temperature data, calculate humidity diffusivity .

[0202] 4. Execution steps of and Calculate the optimization factor and and update Matrix: Calculating the residual for humidity prediction Calculate the kurtosis of the residual sequence and skewness Constructing non-Gaussian indicators Calculate the optimization factor Calculate the residual spatial gradient ∇ and its modulus Calculate the expected residual gradient Constructing local anomaly indicators Calculate the optimization factor Update the noise covariance matrix during the process: .

[0203] 5. Perform the Kalman filter update step: Calculate the Kalman gain. Update state estimate: Update the state estimate covariance: .

[0204] The above technical solutions only embody the preferred technical solutions of the present invention. Any modifications that may be made by those skilled in the art to certain parts thereof embody the principles of the present invention and fall within the protection scope of the present invention.

Claims

1. A method for controlling the curing humidity of foamed concrete, characterized in that, Includes the following steps: step Data acquisition, data preprocessing, and model initialization; humidity data is acquired by humidity sensors placed inside and on the surface of foamed concrete components, and ambient temperature and humidity are measured and recorded simultaneously to form time series data; outlier detection is performed on the acquired data, obviously erroneous data is removed and suspicious data is corrected, and the moving average filtering method is used to smooth the data to reduce the impact of random noise; The foamed concrete component is divided into a three-dimensional mesh, and the state vector is defined as the humidity value of each mesh node. Based on the three-dimensional humidity diffusion equation, the state transition matrix F is constructed by discretization using the finite difference method. F is then corrected according to the boundary conditions to construct an observation matrix that maps the state vector to the sensor measurements. Initial state estimate Initialize the state estimation covariance matrix Initialization process noise covariance matrix and initialization of the measurement noise covariance matrix ; step : a. Based on the statistical characteristics of the humidity prediction residuals, the process noise covariance matrix in step S1 during model initialization is analyzed. Make initial adjustments; calculate temperature prediction residuals. Within a sliding window, the kurtosis and skewness of the residual sequences of each sensor are calculated, and a non-Gaussianity index is constructed based on the kurtosis and skewness. Construct optimization factors ;in , and according to For the process noise covariance matrix Make initial adjustments; b. Based on the spatial gradient of the humidity prediction residual, the process noise covariance matrix of the model initialization in step a is... Further adjustments were made; the spatial gradient of the humidity prediction residual at the 3D sensor location was calculated, and a local anomaly index was constructed based on the actual residual gradient and the expected residual gradient calculated from the prediction state. And construct optimization factors. ;according to Update process noise covariance matrix : step Humidity estimation and maintenance humidity control based on adaptive Kalman filtering.

2. The method for controlling the curing humidity of foamed concrete according to claim 1, characterized in that, In step S1, humidity data is acquired, ambient temperature and humidity are measured by humidity sensors arranged inside and on the surface of the foamed concrete component, and recorded simultaneously over time to form time series data. Data preprocessing: The collected data is processed as follows: outliers are detected and obviously erroneous data is removed, and suspicious data is corrected; the moving average filtering method is used to smooth the data and reduce the impact of random noise. The following operations are performed during model initialization: Define state vector The foamed concrete component is divided into a three-dimensional mesh, and the state vector is... Includes the humidity value for each grid node: ； in, For the first line, number Column, No. The humidity value of the grid nodes in the layer. , , They are respectively The number of grid nodes in the direction; Construct the state transition matrix Based on the three-dimensional humidity diffusion equation, the state transition matrix is ​​obtained by discretization using the finite difference method. Considering the humidity-dependent diffusion coefficient Optionally, the effect of heat release during hydration can be considered, and the boundary conditions can be adjusted accordingly. After correction, using the central difference scheme, the discretized form of the three-dimensional humidity diffusion equation is as follows: ； in, : Time of the first Humidity values ​​of each grid node; Time step; : respectively Grid spacing in the direction; ; The state transition matrix can be obtained by combining the discretized equations of all grid nodes. It is a sparse matrix whose non-zero elements are determined by the above difference scheme and boundary conditions; Constructing the observation matrix Mapping state vectors to sensor measurements. The number of rows equals the number of sensors, the number of columns equals the number of state variables, and each row corresponds to one sensor; Initial state estimate Use a humidity sensor to measure the humidity distribution of the component at the initial moment, as an estimate of the initial state; Initialize the state estimation covariance matrix Setting it to a large diagonal matrix indicates higher uncertainty in the initial estimate; the diagonal elements can be set as the variance of the initial humidity measurements. Initialization process noise covariance matrix : Set as a diagonal matrix, where the diagonal elements represent the initial variance of each state variable. The diagonal elements are usually set to a large value; Initialize the measurement noise covariance matrix Set according to the accuracy specifications of the humidity sensor. It is also a diagonal matrix, with the diagonal elements representing the measurement noise variance of each sensor.

3. The method for controlling the curing humidity of foamed concrete according to claim 1, characterized in that, In step S2, a) based on the statistical characteristics of the humidity prediction residual, the process noise covariance matrix in step S1 is... The specific steps for making initial adjustments are as follows: Based on the statistical properties of humidity prediction residuals The matrix undergoes initial adjustments to determine the humidity prediction residual, which is the difference between sensor measurements and model predictions. The non-Gaussianity of the residual sequence is quantified by calculating its kurtosis and skewness. Kurtosis reflects the peak of the distribution, and skewness reflects its asymmetry. Based on kurtosis and skewness, a non-Gaussianity index is constructed, referred to as: and according to Size dynamic adjustment The matrix is ​​used to calculate the humidity prediction residual, which is the difference between the sensor measurement and the model prediction. ； in, for Humidity prediction residual vector at time step; for A vector of humidity sensor measurements at any given time; The observation matrix; for Time based The predicted state vector at time step; For each sensor Calculate its residual sequence kurtosis within the sliding window and skewness The sliding window size is That is, considering the past The residuals at each time step, using a sliding window: The sliding window is used to calculate the local statistical properties of the residual sequence, namely kurtosis and skewness. The formulas for calculating kurtosis and skewness are as follows: ； ； in, For sensors exist Kurtosis of the residual sequence at time step; For sensors exist Skewness of the residual sequence at time step; For sensors exist Humidity prediction residual at any given time; The mean of the residuals within the sliding window; Construct a non-Gaussianity index based on kurtosis and skewness. : ； in, for The non-Gaussianity index at time; Number of sensors; and The first Kurtosis and skewness weights for each sensor; Constructing optimization factors: ； in, for The optimization factor at any given time; An adaptive forgetting factor; The update formula is: ； in, Forgetting factor coefficient ; for The initial value can be set to or ; Update process noise covariance matrix : ； in, for The process noise covariance matrix updated at each time step; This is the baseline value for the process noise covariance matrix.

4. The method for controlling the curing humidity of foamed concrete according to claim 1, characterized in that, In step S2, b, based on the spatial gradient of the humidity prediction residual, the process noise covariance matrix in step a... Further adjustments will be made, and the specific steps are as follows: In the steps Based on the statistical properties of the humidity prediction residuals, the covariance matrix of kurtosis and skewness against process noise is given. Preliminary adjustments were made to accommodate the non-Gaussian nature of the process noise in the steps. Based on the spatial gradient information of the humidity prediction residual, the process noise covariance matrix is ​​considered. Adjustments are made, and a local anomaly index is constructed based on the spatial gradient of the humidity prediction residual, referred to as: ; Calculate the spatial gradient of the humidity prediction residual. In the three-dimensional case, the sensors are distributed in a three-dimensional grid pattern inside the foamed concrete component. Therefore, for the sensors... ,in (The number of sensors in the three directions, respectively), which in The residual gradient at time step 1 can be expressed as: ； ； ； ； in, For sensors exist Time residual Gradient in direction; For sensors exist Time residual Gradient in direction; For sensors exist Time residual Gradient in direction; For sensors exist The magnitude of the residual gradient at time step; For sensors exist Humidity prediction residual at any given time; and The sensors are respectively in and Spacing in the direction; The expected residual gradient is calculated using the following formula: ； in, For sensors exist The expected residual gradient magnitude at time step [time]. Represents the expected residual; According to Predicting the state at any time The calculated humidity diffusion coefficient; Calculated based on the predicted state The magnitude of the humidity gradient at any given time is calculated similarly to that of the residual gradient, except that the residual is used in the calculation. Replace with predicted humidity value : ； ； ； ； Based on the actual residual gradient and the expected residual gradient, a local anomaly index is constructed, referred to as... : ； in, For sensors exist Local abnormal indicators at any given time; It is a very small positive number; Constructing optimization factors : ； in, for The optimization factor at any given time; The number of effective sensors participating in the calculation; It is an adjustable parameter; Update process noise covariance matrix : ； in, for Time goes through steps Adjusted process noise covariance matrix; for Time goes through steps Adjusted process noise covariance matrix.

5. The method for controlling the curing humidity of foamed concrete according to claim 1, characterized in that, In step S3, based on the humidity estimation results obtained in the previous steps, humidity control is performed during the curing process of foamed concrete. At each time step, the optimal estimate of the internal humidity field of the concrete at the current moment is obtained using an optimized adaptive Kalman filter algorithm to form a humidity estimate. Based on the humidity estimate and the preset humidity control target, it is determined whether spraying is required. The surface humidity estimate determines whether to start the spraying system. When the surface humidity is lower than the set lower limit, spraying is triggered.

6. The method for controlling the curing humidity of foamed concrete according to claim 1, characterized in that, The estimated internal humidity and the calculated humidity gradient information are used to assist in decision-making. When the internal humidity is generally low or there are areas with excessive local humidity gradients, even if the surface humidity meets the requirements, the spraying strategy can be adjusted appropriately to achieve more precise and forward-looking humidity control.

Images