Photonic crystal fiber axis fixing device
The photonic crystal optical fiber axis fixing device automates light spot and brightness adjustments using a control system with a light spot detection module and machine learning, addressing precision and efficiency issues in optical fiber alignment.
Patent Information
- Application Number
- CN202411199943.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-12-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing photonic crystal fiber axis-fixing devices require manual adjustment and lack precision in forming a clear light spot and adjusting light source brightness, leading to inefficiencies in optical fiber connection.
A photonic crystal optical fiber axis fixing device with a control system that includes a light spot detection module, automatic adjustment module, and a prediction and recommendation module, utilizing PID controllers and machine learning algorithms to automate and optimize light spot position and brightness adjustment.
Enables automatic and precise adjustment of light spot position and brightness, improving the accuracy and efficiency of optical fiber axis alignment by leveraging historical data and machine learning for predictive adjustments.
Smart Images

Figure CN119087585A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of optical fiber axis fixing, and in particular to a photonic crystal optical fiber axis fixing device. Background Art
[0002] The photonic crystal fiber axis-fixing device is a device used for optical fiber axis-fixing in photonic crystal fiber connection. Its design and function are aimed at improving the accuracy and efficiency of optical fiber optical path connection. Photonic crystal fiber (also known as holey fiber, air hole-assisted fiber, microstructure fiber) is a special optical fiber with a built-in microstructure, which contains closely arranged small air holes. In the process of connecting photonic crystal fibers, due to the presence of holes in the fiber end face configuration, the connection between the fiber and other optical components becomes difficult, so a specific axis-fixing device is required to ensure the precise alignment of the fiber.
[0003] After searching, the invention patent with Chinese patent number CN105785505A discloses a device and method for locating the axis of a photonic crystal fiber;
[0004] Compared with the prior art, the invention patent with Chinese patent number CN105785505A can align the optical fiber axis through a mobile platform and a rotator based on the real-time information collection of the measuring optical path and the code disk. The present invention assists the optical fiber in selecting and fixing the sensitive axis direction by real-time monitoring the axial position and rotation angle of the optical fiber, and can quickly and effectively find the required optical fiber axis and fix the optical fiber.
[0005] However, in actual use, when the light spot is formed by the illumination light source, the above device needs to be manually adjusted continuously by the staff before it can be used, and the accuracy is not high. Therefore, a photonic crystal fiber axis fixing device is proposed, which can realize automatic and precise adjustment of the light spot position and the brightness of the light source. Summary of the invention
[0006] The purpose of the present invention is to realize automatic and accurate adjustment of the light spot position and the brightness of the light source, and to propose a photonic crystal optical fiber axis fixing device.
[0007] In order to achieve the above object, the present invention adopts the following technical solutions:
[0008] A photonic crystal optical fiber axis fixing device comprises an optical fiber axis fixing device body and an optical fiber, a light spot detection module is installed inside the optical fiber axis fixing device body, and a control system for the light spot detection module is arranged on the optical fiber axis fixing device body, and the control system comprises:
[0009] The initial setting module is responsible for the overall initialization setting of the system;
[0010] A light spot detection module is used to capture images and then analyze the clarity, brightness, position and size characteristics of the light spot;
[0011] The automatic adjustment module is used to calculate and send instructions for adjusting the light source intensity and the light spot position according to the light spot feature information provided by the light spot detection module in combination with preset algorithms and parameters;
[0012] The mobile platform control module receives the instruction from the automatic adjustment module and controls the movement of the mobile platform of the main body of the optical fiber axis fixing device, thereby adjusting the position of the light spot to ensure the accuracy of the light spot;
[0013] The data storage module is responsible for storing historical data during the system adjustment process, such as light source brightness and spot position, to provide data support for the later stage;
[0014] The data analysis module is used to analyze historical adjustment data and extract valuable information, such as the optimal value of light source brightness and adjustment trend, to provide a basis for later analysis;
[0015] The prediction and recommendation module is used to predict the optimal light source brightness value and light spot position based on the results provided by the data analysis module, and to obtain the recommended value;
[0016] The light source control module is responsible for receiving instructions from the automatic adjustment module or the prediction and recommendation module, controlling the brightness output of the light source, and ensuring that the brightness of the light source is consistent with the expected value.
[0017] The above technical solution further includes:
[0018] The light spot detection module includes a light spot analysis unit, an instruction generation unit and an instruction sending unit. The light spot analysis unit is responsible for receiving the light spot image feature information transmitted by the light spot detection module and performing in-depth analysis on the information. The instruction generation unit generates corresponding adjustment instructions according to the analysis results of the light spot analysis unit. The instruction sending unit is responsible for sending the adjustment instructions to the light source control module and the mobile platform control module.
[0019] The light spot analysis unit analyzes the characteristic information of the light spot according to the PID controller to determine whether the brightness of the light spot reaches a preset threshold and whether the position of the light spot deviates from a predetermined area. The calculation formula of the PID controller is as follows;
[0020] Output value = Kp*e(t)+Ki*∫e(t)dt+Kd*de(t) / dt
[0021] Among them, e(t) represents the gap between the target value and the current value, ∫e(t)dt represents the integral term of the deviation, de(t) / dt represents the differential term of the deviation, Kp, Ki and Kd are the proportional, integral and differential coefficients respectively, which are used to adjust the response speed, stability and anti-interference ability of the controller.
[0022] The light source control module includes a brightness adjustment unit and a brightness control unit. The brightness adjustment unit is responsible for receiving instructions from the automatic adjustment module or the prediction and recommendation module. The brightness control unit is responsible for adjusting the brightness. The mobile platform control module includes an instruction receiving unit and a platform control unit. The instruction receiving unit is responsible for receiving instructions sent by the automatic adjustment module. The platform control unit is responsible for controlling the movement of the mobile platform on the main body of the optical fiber axis fixing device according to the instructions.
[0023] The data analysis module includes a data preprocessing unit, a data analysis unit and a result output unit. The data preprocessing unit is responsible for cleaning and organizing historical data, the data analysis unit is responsible for conducting in-depth analysis of the data using an analysis algorithm, and the result output unit is responsible for outputting the analysis results to the prediction and recommendation module.
[0024] The data preprocessing unit includes processing missing values, processing duplicate values, processing outliers, data standardization and normalization, and feature selection. The analysis algorithm of the data analysis unit is linear regression, and the formula of the linear regression is:
[0025] b=(x1y1+x2y2+...xnyn-nXY) / (x1+x2+...xn-nX)
[0026] Among them, b is the regression coefficient, X and Y are the average values of the independent variable and the dependent variable respectively, xi and yi are the specific values of each variable, and n is the number of data sets.
[0027] The prediction and recommendation module includes a prediction algorithm unit, a recommendation generation unit and a recommendation sending unit. The prediction algorithm is used to make predictions. The recommendation generation unit is responsible for generating recommendation values according to the prediction results of the prediction algorithm unit. The recommendation sending unit is responsible for sending the recommendation values to the light source control module.
[0028] The prediction algorithm unit performs prediction through a convolutional neural network, and the algorithm steps of the convolutional neural network are:
[0029] Input data representation: The input data is a multidimensional array, denoted as X, where the dimensions of X are batch size, number of channels, height, and width, which represent the number of input samples, number of channels, image height, and image width, respectively;
[0030] Convolution operation: The convolution kernel is used to perform a sliding window convolution operation on the input data to extract the features in the input data. The dimensions of the convolution kernel are the number of channels, the height of the convolution kernel, and the width of the convolution kernel, denoted as W. The convolution operation can be expressed by the following formula:
[0031] Z[i,j]=sum(sum(sum(X[:,:,i:i+h,j:j+w]*W)))+b
[0032] Among them, Z is the output of the convolution operation, representing the feature map, i and j represent the position of the feature map, h and w represent the height and width of the convolution kernel respectively, and b is the bias term;
[0033] Activation function: After the convolution operation, ReLU is applied to perform nonlinear transformation on the feature map to increase the network’s expressive power;
[0034] Pooling operation: By selecting the average value in the local perception area as the pooling result, the spatial dimension of the feature map is reduced and important feature information is retained;
[0035] Fully connected operation: After multiple convolution and pooling operations, the feature map is expanded into a one-dimensional vector and passed through a fully connected layer for classification or regression tasks. Each neuron in the fully connected layer is connected to all neurons in the previous layer, and the mapping relationship between input and output is established by learning weights and biases.
[0036] Loss function: In CNN, the cross entropy loss function is used to measure the gap between the model's prediction results and the true label;
[0037] Back propagation algorithm: Use the back propagation algorithm to calculate the gradient of the loss function with respect to the model parameters, and use gradient descent to update the model parameters to continuously optimize the performance of the model;
[0038] The recommendation generation unit makes recommendations through a machine learning model, and the implementation steps are as follows:
[0039] Data collection and preprocessing:
[0040] Collect data on light source brightness adjustment from historical records, including time, environment, and activity factors, as well as corresponding brightness adjustment values, check whether the data contains missing values, outliers, or duplicates, and process them accordingly, formatting the data into a format that can be processed by the machine learning model, such as converting text data into numerical data;
[0041] Feature Engineering:
[0042] Extract features that affect the predicted light source brightness from the raw data, such as time, environment, and activity. Convert categorical features into numerical representations. For numerical features, standardize or normalize them to eliminate scale differences between different features.
[0043] Model training:
[0044] The data set is divided into training set, validation set and test set. The training set is used to train the decision tree model, the validation set is used to adjust the model parameters and hyperparameters, and the test set is used to evaluate the performance of the decision tree model. The training data is used to train the decision tree model.
[0045] Model evaluation and optimization:
[0046] The mean square error is selected to evaluate the performance of the model and measure the deviation between the brightness value predicted by the model and the true value. According to the evaluation results, the model is tuned, including adjusting the parameters of the model, changing the structure of the model, and trying different feature combinations.
[0047] The algorithm formula of the mean square error is:
[0048] text{MSE}=\frac{1}{n}\sum_{i=1}^{n}(y_i-\hat{y}_i)^2
[0049] Where n is the number of samples, y_i is the true value of the i-th sample, and hat{y}_i is the predicted value of the i-th sample;
[0050] Application and deployment:
[0051] The department deploys the trained model to make predictions on new data when needed. When encountering a new situation, relevant factors are input into the model, and the model will output a recommended brightness value. The performance of the model is monitored regularly.
[0052] The present invention has the following beneficial effects:
[0053] 1. In the present invention, in the process of optical fiber axis positioning, the setting of the light source and the optical path is a crucial step. By automatically adjusting the brightness of the lighting source and the position of the light spot, a clear light spot can be formed on the optical fiber support table, and automatic and precise adjustment of the light spot position and the brightness of the light source can be achieved.
[0054] 2. In the present invention, the appropriate brightness value and position can be quickly predicted and recommended based on the historical data of automatic adjustment and machine learning. Through data analysis and machine learning, the adjustment strategy can be continuously learned and optimized to gradually improve the accuracy of prediction and adjustment. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] Figure 1 This is a schematic diagram of the overall structure of a photonic crystal fiber axis fixing device proposed by the present invention;
[0056] Figure 2 It is a schematic diagram of the system block in the present invention.
[0057] In the figure: 1. Initial setting module; 2. Light source control module; 3. Light spot detection module; 4. Automatic adjustment module; 5. Mobile platform control module; 6. Data storage module; 7. Data analysis module; 8. Prediction and recommendation module; 9. Fiber optic axis fixing device body; 10. Optical fiber; 11. Light source. DETAILED DESCRIPTION
[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0059] Embodiment 1
[0060] like Figure 1-Figure 2 As shown, a photonic crystal fiber axis fixing device proposed by the present invention includes an optical fiber axis fixing device body 9 and an optical fiber 10, a light spot detection module 3 is installed inside the optical fiber axis fixing device body 9, and a control system for the light spot detection module 3 is provided on the optical fiber axis fixing device body 9, and the control system includes:
[0061] Initial setting module 1, responsible for the overall initialization setting of the system;
[0062] The light spot detection module 3 is used to capture the image and then analyze the characteristics of the light spot;
[0063] The automatic adjustment module 4 is used to calculate and send instructions for adjusting the light source intensity and the light spot position according to the light spot feature information provided by the light spot detection module 3;
[0064] The mobile platform control module 5 receives the instruction from the automatic adjustment module 4 and controls the movement of the mobile platform on the optical fiber axis fixing device body 9;
[0065] Data storage module 6, used for storing historical data in the process of system adjustment;
[0066] Data analysis module 7, used for analyzing historical adjustment data;
[0067] The prediction and recommendation module 8 is used to predict the optimal light source brightness value and light spot position based on the results provided by the data analysis module 7, and obtain the recommended value;
[0068] The light source control module 2 is responsible for receiving instructions from the automatic adjustment module 4 or the prediction and recommendation module 8 to control the brightness output of the light source;
[0069] The light spot detection module 3 includes a light spot analysis unit, an instruction generation unit and an instruction sending unit. The light spot analysis unit is responsible for receiving the light spot image feature information transmitted by the light spot detection module and performing in-depth analysis on the information. The instruction generation unit generates corresponding adjustment instructions according to the analysis results of the light spot analysis unit. The instruction sending unit is responsible for sending the adjustment instructions to the light source control module 2 and the mobile platform control module 5. The light spot analysis unit analyzes the feature information of the light spot according to the PID controller to determine whether the brightness of the light spot reaches a preset threshold and whether the position of the light spot deviates from the predetermined area. The calculation formula of the PID controller is as follows;
[0070] Output value = Kp*e(t)+Ki*∫e(t)dt+Kd*de(t) / dt
[0071] Among them, e(t) represents the difference between the target value and the current value, ∫e(t)dt represents the integral term of the deviation, de(t) / dt represents the differential term of the deviation, Kp, Ki and Kd are the proportional, integral and differential coefficients respectively, which are used to adjust the response speed, stability and anti-interference ability of the controller;
[0072] The light source control module 2 includes a brightness adjustment unit and a brightness control unit. The brightness adjustment unit is responsible for receiving instructions from the automatic adjustment module 4 or the prediction and recommendation module 8. The brightness control unit is responsible for adjusting the brightness. The mobile platform control module 5 includes an instruction receiving unit and a platform control unit. The instruction receiving unit is responsible for receiving instructions sent by the automatic adjustment module 4. The platform control unit is responsible for controlling the movement of the mobile platform on the optical fiber axis fixing device body 9 according to the instructions.
[0073] In this embodiment, when the optical fiber 10 is being fixed, the lighting source 11 can be turned on, and the initial setting module 1 is initialized, and then the lighting source 11 is initially set to a medium light intensity through the light source control module 2. At this time, the light spot detection module 3 captures the image, and then analyzes the clarity, brightness, position and size of the light spot, and then transmits the characteristic information of the light spot to the automatic adjustment module 4. At this time, the automatic adjustment module 4 calculates the required adjustment of the light source intensity and the light spot position according to the light spot characteristic information provided by the light spot detection module 3, and then sends the instruction to the light source control module 2 and the mobile platform control module 5, through The mobile platform control module 5 controls the movement of the mobile platform on the optical fiber axis fixing device body 9, thereby adjusting the position of the light spot to ensure the accuracy of the light spot. Then, after the light source control module 2 receives the instruction from the automatic adjustment module 4, it controls the light source brightness output of the lighting source 11 to ensure that the light source is consistent with the expected value. In this way, by automatically adjusting the brightness of the lighting source 11 and the position of the light spot, a clear light spot can be formed on the optical fiber support table, and automatic and precise adjustment of the light spot position and the brightness of the light source can be achieved. Then, the optical fiber 10 is placed on the optical fiber axis fixing device body 9 with a clamp V-groove to carry out the axis fixing work.
[0074] Embodiment 2
[0075] like Figure 1-Figure 2 As shown, based on the first embodiment, the data analysis module 7 includes a data preprocessing unit, a data analysis unit and a result output unit. The data preprocessing unit is responsible for cleaning and sorting the historical data, the data analysis unit is responsible for using the analysis algorithm to conduct in-depth analysis of the data, and the result output unit is responsible for outputting the analysis results to the prediction and recommendation module.
[0076] The data preprocessing unit includes processing missing values, processing duplicate values, processing outliers, data standardization and normalization, and feature selection. The analysis algorithm of the data analysis unit is linear regression, and the formula of linear regression is:
[0077] b=(x1y1+x2y2+...xnyn-nXY) / (x1+x2+...xn-nX)
[0078] Wherein, b is the regression coefficient, X and Y are the average values of the independent variable and the dependent variable respectively, xi and yi are the specific values of the respective variables, n is the number of data groups, and the prediction and recommendation module 8 includes a prediction algorithm unit, a recommendation generation unit and a recommendation sending unit. The prediction algorithm unit is responsible for making predictions using the prediction algorithm, the recommendation generation unit is responsible for generating a recommendation value according to the prediction result of the prediction algorithm unit, and the recommendation sending unit is responsible for sending the recommendation value to the light source control module 2. The prediction algorithm unit makes predictions through a convolutional neural network, and the algorithm steps of the convolutional neural network are as follows:
[0079] Input data representation: The input data is a multidimensional array, denoted as X, where the dimensions of X are batch size, number of channels, height, and width, which represent the number of input samples, number of channels, image height, and image width, respectively;
[0080] Convolution operation: The convolution kernel is used to perform a sliding window convolution operation on the input data to extract the features in the input data. The dimensions of the convolution kernel are the number of channels, the height of the convolution kernel, and the width of the convolution kernel, denoted as W. The convolution operation can be expressed by the following formula:
[0081] Z[i,j]=sum(sum(sum(X[:,:,i:i+h,j:j+w]*W)))+b
[0082] Among them, Z is the output of the convolution operation, representing the feature map, i and j represent the position of the feature map, h and w represent the height and width of the convolution kernel respectively, and b is the bias term;
[0083] Activation function: After the convolution operation, ReLU is applied to perform nonlinear transformation on the feature map to increase the network’s expressive power;
[0084] Pooling operation: By selecting the average value in the local perception area as the pooling result, the spatial dimension of the feature map is reduced and important feature information is retained;
[0085] Fully connected operation: After multiple convolution and pooling operations, the feature map is expanded into a one-dimensional vector and passed through a fully connected layer for classification or regression tasks. Each neuron in the fully connected layer is connected to all neurons in the previous layer, and the mapping relationship between input and output is established by learning weights and biases.
[0086] Loss function: In CNN, the cross entropy loss function is used to measure the gap between the model's prediction results and the true label;
[0087] Back propagation algorithm: Use the back propagation algorithm to calculate the gradient of the loss function with respect to the model parameters, and use gradient descent to update the model parameters to continuously optimize the performance of the model;
[0088] The recommendation generation unit makes recommendations through a machine learning model. The implementation steps are as follows:
[0089] Data collection and preprocessing:
[0090] Collect data on light source brightness adjustments from historical records, check whether the data contains missing values, outliers, or duplicates, and process them accordingly to format the data into a format that can be processed by the machine learning model;
[0091] Feature Engineering:
[0092] Extract features that affect the predicted light source brightness from the original data, convert the categorical features into numerical representations, and standardize or normalize the numerical features to eliminate the scale differences between different features;
[0093] Model training:
[0094] The data set is divided into training set, validation set and test set. The training set is used to train the decision tree model, the validation set is used to adjust the model parameters and hyperparameters, and the test set is used to evaluate the performance of the decision tree model. The training data is used to train the decision tree model.
[0095] Model evaluation and optimization:
[0096] The mean square error is selected to evaluate the performance of the model and measure the deviation between the brightness value predicted by the model and the true value. According to the evaluation results, the model is tuned, including adjusting the parameters of the model, changing the structure of the model, and trying different feature combinations.
[0097] The algorithm formula for the mean square error is:
[0098] text{MSE}=\frac{1}{n}\sum_{i=1}^{n}(y_i-\hat{y}_i)^2
[0099] Where n is the number of samples, y_i is the true value of the i-th sample, and hat{y}_i is the predicted value of the i-th sample;
[0100] Application and deployment:
[0101] The department deploys the trained model to make predictions on new data when needed. When encountering a new situation, relevant factors are input into the model, and the model will output a recommended brightness value. The performance of the model is monitored regularly.
[0102] In this embodiment, after adjustment and use, the historical data of the system adjustment process, such as light source brightness, light spot position, etc., is first stored through the data storage module 6 to provide data support for the data analysis module 7, and then the historical adjustment data is analyzed by the data analysis module 7 to extract valuable information, such as the optimal value of the light source brightness, adjustment trend, etc., to provide an analysis basis for the prediction and recommendation module 8. After that, the prediction and recommendation module 8 can predict the light source intensity and light spot position under corresponding conditions based on the results provided by the data analysis module 7, and then send instructions to the light source control module 2 and the mobile platform control module 5 to adjust the position of the light spot and the light source intensity, so that the appropriate brightness value and position can be quickly predicted and recommended based on historical data and machine learning, and the adjustment strategy can be continuously learned and optimized to gradually improve the accuracy of prediction and adjustment.
[0103] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A photonic crystal optical fiber axis fixing device, comprising an optical fiber axis fixing device body (9) and an optical fiber (10), characterized in that: A light spot detection module (3) is installed on the inner side of the optical fiber axis positioning device body (9), and a control system for the light spot detection module (3) is arranged on the optical fiber axis positioning device body (9), and the control system comprises: Initial setting module (1), responsible for the overall initialization setting of the system; A light spot detection module (3), used for capturing images and then analyzing the characteristics of the light spots; An automatic adjustment module (4) is used to calculate and send instructions for adjusting the light source intensity and the light spot position according to the light spot feature information provided by the light spot detection module (3); A mobile platform control module (5) receives instructions from the automatic adjustment module (4) and controls the movement of the mobile platform on the optical fiber axis fixing device body (9); A data storage module (6), used for storing historical data during the system adjustment process; A data analysis module (7), used for analyzing historical adjustment data; A prediction and recommendation module (8), used for predicting the optimal light source brightness value and light spot position based on the results provided by the data analysis module (7), and obtaining a recommended value; The light source control module (2) is responsible for receiving instructions from the automatic adjustment module (4) or the prediction and recommendation module (8) to control the brightness output of the light source.
2. A photonic crystal fiber axis fixing device according to claim 1, characterized in that: The light spot detection module (3) comprises a light spot analysis unit, an instruction generation unit and an instruction sending unit. The light spot analysis unit is responsible for receiving the light spot image feature information transmitted by the light spot detection module and performing in-depth analysis on the information. The instruction generation unit generates corresponding adjustment instructions based on the analysis results of the light spot analysis unit. The instruction sending unit is responsible for sending the adjustment instructions to the light source control module (2) and the mobile platform control module (5).
3. A photonic crystal fiber axis fixing device according to claim 2, characterized in that: The light spot analysis unit analyzes the characteristic information of the light spot according to the PID controller to determine whether the brightness of the light spot reaches a preset threshold and whether the position of the light spot deviates from a predetermined area. The calculation formula of the PID controller is as follows; Output value = Kp*e(t)+Ki*∫e(t)dt+Kd*de(t) / dt Among them, e(t) represents the gap between the target value and the current value, ∫e(t)dt represents the integral term of the deviation, de(t) / dt represents the differential term of the deviation, Kp, Ki and Kd are the proportional, integral and differential coefficients respectively, which are used to adjust the response speed, stability and anti-interference ability of the controller.
4. A photonic crystal fiber axis fixing device according to claim 1, characterized in that: The light source control module (2) comprises a brightness adjustment unit and a brightness control unit, the brightness adjustment unit is responsible for receiving instructions from the automatic adjustment module (4) or the prediction and recommendation module (8), and the brightness control unit is responsible for adjusting the brightness; the mobile platform control module (5) comprises an instruction receiving unit and a platform control unit, the instruction receiving unit is responsible for receiving instructions sent by the automatic adjustment module (4), and the platform control unit is responsible for controlling the movement of the mobile platform on the optical fiber axis fixing device body (9) according to the instructions.
5. The photonic crystal fiber axis fixing device according to claim 1, characterized in that: The data analysis module (7) comprises a data preprocessing unit, a data analysis unit and a result output unit. The data preprocessing unit is responsible for cleaning and sorting historical data, the data analysis unit is responsible for performing in-depth analysis on the data using an analysis algorithm, and the result output unit is responsible for outputting the analysis results to the prediction and recommendation module.
6. A photonic crystal fiber axis fixing device according to claim 5, characterized in that: The data preprocessing unit includes processing missing values, processing duplicate values, processing outliers, data standardization and normalization, and feature selection. The analysis algorithm of the data analysis unit is linear regression, and the formula of the linear regression is: b=(x1y1+x2y2+...xnyn-nXY) / (x1+x2+...xn-nX) Among them, b is the regression coefficient, X and Y are the average values of the independent variable and the dependent variable respectively, xi and yi are the specific values of each variable, and n is the number of data sets.
7. A photonic crystal fiber axis fixing device according to claim 1, characterized in that: The prediction and recommendation module (8) comprises a prediction algorithm unit, a recommendation generation unit and a recommendation sending unit, wherein the prediction algorithm unit is responsible for making predictions using a prediction algorithm, the recommendation generation unit is responsible for generating a recommendation value based on the prediction result of the prediction algorithm unit, and the recommendation sending unit is responsible for sending the recommendation value to the light source control module (2).
8. A photonic crystal fiber axis fixing device according to claim 7, characterized in that: The prediction algorithm unit performs prediction through a convolutional neural network, and the algorithm steps of the convolutional neural network are: Input data representation: The input data is a multidimensional array, denoted as X, where the dimensions of X are (batch size, number of channels, height, width), representing the number of input samples, number of channels, image height, and image width, respectively; Convolution operation: The convolution kernel is used to perform a sliding window convolution operation on the input data to extract the features in the input data. The dimension of the convolution kernel is (number of channels, convolution kernel height, convolution kernel width), denoted as W. The convolution operation can be expressed by the following formula: Z[i,j]=sum(sum(sum(X[:,:,i:i+h,j:j+w]*W)))+b Among them, Z is the output of the convolution operation, representing the feature map, i and j represent the position of the feature map, h and w represent the height and width of the convolution kernel respectively, and b is the bias term; Activation function: After the convolution operation, ReLU is applied to perform nonlinear transformation on the feature map to increase the network's expressive power; Pooling operation: By selecting the average value in the local perception area as the pooling result, the spatial dimension of the feature map is reduced and important feature information is retained; Fully connected operation: After multiple convolution and pooling operations, the feature map is expanded into a one-dimensional vector and passed through a fully connected layer for classification or regression tasks. Each neuron in the fully connected layer is connected to all neurons in the previous layer, and the mapping relationship between input and output is established by learning weights and biases. Loss function: In CNN, the cross entropy loss function is used to measure the gap between the model's prediction results and the true label; Back propagation algorithm: Use the back propagation algorithm to calculate the gradient of the loss function with respect to the model parameters, and use gradient descent to update the model parameters to continuously optimize the performance of the model; The recommendation generation unit makes recommendations through a machine learning model, and the implementation steps are as follows: Data collection and preprocessing: Collect data on light source brightness adjustments from historical records, check whether the data contains missing values, outliers, or duplicates, and process them accordingly to format the data into a format that can be processed by the machine learning model; Feature Engineering: Extract features that affect the predicted light source brightness from the original data, convert the categorical features into numerical representations, and standardize or normalize the numerical features to eliminate the scale differences between different features; Model training: The data set is divided into training set, validation set and test set. The training set is used to train the decision tree model, the validation set is used to adjust the model parameters and hyperparameters, and the test set is used to evaluate the performance of the decision tree model. The training data is used to train the decision tree model. Model evaluation and optimization: The mean square error is selected to evaluate the performance of the model and measure the deviation between the brightness value predicted by the model and the true value. According to the evaluation results, the model is tuned, including adjusting the parameters of the model, changing the structure of the model, and trying different feature combinations. The algorithm formula of the mean square error is: text{MSE}=\frac{1}{n}\sum_{i=1}^{n}(y_i-\hat{y}_i)^2 Where n is the number of samples, y_i is the true value of the i-th sample, and hat{y}_i is the predicted value of the i-th sample; Application and deployment: The department deploys the trained model to make predictions on new data when needed. When encountering new situations, relevant factors are input into the model, and the model will output a recommended brightness value. The performance of the model is monitored regularly.
Citation Information
Cited By
High-density and miniaturized connector
CN121899989A
