Optimization method and device for concrete temperature monitoring based on optical fiber sensor and machine learning, and medium

Abstract

The application relates to a concrete temperature monitoring optimization method and device based on an optical fiber sensor and machine learning and a medium, wherein the method comprises the following steps: selecting and setting an optical fiber sensor; performing a simulation experiment of concrete temperature monitoring, acquiring accurate temperature monitoring data and transmission spectrum data of current concrete by using the optical fiber sensor; performing moving average algorithm processing on the temperature monitoring data to obtain characteristic temperature data; taking the measurement data of multiple peak points of the transmission spectrum as the input of a deep neural network model, taking the characteristic temperature data processed by the moving average algorithm as the output of the deep neural network model, training the model; acquiring the transmission spectrum data of the concrete by using the optical fiber sensor in a normal environment, inputting the trained deep neural network model, and outputting a temperature prediction value. Compared with the prior art, the application has the advantages of improving the economy and accuracy of concrete temperature monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of construction quality monitoring and evaluation, and in particular to a concrete temperature monitoring optimization method, device and medium based on optical fiber sensors and machine learning. Background Art

[0002] With continued socioeconomic growth, the demand for large and high-rise buildings is increasing. These buildings often utilize mass concrete as their foundation structure. However, during the pouring and hardening process of mass concrete, the hydration reaction generates a large amount of heat, leading to a dramatic increase in the temperature difference between the concrete's interior and exterior. This significant temperature difference, combined with the concrete's inherently low tensile strength, can easily cause cracks, compromising the structural integrity and potentially severely impacting the quality of the entire project, leading to serious consequences. Therefore, effectively monitoring and analyzing the temperature field changes of mass concrete during pouring is crucial for guiding crack prevention and control efforts and ensuring pouring quality. Traditional temperature monitoring methods rely primarily on point-type temperature sensors, which require manual periodic checks and carry the risk of missed detections, sensor failure, and interference with construction during installation. Furthermore, the data collected by this method is limited and discontinuous, making it difficult to fully reflect the temperature field changes within the concrete. Summary of the Invention

[0003] The purpose of the present invention is to overcome the problems of low monitoring and processing efficiency in the above-mentioned prior art, which cannot accurately and economically monitor concrete temperature, resulting in excessive temperature differences between the inside and outside of the concrete causing cracks, affecting structural integrity and project quality, and to provide a concrete temperature monitoring optimization method, device and medium based on optical fiber sensors and machine learning.

[0004] The purpose of the present invention can be achieved by the following technical solutions:

[0005] A concrete temperature monitoring optimization method based on optical fiber sensors and machine learning includes the following steps:

[0006] S1. Select and set up the optical fiber sensor;

[0007] S2. Conduct a simulation experiment on concrete temperature monitoring and use fiber optic sensors to obtain accurate temperature monitoring data and transmission spectrum data of the current concrete;

[0008] S3. Process the temperature monitoring data using a moving average algorithm to obtain characteristic temperature data;

[0009] S4. Using the measured data of multiple peak points of the transmission spectrum as the input of the deep neural network model, and using the characteristic temperature data processed by the moving average algorithm as the output of the deep neural network model to train the model;

[0010] S5. Use the fiber optic sensor to obtain the transmission spectrum data of concrete in a normal environment, input it into the trained deep neural network model, and output the temperature prediction value.

[0011] The step S1 specifically includes: selecting a tilted Bragg grating sensor as the optical fiber sensor, determining the grating tilt angle and gauge length of the sensor, as well as the wavelength range, wavelength resolution, and wavelength scanning speed of the transmission spectrum.

[0012] In step S2, the simulation experiment simulates a large-volume concrete structure in a three-dimensional space, and the optical fiber sensors with calibrated positions are arranged in the simulated large-volume concrete structure space.

[0013] In step S3, the temperature monitoring data is processed by a moving average algorithm in the time, space and quality dimensions, and the characteristic temperature data of the sensor at a certain moment and a certain spatial point are represented by the arithmetic average of the original monitoring data of all valid sensors in the time, space and quality dimensions respectively.

[0014] The moving average calculation formulas for the time, space, and quality dimensions are as follows:

[0015]

[0016] Where t is the moment, s is the spatial point, o is the effective sensor in the monitored object, Δ(t) is the time interval containing the moment, Ω(s) is the spatial unit containing the spatial point, N(o) is all the temperature sensors deployed in the monitored object, n(t), n(s), and n(o) are the number of temperature values ​​collected in the time interval, spatial unit, and monitored object, respectively. T i (t), T j (s), T k (o) are the corresponding original temperature measurements, T(t), T(s), and T(o) are the characteristic temperature data of the moment, the spatial point, and the monitored object, respectively.

[0017] The input of the deep neural network model is the wavelengths of multiple peak points of the transmission spectrum.

[0018] The training of the deep neural network model specifically includes the following steps:

[0019] Collect transmission spectrum wavelength data recorded by the tilted Bragg grating sensor under different temperature and strain conditions;

[0020] A specific number of wavelength peak points are selected as model input, and the corresponding characteristic temperature data is used as model output. The data set is divided into training set, validation set, and test set in proportion. The model is trained using the training set and validated using the validation set. The model with the best performance on the test set is selected as the optimal model to complete the model training.

[0021] The deep neural network model uses the LM algorithm to minimize the loss function to implement model training.

[0022] A concrete temperature monitoring and optimization device based on optical fiber sensors and machine learning includes a memory, a processor, and a program stored in the memory. When the processor executes the program, the method described above is implemented.

[0023] A storage medium stores a program thereon, and when the program is executed, the method described above is implemented.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] This invention transfers the deep neural network model in the field of machine learning and applies it to the field of construction engineering, optimizes the existing concrete temperature monitoring method, improves the efficiency of concrete health monitoring, can accurately and economically monitor the internal temperature of concrete in real time, and improves the accuracy of concrete health detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flow chart of the method of the present invention;

[0027] Figure 2 Schematic diagram of the measurement of the TFBG sensor of the present invention;

[0028] Figure 3 This is a schematic diagram of the arrangement of optical fiber sensors for the simulation experiment of the present invention;

[0029] Figure 4 Schematic diagram of the neuron structure of a deep neural network;

[0030] Figure 5 Schematic diagram of the process of data processing and model training of the deep neural network model of the present invention. DETAILED DESCRIPTION

[0031] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0032] This embodiment provides a concrete temperature monitoring optimization method based on optical fiber sensor and machine learning, such as Figure 1 As shown, the following steps are included:

[0033] S1. Select and set up the fiber optic sensor.

[0034] Fiber optic sensors, with their high precision, excellent safety, anti-interference capabilities, stability, and durability, can provide a variety of information, including temperature, strain, displacement, subsidence, tilt, and acceleration. Monitoring systems developed based on fiber optic sensing technology have been widely used in various civil engineering applications. Currently, widely used fiber optic sensors include Bragg gratings, long-period fiber gratings, Fabry-Perot (FP), and fiber structures based on multimode interferometry (MMI). Fiber Bragg grating (FBG) sensors are widely used for temperature and strain measurement. A fiber Bragg grating (FBG) is fabricated by exposing the core of a single-mode optical fiber to intense ultraviolet light in a periodic pattern. A small amount of light is reflected at each spatially periodic refractive index change. When the grating period is approximately half the wavelength of the incident light, all reflected light coherently combines into a single, large reflection with a specific wavelength. This condition is known as the Bragg condition, and the wavelength at which the incident light is reflected is called the Bragg wavelength, which varies with the sensor's axial strain and temperature. Tilted Fiber Bragg Gate temperature sensors (TFBG) are sensors that can measure temperature and strain simultaneously without the need for additional sensors.

[0035] In this embodiment, a tilted Bragg grating sensor is selected as the optical fiber sensor, and the grating tilt angle and gauge length of the sensor, as well as the wavelength range, wavelength resolution, and wavelength scanning speed of the transmission spectrum are determined.

[0036] Specifically, the following case can be used as a reference: A TFBG sensor has a tilt angle of 4° and a gauge length of 6 mm. The measurement system diagram is as follows: Figure 2 . A tunable laser and a power meter are used to measure the transmission spectrum of the TFBG sensor in the wavelength range of 1510-1560nm. The wavelength resolution is set to 1pm. The wavelength scanning speed is set to 10nm / s. By placing an optical circulator between the light source and the sensor, the transmitted light and the reflected light are obtained at the same time. An optical isolator is placed between the sensor and the power meter to prevent the light from being reflected back due to reflection at the connection point between the optical fiber and the power meter. The light source emits incident light, which is divided into transmitted light and reflected light by the optical circulator. The transmitted light passes through the Bragg grating sensor (under different temperature conditions) and is received by the power meter to obtain different transmission spectra under different temperature conditions.

[0037] S2. Conduct a simulation experiment of concrete temperature monitoring and use optical fiber sensors to obtain accurate temperature monitoring data and transmission spectrum data of the current concrete.

[0038] The simulation experiment simulates a large concrete structure in a three-dimensional space, and the calibrated optical fiber sensors are placed in the simulated large concrete structure space. Specifically, the following case can be used as a reference: a 4m×1.5m×0.8m three-dimensional space is used to simulate a large concrete structure. The sensor layout model is as follows: Figure 3 As shown, there are three layers in total, and each layer is wired in a folded manner.

[0039] S3. Process the temperature monitoring data using a moving average algorithm to obtain characteristic temperature data.

[0040] Moving average (MA) is a commonly used technique in data analysis. The basic idea is that the object being averaged changes dynamically over time. Traditional moving averages are widely used in time series data analysis, industrial filtering, stock forecasting, and other fields. Depending on the size of the average window, moving averages can be divided into short-term, medium-term, and long-term moving averages. Depending on the averaging method, moving averages can be further divided into simple moving averages, weighted moving averages, and other types. Its general calculation formula can be expressed as:

[0041] FMA=(f(x)+f(x)+f(x)+…+

[0042] f(xN)) / N.

[0043] Among them: F MA is the moving average value of variable x, N is the size of the moving average window, and f(x) is the average calculation function.

[0044] The moving average method was originally used to analyze stock price fluctuations by breaking them down into a combination of primary trends, secondary trends, and daily fluctuations. In the field of monitoring data analysis, this method can be understood as identifying components within the data, including the true value, random error, and systematic error. Random errors are ubiquitous and frequent, typically resulting from random fluctuations caused by uncertainties in the measurement process. They tend to cancel each other out and generally follow a normal distribution. Systematic errors, on the other hand, are often caused by improper instrument installation or abnormalities in the monitoring system, manifesting as regular deviations.

[0045] Moving average technology excels at filtering high-frequency and periodic errors from data. By extending the traditional time series moving average method to include spatial and mass dimensions, a multidimensional moving average analysis method can be developed, particularly suitable for real-time concrete temperature monitoring data. This multidimensional moving average analysis method not only eliminates the implicit high-frequency random errors in large amounts of real-time temperature monitoring data but also reduces systematic errors in the data by dynamically identifying and removing abnormal temperature sensors.

[0046] The temperature monitoring data is processed by the moving average algorithm in the time, space and quality dimensions, and the characteristic temperature data of the sensor at a certain moment and a certain spatial point are represented by the arithmetic average of the original monitoring data of all valid sensors in the time, space and quality dimensions respectively.

[0047] The moving average calculation formulas for time, space, and quality dimensions are as follows:

[0048]

[0049] Where t is the moment, s is the spatial point, o is the effective sensor in the monitored object, Δ(t) is the time interval containing the moment, Ω(s) is the spatial unit containing the spatial point, N(o) is all the temperature sensors deployed in the monitored object, n(t), n(s), and n(o) are the number of temperature values ​​collected in the time interval, spatial unit, and monitored object, respectively. T i (t), T j (s), T k (o) are the corresponding original temperature measurements, T(t), T(s), and T(o) are the characteristic temperature data of the moment, the spatial point, and the monitored object, respectively.

[0050] In one embodiment, the moving average algorithm specifically includes the following steps:

[0051] S301, the sensor collects raw data;

[0052] S302, the collected data is transmitted online in real time to the original database for storage;

[0053] S303, processing the temperature monitoring data of the original database using time, space, and quality dimension average moving code programs respectively;

[0054] S304, the processed data is stored in a feature database;

[0055] S305, extracting the characteristic temperature of a certain time point, space point, or monitoring object from the characteristic database for use in the model output of step S4.

[0056] S4. Use the measurement data of multiple peak points of the transmission spectrum as the input of the deep neural network model, and use the characteristic temperature data processed by the moving average algorithm as the output of the deep neural network model to train the model.

[0057] Deep Neural Networks (DNN) is a neural network model composed of multiple hidden layers. Each hidden layer consists of multiple neurons, which transmit information and perform calculations through weights and activation functions.

[0058] Deep neural networks, through multiple layers of nonlinear transformations, can learn more abstract and complex feature representations. Each layer transforms input data into a higher-level representation, better capturing the data's characteristics and patterns. By continuously adding hidden layers, the network gradually learns more abstract features, improving the model's expressive power.

[0059] A deep neural network consists of an input layer, hidden layers, and an output layer. The input layer is responsible only for inputting data, has no activation function, and contains neurons equal to the number of features in a single instance. The hidden layer extracts features and must contain an activation function. The output layer outputs the model's predicted values ​​and may contain an activation function. The number of neurons in the output layer is related to the number of label categories.

[0060] Neurons are the most basic units in neural networks. Their structure is as follows: Figure 4 As shown in the figure, x is the input, each connection has a weight w, and the middle node is the artificial neuron node; δ is a nonlinear transformation called an activation function, the purpose of which is to enable artificial neurons to have the ability to represent nonlinear relationships; the parameter b is called the bias; output is the output of the artificial neuron.

[0061] An activation function is a nonlinear function in a neural network that acts on the input signal of a neuron and converts it into its output. The activation function introduces nonlinear transformations into the neural network, increasing the network's expressive power.

[0062] The loss function is a mathematical function that measures the error between the predicted value and the true value to determine whether the model's prediction results are accurate and evaluate the model's performance.

[0063] The tilted Bragg grating sensor (TFBG sensor) can sense wavelength changes in the transmission spectrum and use the wavelengths of multiple peak points of the transmission spectrum as feature values ​​input to machine learning. 70% of the data is used for training, 15% for validation, and 15% for testing.

[0064] like Figure 5 As shown in Figure 2, the training of the deep neural network model specifically includes the following steps:

[0065] S401, collecting transmission spectrum wavelength data recorded by the tilted Bragg grating sensor under different temperature and strain conditions;

[0066] S402: Select a specific number of wavelength peak points as model input and the corresponding characteristic temperature data as model output. Divide the data set into training set, validation set, and test set in a ratio of 70%, 15%, and 15%. Use the training set to train the model, use the validation set to validate the model, and select the model with the best performance on the test set as the optimal model to complete model training.

[0067] This embodiment uses the Levenberg-Marquardt algorithm (LM algorithm) as a solution to reduce the loss function. The LM algorithm uses both gradient descent and Gauss-Newton methods to reduce the loss function. Gradient descent is less affected by the location of the loss function minimum, while Gauss-Newton is faster and more accurate near the minimum. The LM algorithm combines the advantages of both methods.

[0068] The number of layers of the deep neural network is set to 30, and the network code is written using Python and its libraries.

[0069] The root mean square error (RMSE) is used to evaluate the performance of the model. The calculation formula is as follows:

[0070]

[0071] Among them, N is the number of predicted samples, observed t Indicates the accurate temperature monitoring value of the t-th sample, predicted t represents the model-predicted temperature value for the tth sample.

[0072] S5. Use the fiber optic sensor to obtain the transmission spectrum data of concrete in a normal environment, input it into the trained deep neural network model, and output the temperature prediction value.

[0073] In specific implementation, using the pouring and curing of large-volume concrete for bridge towers as an example, the selection and parameter setting of TFBG sensors were completed before pouring began. Raw temperature data was collected in a laboratory using a scaled-down spatial simulation experiment, and a moving average algorithm was applied across time, space, and mass dimensions. An appropriate number of peak values ​​were selected from the transmission spectrum data collected by the TFBG sensor system. This temperature data, processed using the moving average algorithm, was then divided into training, test, and validation sets, which served as the input and output of a deep neural network model. After repeated model training and evaluation, the optimal model was identified. During the concrete pouring process of the actual project, actual TFBG sensors were deployed according to the layout used in the simulation experiment. Real-time transmission spectrum data of the concrete was obtained through demodulation and real-time transmission. The trained prediction model was used to predict the actual temperature monitoring value.

[0074] This embodiment provides a concrete temperature monitoring and optimization device based on an optical fiber sensor and machine learning, including a memory, a processor, and a program stored in the memory. When the processor executes the program, the method described above is implemented.

[0075] In a preferred embodiment, the device comprises:

[0076] Data acquisition module: Select and set up the fiber optic sensor; conduct a simulation experiment of concrete temperature monitoring, and use the fiber optic sensor to obtain accurate temperature monitoring data and transmission spectrum data of the current concrete;

[0077] Temperature data processing module: Process the temperature monitoring data using a moving average algorithm to obtain characteristic temperature data;

[0078] Model training module: The measured data of multiple peak points of the transmission spectrum are used as the input of the deep neural network model, and the characteristic temperature data processed by the moving average algorithm is used as the output of the deep neural network model to train the model;

[0079] Temperature monitoring module: Use fiber optic sensors to obtain the transmission spectrum data of concrete in a normal environment, input it into the trained deep neural network model, and output the temperature prediction value.

[0080] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0081] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0082] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A concrete temperature monitoring optimization method based on optical fiber sensor and machine learning, characterized in that: The following steps are involved: S1. Select and set up the optical fiber sensor; S2. Conduct a simulation experiment on concrete temperature monitoring and use fiber optic sensors to obtain accurate temperature monitoring data and transmission spectrum data of the current concrete; S3, performing moving average algorithm processing on the temperature monitoring data to obtain characteristic temperature data; S4. Using the measured data of multiple peak points of the transmission spectrum as the input of the deep neural network model, and using the characteristic temperature data processed by the moving average algorithm as the output of the deep neural network model to train the model; S5. Use the fiber optic sensor to obtain the transmission spectrum data of concrete in a normal environment, input it into the trained deep neural network model, and output the temperature prediction value; In step S3, the temperature monitoring data is processed by a moving average algorithm in the time, space, and quality dimensions, and the characteristic temperature data of the sensor at a certain time and a certain spatial point are represented by the arithmetic average of the original monitoring data of all valid sensors in the time, space, and quality dimensions respectively; The input of the deep neural network model is the wavelengths of multiple peak points of the transmission spectrum; The training of the deep neural network model is specifically The following steps are involved: Collect transmission spectrum wavelength data recorded by the tilted Bragg grating sensor under different temperature and strain conditions; A specific number of wavelength peak points are selected as model input, and the corresponding characteristic temperature data is used as model output. The data set is divided into training set, validation set, and test set in proportion. The model is trained using the training set and validated using the validation set. The model with the best performance on the test set is selected as the optimal model to complete the model training.

2. The concrete temperature monitoring optimization method based on optical fiber sensor and machine learning according to claim 1 is characterized in that: The step S1 specifically includes: selecting a tilted Bragg grating sensor as the optical fiber sensor, determining the grating tilt angle and gauge length of the sensor, as well as the wavelength range, wavelength resolution, and wavelength scanning speed of the transmission spectrum.

3. The concrete temperature monitoring optimization method based on optical fiber sensor and machine learning according to claim 1 is characterized in that: In step S2, the simulation experiment simulates a large-volume concrete structure in a three-dimensional space, and the optical fiber sensors with calibrated positions are arranged in the simulated large-volume concrete structure space.

4. The concrete temperature monitoring optimization method based on optical fiber sensor and machine learning according to claim 1 is characterized in that: The moving average calculation formulas for the time, space, and quality dimensions are as follows: in For the moment, is a spatial point, To monitor the effective sensors in the object, is the time interval that includes this moment, is the spatial unit containing the spatial point, For all temperature sensors deployed in the monitoring object, 、 、 are the number of temperature values ​​collected within the time interval, space unit and monitoring object, respectively. 、 、 are the corresponding original temperature measurements, 、 、 They are the characteristic temperature data of the moment, the spatial point and the monitored object respectively.

5. The concrete temperature monitoring optimization method based on optical fiber sensor and machine learning according to claim 1 is characterized in that: The deep neural network model uses the LM algorithm to minimize the loss function to implement model training.

6. A concrete temperature monitoring and optimization device based on optical fiber sensors and machine learning, comprising a memory, a processor, and a program stored in the memory, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 5 is implemented.

7. A storage medium having a program stored thereon, characterized in that: When the program is executed, the method according to any one of claims 1 to 5 is implemented.

Images