Intelligent early warning system for mold core deviation

Abstract

The present invention relates to the field of core offset monitoring technology, specifically to an intelligent early warning system for mold core offset, the system comprising: a core trajectory dynamic analysis module, a morphological feature extraction module, an offset dynamic identification module, a signal diagnosis enhancement module, an early warning judgment processing module, and a dynamic adjustment execution module. In the present invention, the cubic Bezier curve technology and the gated cyclic unit algorithm are used to perform continuous dynamic tracking and deep learning analysis of the core position changes, which can not only accurately identify the offset trend, but also issue an early warning and guide adjustments in time before the offset reaches a level that may affect product quality, thereby significantly reducing production defects and improving production efficiency and product quality. In addition, by combining real-time data analysis and historical trend comparison to set the early warning threshold, the logic and accuracy of the early warning judgment are enhanced, and the control capability and efficiency of the manufacturing process are further improved.

Description

Technical Field

[0001] The invention relates to the technical field of core deviation monitoring, and in particular to an intelligent early warning system for mold core deviation. Background Art

[0002] The core shift monitoring technology field is a high-precision technology focused on detecting and managing the core position shift in the mold. In the manufacturing industry, especially in high-precision manufacturing processes, the accurate position of the mold core is crucial to ensure product quality. The core shift monitoring technology field covers a range of sensor technologies, data processing algorithms and early warning mechanisms to monitor the position changes of the core in real time to ensure processing accuracy and product quality.

[0003] Among them, the mold core deviation intelligent early warning system is an automated system that uses advanced sensing technology and data analysis to predict and prevent core position deviation. Its main purpose is to accurately predict the deviation trend by real-time monitoring and analysis of the core position information, so as to issue an immediate warning and take corresponding measures before the deviation reaches a level that may affect product quality. Such a system aims to reduce or eliminate production defects caused by core position deviation, improve production efficiency and product quality, and ultimately achieve the effect of reducing production costs and improving the competitiveness of the manufacturing industry.

[0004] Traditional systems often rely on relatively simple monitoring methods and static threshold settings, lacking continuous dynamic analysis and deep learning capabilities for the core's motion trajectory, which leads to obvious deficiencies in identifying subtle deviation trends and dynamically adjusting the core position. Due to the lack of efficient morphological feature extraction and dynamic deviation identification mechanisms, traditional systems are difficult to accurately warn in the early stages of deviation, thereby increasing the defect rate and cost in the production process. For example, a slight core deviation that is not discovered and adjusted in time may lead to inconsistent finished product dimensions, affect product appearance and performance, and ultimately lead to a decrease in finished product qualification rate and customer dissatisfaction. In addition, due to the lack of the ability to comprehensively analyze historical and real-time data, the traditional early warning judgment logic has insufficient early warning efficiency and accuracy, further limiting the improvement of production efficiency and quality. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose an intelligent early warning system for mold core deviation.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme, the mold core deviation intelligent early warning system comprises:

[0007] The core trajectory dynamic analysis module performs noise filtering and data smoothing based on the mold core position data collected by the displacement sensor, and uses the cubic Bezier curve technology to generate continuous dynamic trajectory graphics according to the position data points;

[0008] The morphological feature extraction module performs edge detection and morphological transformation analysis based on the continuous dynamic trajectory graph, identifies key morphological features in the graphic trajectory, and obtains a key morphological feature set;

[0009] The displacement dynamics recognition module uses the gated recurrent unit algorithm to perform deep learning analysis based on the key morphological feature set, analyzes the displacement dynamics according to the morphological feature change trend, and obtains the displacement dynamics prediction analysis result;

[0010] The signal diagnosis enhancement module performs signal processing based on the deviation dynamic prediction analysis result, including signal amplification and noise filtering, to obtain an enhanced deviation signal;

[0011] The early warning determination processing module sets an early warning threshold based on the enhanced offset signal, combines real-time data analysis and historical trend comparison, determines whether the early warning condition is met, and generates early warning trigger parameters;

[0012] The dynamic adjustment execution module adjusts the operating parameters of the mold core and provides operator intervention guidance based on the early warning trigger parameters to obtain adjustment execution feedback.

[0013] As a further solution of the present invention, the continuous dynamic trajectory graph includes the trajectory starting point position, the trajectory end point position, and the trajectory shape change; the key morphological feature set includes the key curvature point, the target direction offset, and the closed area; the offset dynamic prediction analysis result includes the predicted offset time point, offset direction, and offset degree; the enhanced offset signal includes the signal amplitude, signal frequency range, and signal clarity index; the early warning trigger parameters include the early warning threshold, the trigger time window, and the trigger signal characteristics; the adjustment execution feedback includes the position deviation before and after adjustment, the adjustment response time, and the intervention measure effect feedback.

[0014] As a further solution of the present invention, the core trajectory dynamic analysis module includes:

[0015] The data acquisition submodule performs real-time acquisition of mold core position data based on displacement sensors, records the position information of each data point through time series analysis, and generates a position data set;

[0016] The noise filtering submodule performs frequency domain analysis and low-pass filtering based on the position data set to remove high-frequency noise in the data, and then restores the data to the time domain, performs smoothing processing, and generates a smoothed position data set;

[0017] The trajectory generation submodule executes the construction of continuous dynamic trajectory based on the smooth position data set, draws the continuous path of the core movement by using the cubic Bezier curve technology, describes the dynamic changes, and generates the continuous dynamic trajectory graphics;

[0018] The cubic Bezier curve technique uses the formula

[0019]

[0020] Compute continuous paths where is the parameter value on the curve, between 0 and 1, As a starting point, is the first control point, is the second control point, For the end point, are weight coefficients, which respectively adjust the influence of the starting point and the two control points on the curve shape.

[0021] As a further solution of the present invention, the morphological feature extraction module includes:

[0022] The trajectory graphic analysis submodule performs basic morphology and structural analysis of the graphic trajectory based on the continuous dynamic trajectory graphic, analyzes the structured representation of the trajectory data through graphic drawing and array operation, analyzes the basic morphology of the trajectory, extracts structural features, and generates trajectory structure information;

[0023] The edge detection submodule processes the graphics using an edge detection algorithm based on the trajectory structure information, highlights the edge features of the graphics, identifies the boundaries of the trajectory graphics through edge enhancement processing, and generates edge feature information;

[0024] The morphological transformation analysis submodule performs morphological transformation analysis based on the edge feature information, strengthens the connection between graphic features and eliminates interference, identifies and extracts key morphological features in the graphic trajectory, and generates a key morphological feature set.

[0025] As a further solution of the present invention, the offset dynamic identification module includes:

[0026] The morphological feature extraction submodule performs grayscale conversion and edge detection on the input image based on the key morphological feature set, extracts key points and descriptors of the image, and generates a morphological feature data set;

[0027] The dynamic trend analysis submodule performs time series processing of the data based on the morphological feature data set, analyzes the time series data of the morphological features, performs trend change analysis, and generates trend change analysis results;

[0028] The offset prediction analysis submodule uses a time series prediction analysis method based on the trend change analysis results to construct a gated recurrent unit network, define a model structure, perform prediction analysis of future offset dynamics, and generate offset dynamic prediction analysis results;

[0029] The time series forecasting analysis method adopts the formula

[0030]

[0031] in, is the hidden state at the current moment, To update the gate, is the hidden state at the previous moment, is the candidate hidden state at the current moment, is the weight coefficient for adjusting the hidden state at the previous moment, is the weight coefficient for strengthening the current candidate hidden state, is the input weight matrix, is the input at the current moment, is the introduced nonlinear adjustment coefficient.

[0032] As a further solution of the present invention, the signal diagnosis enhancement module includes:

[0033] The signal amplification submodule amplifies the signal based on the deviation dynamic prediction analysis result, adjusts the amplification factor to match the signal characteristics, enhances the target frequency band in the signal, and generates an amplified deviation signal;

[0034] The noise filtering submodule applies noise filtering technology based on the amplified offset signal, adopts a filtering algorithm to remove non-target noise in the signal, retains key signal components, identifies and isolates useful signals from background noise, and generates a filtered offset signal;

[0035] The signal integration submodule performs signal integration analysis based on the filtered offset signal, optimizes signal quality through data fusion technology, merges multiple signal source data, and generates an enhanced offset signal.

[0036] As a further solution of the present invention, the early warning determination processing module includes:

[0037] The signal analysis submodule analyzes the signal based on the enhanced offset signal, identifies the basic properties and key indicators of the signal, highlights the key information through signal feature extraction, and obtains the signal feature extraction result;

[0038] The trend identification submodule performs trend analysis on the real-time data based on the signal feature extraction result, compares it with the historical data, identifies the data fluctuation or deviation trend, and generates trend comparison analysis results;

[0039] The threshold evaluation submodule evaluates the warning threshold based on the trend comparison analysis results to determine whether the current data meets the preset warning conditions. If so, the warning mechanism is activated to generate warning trigger parameters.

[0040] As a further solution of the present invention, the dynamic adjustment execution module includes:

[0041] The early warning trigger parameter analysis submodule analyzes potential anomalies in the operation state of the mold core based on the early warning trigger parameters, uses data analysis methods to screen key operation indicators, identifies parameters that deviate from the normal range, and generates early warning trigger analysis results;

[0042] The operating parameter adjustment submodule dynamically adjusts the operating parameters of the mold core based on the early warning trigger analysis results, optimizes the parameter settings, including adjusting the temperature, pressure and speed, matching the differentiated operating conditions, and generating adjusted parameter results;

[0043] The operator intervention guidance submodule formulates operator intervention guidance based on the adjusted parameter results, guides the operator to perform manual intervention and equipment adjustment according to the adjustment plan, and generates intervention guidance results;

[0044] The adjustment execution feedback submodule collects and analyzes feedback information from operators after they follow the instructions based on the intervention guidance results, integrates the data of the execution operations, and obtains adjustment execution feedback.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are:

[0046] In the present invention, the cubic Bezier curve technology and gated cyclic unit algorithm are used to continuously and dynamically track the core position changes and conduct deep learning analysis, which can not only accurately identify the deviation trend, but also issue early warnings and guide adjustments in time before the deviation reaches a level that may affect product quality, thereby significantly reducing production defects and improving production efficiency and product quality. In addition, by combining real-time data analysis and historical trend comparison to set the early warning threshold, the logic and accuracy of the early warning judgment are enhanced, further improving the control ability and efficiency of the manufacturing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a system flow chart of the present invention;

[0048] Figure 2 It is a schematic diagram of the system framework of the present invention;

[0049] Figure 3 It is a flow chart of the core trajectory dynamic analysis module of the present invention;

[0050] Figure 4 It is a flow chart of the morphological feature extraction module of the present invention;

[0051] Figure 5 This is a flow chart of the offset dynamic identification module of the present invention;

[0052] Figure 6 This is a flow chart of the signal diagnosis enhancement module of the present invention;

[0053] Figure 7This is a flow chart of the early warning determination processing module of the present invention;

[0054] Figure 8 This is a flow chart of the dynamic adjustment execution module of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0056] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined. Embodiment 1

[0057] See also Figure 1 to Figure 2 , the mold core deviation intelligent warning system includes:

[0058] The core trajectory dynamic analysis module performs noise filtering and data smoothing based on the mold core position data collected by the displacement sensor, and uses the cubic Bezier curve technology to generate continuous dynamic trajectory graphics according to the position data points;

[0059] The morphological feature extraction module performs edge detection and morphological transformation analysis based on the continuous dynamic trajectory graphics, identifies the key morphological features in the graphics trajectory, and obtains the key morphological feature set;

[0060] The displacement dynamics recognition module uses the gated recurrent unit algorithm to perform deep learning analysis based on the key morphological feature set, analyzes the displacement dynamics according to the changing trend of the morphological features, and obtains the displacement dynamics prediction analysis results;

[0061] The signal diagnosis enhancement module performs signal processing based on the dynamic prediction analysis results of the offset, including signal amplification and noise filtering, to obtain an enhanced offset signal;

[0062] The early warning judgment processing module sets the early warning threshold based on the enhanced offset signal, combines real-time data analysis and historical trend comparison, determines whether the early warning conditions are met, and generates early warning trigger parameters;

[0063] The dynamic adjustment execution module adjusts the operating parameters of the mold core based on the early warning trigger parameters and provides operator intervention guidance to obtain adjustment execution feedback.

[0064] The continuous dynamic trajectory graphics include the trajectory starting point position, trajectory end point position, and trajectory shape change. The key morphological feature set includes key curvature points, target direction offset, and closed area. The dynamic prediction analysis results of the offset include the predicted offset time point, offset direction, and offset degree. The enhanced offset signal includes signal amplitude, signal frequency range, and signal clarity index. The early warning trigger parameters include early warning threshold, trigger time window, and trigger signal characteristics. The adjustment execution feedback includes adjustment of the position deviation before and after, adjustment of response time, and feedback on the effectiveness of intervention measures.

[0065] See also Figure 3 , the core trajectory dynamic analysis module includes:

[0066] The data acquisition submodule performs real-time acquisition of mold core position data based on displacement sensors, records the position information of each data point through time series analysis, and generates a position data set;

[0067] The data acquisition submodule is based on displacement sensors and adopts real-time data acquisition technology. It uses the Python programming language combined with the Pandas library to perform time series analysis on the collected data. It applies a custom position information acquisition function to each row of data through the DataFrame.apply function, sets the sampling frequency to 100 Hz, realizes real-time recording of data point position information, and performs array processing of position information to generate a position data set.

[0068] The noise filtering submodule performs frequency domain analysis and low-pass filtering based on the position data set to remove high-frequency noise in the data, and then restores the data to the time domain and performs smoothing to generate a smoothed position data set;

[0069] The noise filtering submodule is based on the position data set and adopts the Fourier transform method. It uses the fft and ifft functions in the NumPy library to perform frequency domain analysis, performs fast Fourier transform on the position data set to convert it to the frequency domain, sets the low-pass filter threshold to 500 Hz, uses the logical_and function combined with the fftshift function to remove high-frequency noise from the frequency domain data, and uses the ifft function to restore the data to the time domain. It uses the savgol_filter function of the signal module in the SciPy library for smoothing, sets the window length to 51 and the polynomial order to 3 to remove high-frequency noise in the data, and performs smoothing on the time domain data to generate a smoothed position data set.

[0070] The trajectory generation submodule performs the construction of continuous dynamic trajectory based on the smooth position data set, draws the continuous path of core movement using cubic Bezier curve technology, describes the dynamic changes, and generates continuous dynamic trajectory graphics;

[0071] The trajectory generation submodule is based on the smooth position data set, adopts the cubic Bezier curve technology, uses the Matplotlib library for graphics drawing, constructs continuous dynamic trajectories for the data points in the smooth position data set, sets the control point calculation formula of the Bezier curve, draws the continuous path of the core movement through the BezierPath class and its add_patch function, describes the dynamic changes, and generates continuous dynamic trajectory graphics.

[0072] The cubic Bezier curve technique uses the formula

[0073]

[0074] Compute continuous paths where is the parameter value on the curve, between 0 and 1, As a starting point, is the first control point, is the second control point, For the end point, are weight coefficients, which respectively adjust the influence of the starting point and the two control points on the curve shape.

[0075] first, As a curve parameter, it determines the position of each point on the curve throughout the whole process. Its value changes from 0 to 1, indicating that the position of each point on the curve is from the starting point to the end. To the end process. , , , These four points represent the starting point, the first control point, the second control point, and the end point. They are the basis for constructing a Bezier curve. These points can be used to determine the approximate direction and shape of the curve. Next, the weight coefficient is introduced. They are used to adjust the starting point and the influence of the two control points on the curve shape. The introduction of these three coefficients provides the possibility of fine-tuning the curve, making the curve more suitable for the actual trajectory. By adjusting the control point, you can control the curvature of the curve near the starting point, when passing through the control point, and when approaching the end point, thereby achieving more complex and varied trajectory shapes.

[0076] See also Figure 4 , the morphological feature extraction module includes:

[0077] The trajectory graphics analysis submodule performs basic morphology and structural analysis of the graphic trajectory based on the continuous dynamic trajectory graphics. It analyzes the structured representation of the trajectory data through graphics drawing and array operations, analyzes the basic morphology of the trajectory, extracts structural features, and generates trajectory structure information.

[0078] The trajectory graphics analysis submodule is based on continuous dynamic trajectory graphics. It adopts a graphic structured analysis method, uses the Python programming language combined with the OpenCV library for graphics drawing and array operations, grays the continuous dynamic trajectory graphics, converts the graphics into a grayscale image through the cv2.cvtColor function, and then uses the cv2.findContours function to identify the contour of the graphic trajectory. The structured representation of the trajectory data is analyzed, and the trajectory contour is approximated by polygons through the cv2.approxPolyDP function. The approximation accuracy ratio is set to 0.01 to extract structural features and generate trajectory structure information.

[0079] The edge detection submodule processes the graphics based on the trajectory structure information and applies the edge detection algorithm to highlight the edge features of the graphics. It identifies the boundaries of the trajectory graphics through edge enhancement processing and generates edge feature information.

[0080] The edge detection submodule uses the Canny edge detection algorithm based on the trajectory structure information and the cv2.Canny function of the OpenCV library to process the graphics. The low threshold is set to 100 and the high threshold is set to 200. The edge features of the graphics are identified by setting the high and low thresholds, and edge enhancement processing is performed to identify the boundaries of the trajectory graphics and generate edge feature information.

[0081] The morphological transformation analysis submodule performs morphological transformation analysis based on edge feature information, strengthens the connection between graphic features and eliminates interference, identifies and extracts key morphological features in the graphic trajectory, and generates a key morphological feature set;

[0082] The morphological transformation analysis submodule is based on edge feature information and adopts morphological operation methods. It uses the cv2.dilate function and cv2.erode function in the OpenCV library to perform morphological transformation analysis, sets the kernel shape to a rectangle, and the kernel size to a 5x5 pixel structural element. The structural element is created through the cv2.getStructuringElement function. First, the dilation operation is applied to strengthen the connection between graphic features, and then the erosion operation is used to eliminate interference, so as to identify and extract the key morphological features in the graphic trajectory and generate a set of key morphological features.

[0083] See also Figure 5 , the offset dynamic recognition module includes:

[0084] The morphological feature extraction submodule performs grayscale conversion and edge detection on the input image based on the key morphological feature set, extracts the key points and descriptors of the image, and generates a morphological feature dataset;

[0085] The morphological feature extraction submodule is based on the key morphological feature set and adopts digital image processing technology. It uses the Python programming language combined with the OpenCV library to convert the input image into grayscale. The grayscale processing is performed through the cv2.cvtColor function, and the conversion code is set to cv2.COLOR_BGR2GRAY. Then the SIFT (Scale Invariant Feature Transform) algorithm is adopted. The SIFT detector is initialized using the cv2.SIFT_create function. The key points and descriptors of the grayscale image are extracted. This operation is performed through the detectAndCompute function. There is no need to preset the threshold. The key points and descriptors of the image are directly extracted to generate a morphological feature dataset.

[0086] The dynamic trend analysis submodule performs time series processing of data based on the morphological feature data set, analyzes the time series data of morphological features, performs trend change analysis, and generates trend change analysis results;

[0087] The dynamic trend analysis submodule is based on the morphological feature dataset and uses time series analysis technology. It uses the Python programming language combined with the Pandas library to perform time series processing of the data. The morphological feature dataset is stored in the DataFrame structure and the timestamp is used as the index. The moving average method is then used to set the window size to 5 using the rolling function to calculate the average value within the time window, perform trend change analysis, and use the plot function to draw a trend graph of the morphological feature changes over time to generate trend change analysis results.

[0088] Based on the trend change analysis results, the offset prediction analysis submodule adopts the time series prediction analysis method to construct a gated recurrent unit network, define the model structure, perform prediction analysis of future offset dynamics, and generate offset dynamic prediction analysis results;

[0089] The offset prediction analysis submodule builds a gated recurrent unit network based on the trend change analysis results. It uses the Python programming language combined with the TensorFlow and Keras libraries to define the model structure. It uses the Sequential model to add a GRU layer, sets the number of units to 128, and selects relu as the activation function to capture the long-term dependencies of time series data. Then it adds a Dense layer as the output layer, sets the number of output units to 1, and selects linear as the activation function to perform future offset dynamics prediction analysis. It sets the optimizer to Adam and the loss function to mean_squared_error. It trains the model through the fit function, sets the batch size to 32, and the training rounds to 100 to generate offset dynamics prediction analysis results.

[0090] The time series forecasting analysis method uses the formula

[0091]

[0092] in, is the hidden state at the current moment, To update the gate, is the hidden state at the previous moment, is the candidate hidden state at the current moment, is the weight coefficient for adjusting the hidden state at the previous moment, is the weight coefficient for strengthening the current candidate hidden state, is the input weight matrix, is the input at the current moment, is the introduced nonlinear adjustment coefficient.

[0093] By introducing the weight coefficient and , and the nonlinear adjustment coefficient And the input weight matrix , improve the model's adaptability to the characteristics of time series data and prediction accuracy. Among them, Adjust the influence of the previous hidden state. Strengthen the role of the current candidate hidden state, Combination Function and input weight matrix Introducing additional nonlinear characteristics enhances the model's ability to capture the complex dynamics of time series. The specific methodological steps for confirming the coefficient values ​​include using cross-validation to optimize , , The value of ensures that the model can show the best performance on different data sets. The determination of depends on the dimension of the input data and the specific configuration of the network structure.

[0094] See also Figure 6 , the signal diagnosis enhancement module includes:

[0095] The signal amplification submodule amplifies the signal based on the dynamic prediction analysis results of the offset, adjusts the amplification factor to match the signal characteristics, enhances the target frequency band in the signal, and generates an amplified offset signal;

[0096] The signal amplification submodule is based on the results of dynamic prediction analysis of the offset and adopts digital signal processing technology. It uses the Python programming language combined with the SciPy library to amplify the signal. The signal.lfilter function performs low-pass filtering to enhance the target frequency band in the signal. The filter coefficient is set to a 5th-order low-pass filter using the signal.butter function, and the cutoff frequency is set to 0.5 times the Nyquist frequency to adjust the amplification factor to match the signal characteristics, enhance the target frequency band in the signal, and generate an amplified offset signal.

[0097] The noise filtering submodule is based on the amplified offset signal, applies noise filtering technology, uses filtering algorithms to remove non-target noise from the signal, retains key signal components, identifies and isolates useful signals from background noise, and generates filtered offset signals;

[0098] The noise filtering submodule is based on the amplified offset signal, applies noise filtering technology, and uses the Gaussian filtering algorithm to remove non-target noise in the signal. It uses the cv2.GaussianBlur function of the OpenCV library for processing, sets the kernel size to (5,5) and the standard deviation to 0, and removes high-frequency noise in the signal by adjusting the kernel size and standard deviation, retains key signal components, identifies and isolates useful signals from background noise, and generates a filtered offset signal.

[0099] The signal integration submodule performs signal integration analysis based on the filtered offset signal, optimizes the signal quality through data fusion technology, merges multiple signal source data, and generates an enhanced offset signal;

[0100] The signal integration submodule performs signal integration analysis based on filtered offset signals, optimizes signal quality through data fusion technology, and uses weighted averaging to merge multiple signal source data. It is implemented in Python programming language and uses the np.average function of the numpy library to perform weighted averaging on multiple signal sources. The weights are set to dynamically adjust according to the signal quality to optimize the quality of the integrated signal and generate an enhanced offset signal.

[0101] See also Figure 7 , the early warning judgment processing module includes:

[0102] The signal analysis submodule analyzes the signal based on the enhanced offset signal, identifies the basic properties and key indicators of the signal, highlights the key information through signal feature extraction, and obtains the signal feature extraction result;

[0103] The signal analysis submodule is based on enhanced offset signals and adopts fast Fourier transform (FFT) technology. It uses Python programming language combined with NumPy library to analyze the signal. The fast Fourier transform of the signal is performed through np.fft.fft function to convert the signal from time domain to frequency domain to identify the basic properties and key indicators of the signal. Then, the peak detection algorithm is used to identify the main frequency components in the frequency domain using the signal.find_peaks function in the SciPy library. The minimum height and distance parameters are set to highlight the key information and generate signal feature extraction results.

[0104] The trend identification submodule performs trend analysis on real-time data based on the signal feature extraction results, compares it with historical data, identifies data fluctuation or deviation trends, and generates trend comparison analysis results;

[0105] The trend identification submodule uses the autoregressive moving average (ARMA) model based on the signal feature extraction results, and uses the Python programming language combined with the Statsmodels library to perform trend analysis on real-time data. The model is constructed through the ARMA function, and the order parameter of the model is set to (2,1) to match the time series characteristics of the data. The model fitting process is executed, and the real-time data is compared with the historical data to identify data fluctuations or deviation trends, and generate trend comparison analysis results.

[0106] The threshold evaluation submodule evaluates the warning threshold based on the trend comparison analysis results to determine whether the current data reaches the preset warning conditions. If so, the warning mechanism is activated and warning trigger parameters are generated.

[0107] The threshold assessment submodule is based on the trend comparison analysis results and adopts a dynamic threshold assessment method. It is implemented in the Python programming language. By calculating the standard deviation and average value of real-time data, the dynamic threshold is set to the average value plus two times the standard deviation. The current data is evaluated to determine whether the preset warning conditions are met. If the current data exceeds the dynamic threshold, the warning mechanism is activated and the warning trigger parameters are generated.

[0108] See also Figure 8 , the dynamic adjustment execution module includes:

[0109] The early warning trigger parameter analysis submodule analyzes potential anomalies in the operation status of the mold core based on the early warning trigger parameters, uses data analysis methods to screen key operation indicators, identifies parameters that deviate from the normal range, and generates early warning trigger analysis results;

[0110] The early warning trigger parameter analysis submodule adopts statistical analysis methods based on the early warning trigger parameters, uses the Python programming language combined with the Pandas library to analyze the data in the mold core operation status, and calculates the statistical summary of key operating indicators through the DataFrame.describe function, including the mean, standard deviation, minimum and maximum values, to screen out parameters that deviate from the normal range. The box plot analysis method is further used to use the boxplot function of the matplotlib library to visualize the key operating indicators, identify outliers, and generate early warning trigger analysis results.

[0111] The operating parameter adjustment submodule dynamically adjusts the operating parameters of the mold core based on the early warning trigger analysis results, optimizes the parameter settings, including adjusting the temperature, pressure and speed, matching the differentiated operating conditions, and generating the adjusted parameter results;

[0112] The operating parameter adjustment submodule dynamically adjusts the operating parameters of the mold core based on the early warning trigger analysis results, uses the PID control algorithm to optimize the parameter settings, uses the Python programming language to implement the PID controller, and sets the proportional coefficient K p , integral coefficient K i , differential coefficient K d They are 1.0, 0.1, and 0.01 respectively. Dynamic adjustments are made to the temperature, pressure, and speed parameters. By calculating the deviation value and adjustment amount, the operating parameters are automatically adjusted to match the differentiated operating conditions and generate the adjusted parameter results.

[0113] The operator intervention guidance submodule formulates operator intervention guidance based on the adjusted parameter results, guides the operator to perform manual intervention and equipment adjustment according to the adjustment plan, and generates intervention guidance results;

[0114] The operator intervention guidance submodule formulates operator intervention guidelines based on the adjusted parameter results, adopts the decision support system method to guide the operator to perform manual intervention and equipment adjustment according to the adjustment plan, and uses the Python programming language combined with a simple rule engine to define rules such as increasing the cooling rate if the temperature deviation is greater than the set threshold. It provides operators with specific intervention steps and adjustment suggestions, and generates intervention guidance results.

[0115] The adjustment execution feedback submodule collects and analyzes the feedback information of operators after they follow the instructions based on the intervention guidance results, integrates the data of the execution operations, and obtains the adjustment execution feedback;

[0116] The adjustment execution feedback submodule collects and analyzes the feedback information from operators after they follow the instructions based on the results of the intervention guidance. It adopts the feedback loop analysis method and uses the Python programming language to integrate the data. It evaluates the effectiveness of the adjustment by comparing the operating parameters and equipment performance data before and after the adjustment, integrates the data of the execution operation, and obtains the adjustment execution feedback.

[0117] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. Intelligent early warning system for mold core deviation, characterized by: The system comprises: The core trajectory dynamic analysis module performs noise filtering and data smoothing based on the mold core position data collected by the displacement sensor, and uses the cubic Bezier curve technology to generate continuous dynamic trajectory graphics according to the position data points; The morphological feature extraction module performs edge detection and morphological transformation analysis based on the continuous dynamic trajectory graph, identifies key morphological features in the graphic trajectory, and obtains a key morphological feature set; The displacement dynamics recognition module uses the gated recurrent unit algorithm to perform deep learning analysis based on the key morphological feature set, analyzes the displacement dynamics according to the morphological feature change trend, and obtains the displacement dynamics prediction analysis result; The signal diagnosis enhancement module performs signal processing based on the deviation dynamic prediction analysis result, including signal amplification and noise filtering, to obtain an enhanced deviation signal; The early warning determination processing module sets an early warning threshold based on the enhanced offset signal, combines real-time data analysis and historical trend comparison, determines whether the early warning condition is met, and generates early warning trigger parameters; The dynamic adjustment execution module adjusts the operating parameters of the mold core and provides operator intervention guidance based on the early warning trigger parameters to obtain adjustment execution feedback; The displacement dynamic identification module comprises: The morphological feature extraction submodule performs grayscale conversion and edge detection on the input image based on the key morphological feature set, extracts key points and descriptors of the image, and generates a morphological feature data set; The dynamic trend analysis submodule performs time series processing of the data based on the morphological feature data set, analyzes the time series data of the morphological features, performs trend change analysis, and generates trend change analysis results; The displacement prediction analysis submodule uses a time series prediction analysis method based on the trend change analysis results to construct a gated recurrent unit network, define a model structure, perform prediction analysis of future displacement dynamics, and generate displacement dynamic prediction analysis results; The time series forecasting analysis method adopts the formula ； in, is the hidden state at the current moment, To update the gate, is the hidden state at the previous moment, is the candidate hidden state at the current moment, is the weight coefficient for adjusting the hidden state at the previous moment, is the weight coefficient for strengthening the current candidate hidden state, is the input weight matrix, is the input at the current moment, is the introduced nonlinear adjustment coefficient.

2. The mold core deviation intelligent early warning system according to claim 1 is characterized in that: The continuous dynamic trajectory graph includes the trajectory starting point position, the trajectory end point position, and the trajectory shape change; the key morphological feature set includes the key curvature point, the target direction offset, and the closed area; the offset dynamic prediction analysis result includes the predicted offset time point, offset direction, and offset degree; the enhanced offset signal includes the signal amplitude, signal frequency range, and signal clarity index; the early warning trigger parameters include the early warning threshold, the trigger time window, and the trigger signal characteristics; the adjustment execution feedback includes the position deviation before and after adjustment, the adjustment response time, and the intervention measure effect feedback.

3. The mold core deviation intelligent early warning system according to claim 1 is characterized in that: The core trajectory dynamic analysis module includes: The data acquisition submodule performs real-time acquisition of mold core position data based on displacement sensors, records the position information of each data point through time series analysis, and generates a position data set; The noise filtering submodule performs frequency domain analysis and low-pass filtering based on the position data set to remove high-frequency noise in the data, and then restores the data to the time domain, performs smoothing processing, and generates a smoothed position data set; The trajectory generation submodule executes the construction of continuous dynamic trajectory based on the smooth position data set, draws the continuous path of the core movement by using the cubic Bezier curve technology, describes the dynamic changes, and generates the continuous dynamic trajectory graphics; The cubic Bezier curve technique uses the formula ； Compute continuous paths where is the parameter value on the curve, between 0 and 1, As a starting point, is the first control point, is the second control point, For the end point, are weight coefficients, which respectively adjust the influence of the starting point and the two control points on the curve shape.

4. The mold core deviation intelligent early warning system according to claim 1, characterized in that: The morphological feature extraction module comprises: The trajectory graphic analysis submodule performs basic morphology and structural analysis of the graphic trajectory based on the continuous dynamic trajectory graphic, analyzes the structured representation of the trajectory data through graphic drawing and array operation, analyzes the basic morphology of the trajectory, extracts structural features, and generates trajectory structure information; The edge detection submodule processes the graphics using an edge detection algorithm based on the trajectory structure information, highlights the edge features of the graphics, identifies the boundaries of the trajectory graphics through edge enhancement processing, and generates edge feature information; The morphological transformation analysis submodule performs morphological transformation analysis based on the edge feature information, strengthens the connection between graphic features and eliminates interference, identifies and extracts key morphological features in the graphic trajectory, and generates a key morphological feature set.

5. The mold core deviation intelligent early warning system according to claim 1, characterized in that: The signal diagnosis enhancement module comprises: The signal amplification submodule amplifies the signal based on the deviation dynamic prediction analysis result, adjusts the amplification factor to match the signal characteristics, enhances the target frequency band in the signal, and generates an amplified deviation signal; The noise filtering submodule applies noise filtering technology based on the amplified offset signal, adopts a filtering algorithm to remove non-target noise in the signal, retains key signal components, identifies and isolates useful signals from background noise, and generates a filtered offset signal; The signal integration submodule performs signal integration analysis based on the filtered offset signal, optimizes signal quality through data fusion technology, merges multiple signal source data, and generates an enhanced offset signal.

6. The mold core deviation intelligent early warning system according to claim 1, characterized in that: The early warning determination processing module includes: The signal analysis submodule analyzes the signal based on the enhanced offset signal, identifies the basic properties and key indicators of the signal, highlights the key information through signal feature extraction, and obtains the signal feature extraction result; The trend identification submodule performs trend analysis on the real-time data based on the signal feature extraction result, compares it with the historical data, identifies the data fluctuation or deviation trend, and generates trend comparison analysis results; The threshold evaluation submodule evaluates the warning threshold based on the trend comparison analysis results to determine whether the current data reaches the preset warning condition. If so, the warning mechanism is activated to generate warning trigger parameters.

7. The mold core deviation intelligent early warning system according to claim 1, characterized in that: The dynamic adjustment execution module includes: The early warning trigger parameter analysis submodule analyzes potential anomalies in the operation state of the mold core based on the early warning trigger parameters, uses data analysis methods to screen key operation indicators, identifies parameters that deviate from the normal range, and generates early warning trigger analysis results; The operating parameter adjustment submodule dynamically adjusts the operating parameters of the mold core based on the early warning trigger analysis results, optimizes the parameter settings, including adjusting the temperature, pressure and speed, matching the differentiated operating conditions, and generating adjusted parameter results; The operator intervention guidance submodule formulates operator intervention guidance based on the adjusted parameter results, guides the operator to perform manual intervention and equipment adjustment according to the adjustment plan, and generates intervention guidance results; The adjustment execution feedback submodule collects and analyzes feedback information from operators after they follow the instructions based on the intervention guidance results, integrates the data of the execution operations, and obtains adjustment execution feedback.

Images