Fault diagnosis and prediction maintenance method and device for intelligent injection mold

Abstract

The invention relates to the technical field of mold injection molding production, and discloses a fault diagnosis and prediction maintenance method and device for an intelligent injection mold, and the method comprises the steps: obtaining an original sound vibration signal and a real-time sound vibration signal; performing signal decomposition operation on the original acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; performing feature extraction operation on the sound vibration high-frequency component and the sound vibration low-frequency component to obtain a high-frequency feature set and a low-frequency feature set; inputting the high-frequency feature set and the low-frequency feature set into a detection model for model training to obtain an original detection model; inputting data acquired by different injection molding machines and molds into the original detection model for generalization learning to obtain a generalization detection model; and the real-time sound vibration signals are input into the detection model for mold state reasoning, a fault alarm type and a mold fault position are obtained, and fault diagnosis and mold maintenance of the injection mold are conducted according to the fault alarm type and the mold fault position. The method can realize fault diagnosis and maintenance of the injection mold.

Description

Technical Field

[0001] The present invention relates to the technical field of mold injection production, and in particular to a fault diagnosis and predictive maintenance method and device for an intelligent injection mold. Background Art

[0002] At present, intelligent injection molding is a widely used processing method in modern manufacturing industry, and is widely used in many fields such as automobile, electronics, medical, packaging, etc. As the core equipment of injection molding, the performance and reliability of intelligent injection molds directly affect the quality, production efficiency and cost of plastic products.

[0003] In existing injection molding technology, molds are affected by extreme working conditions such as high temperature, high pressure, and high speed. Data collection and stable signal transmission will affect the judgment of on-site personnel when mold failure occurs. At the same time, for some early failures and potential failures, their signal characteristics are not obvious, and they are easy to miss or misjudge.

[0004] There is a problem of inaccurate diagnosis of injection mold faults in the prior art. Summary of the invention

[0005] The present invention provides a fault diagnosis and predictive maintenance method and device for an intelligent injection mold, so as to solve the problem of inaccurate fault diagnosis of the injection mold.

[0006] In the first aspect, in order to solve the above technical problems, the present invention provides a fault diagnosis and predictive maintenance method for an intelligent injection mold, comprising: Obtaining original acoustic vibration signals and real-time acoustic vibration signals; Decomposing the original acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; Perform feature extraction operations on the acoustic vibration high-frequency component and the acoustic vibration low-frequency component respectively to obtain a feature set, wherein the feature set includes a high-frequency feature set and a low-frequency feature set; Inputting the feature set into a preset detection model for model training to obtain an original detection model, wherein the original detection model includes a high-frequency sub-model and a low-frequency sub-model; Inputting the data collected from injection molding machines and molds of different models into the original detection model for generalization learning to obtain a generalized detection model; The real-time acoustic vibration signal is input into the generalized detection model to perform mold state inference, obtain the fault alarm type and the mold fault location, and perform injection mold fault diagnosis and mold maintenance according to the fault alarm type and the mold fault location. In an optional embodiment, the original acoustic vibration signal is subjected to a signal decomposition operation to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component, including: The original acoustic vibration signal is subjected to signal preprocessing and frequency band calibration to obtain a filtered acoustic vibration signal; Perform signal discrete decomposition according to the filtered acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; Among them, the high-frequency component of the acoustic vibration is used to detect tiny cracks inside the mold, and the low-frequency component of the acoustic vibration is used to analyze hydraulic system abnormalities.

[0007] In an optional implementation, the acoustic vibration high-frequency component and the acoustic vibration low-frequency component are subjected to feature extraction operations respectively to obtain a feature set, including: Extract the time domain features of the high-frequency component and the low-frequency component of the acoustic vibration to obtain the high-frequency time domain features and the low-frequency time domain features; Extract the frequency domain features of the high-frequency component and the low-frequency component of the acoustic vibration to obtain the high-frequency frequency domain features and the low-frequency frequency domain features; The high-frequency time domain features and the high-frequency frequency domain features, and the low-frequency time domain features and the low-frequency frequency domain features are standardized and merged to obtain a feature set.

[0008] In an optional implementation, the step of inputting the feature set into a preset detection model for model training to obtain an original detection model includes: Pre-build and train high-frequency detection models and low-frequency detection models using historical data; Inputting the high-frequency feature set into the high-frequency detection model to perform feature dimension verification to obtain a high-frequency sub-model; Inputting the low-frequency feature set into the low-frequency detection model to perform feature dimension verification to obtain a low-frequency sub-model; The high-frequency sub-model and the low-frequency sub-model are loaded into the memory to obtain the original detection model.

[0009] In an optional implementation, data collected from injection molding machines and molds of different models are input into the original detection model for generalization learning to obtain a generalized detection model, including: Sensors transmit detection data from different injection molding machines and molds to computers; The computer inputs the detection data into the original detection model for sample detection, and fine-tunes the model parameters to obtain a generalized detection model.

[0010] In an optional implementation, the inputting the real-time acoustic vibration signal into a generalized detection model to perform mold state inference to obtain a fault alarm type and a mold fault location includes: Obtain real-time acoustic and vibration signals of the mold through edge devices; Preprocessing the real-time acoustic vibration signal to obtain a preprocessed acoustic vibration signal; The preprocessed acoustic vibration signal is input into the generalized detection model to perform mold state inference. If the output result of the generalized detection model is a normal state, the detection continues. If the output result is an abnormal state, the time series analysis method is used to process the abnormal signal and output the fault alarm type and mold fault location.

[0011] In an optional implementation, if the output result is an abnormal state, a time series analysis method is used to process the abnormal signal, and the fault alarm type and mold fault location are output, including: The periodicity and trend characteristics of abnormal signals are extracted through time series analysis methods to obtain signal feature descriptions; By describing the signal characteristics and combining with a preset fault type library, a matching fault type is obtained; Fault diagnosis is performed according to the matching fault type to obtain the fault alarm type and the mold fault location.

[0012] In a second aspect, the present invention provides a fault diagnosis and predictive maintenance device for an intelligent injection mold, comprising: A data acquisition module, used to acquire original acoustic vibration signals and real-time acoustic vibration signals; A signal decomposition module, used for performing a signal decomposition operation on the original acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; A feature extraction module, used to perform feature extraction operations on the acoustic vibration high-frequency component and the acoustic vibration low-frequency component respectively to obtain a feature set, wherein the feature set includes a high-frequency feature set and a low-frequency feature set; A feature conversion module, used for inputting the feature set into a preset detection model for model training to obtain an original detection model, wherein the original detection model includes a high-frequency sub-model and a low-frequency sub-model; A generalized learning module, used for inputting data collected from injection molding machines and molds of different models into the original detection model for generalized learning to obtain a generalized detection model; The mold detection module is used to input the real-time acoustic vibration signal into the generalized detection model to perform mold state reasoning, obtain the fault alarm type and the mold fault location, and perform injection mold fault diagnosis and mold maintenance according to the fault alarm type and the mold fault location.

[0013] In a third aspect, the present invention also provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the fault diagnosis and predictive maintenance method of the intelligent injection mold described in any one of the above is implemented.

[0014] In a fourth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned methods for fault diagnosis and predictive maintenance of intelligent injection molds.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention pre-processes the original acoustic vibration signal. All acoustic vibration signals are cleaned, standardized and frequency-band calibrated to obtain filtered acoustic vibration signals, thereby ensuring the integrity and accuracy of the signal. Through the standardized processing of the data, the system can perform subsequent data analysis and operations more efficiently and accurately, and ultimately realize the fault diagnosis and predictive maintenance of the injection mold; (2) The present invention performs feature extraction operations on the processed acoustic vibration high-frequency components and acoustic vibration low-frequency components, accurately classifies the fault features identified by each component, and divides the feature set into a training set, a validation set, and a test set. Sufficient samples ensure the detection accuracy of the detection model; (3) While ensuring the fault identification accuracy of the original detection model, the present invention continuously collects data generated by injection molding machines and molds of different models for generalized learning, updates the fault feature database, and fine-tunes the model parameters to continuously improve the fault diagnosis capability of the model; (4) The present invention deploys the model to the edge computing device, diagnoses the mold status in real time, and dynamically adjusts the injection molding process parameters according to the diagnosis results. Through the generalized detection model, the real-time mold status is inferred to achieve efficient fault diagnosis and mold status predictive maintenance. At the same time, the operator can quickly locate the fault and maintain the mold according to the fault alarm type and mold fault location, thereby improving the efficiency of mold work, promoting the accuracy of fault diagnosis and predictive maintenance of injection molds, realizing intelligent monitoring and fault warning of injection molds, and improving production efficiency and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic flow chart of a fault diagnosis and predictive maintenance method for an intelligent injection mold provided by a first embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of a fault diagnosis and predictive maintenance device for an intelligent injection mold provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0017] 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.

[0018] Reference Figure 1 The first embodiment of the present invention provides a fault diagnosis and predictive maintenance method for an intelligent injection mold, comprising the following steps: S11, obtaining an original acoustic vibration signal and a real-time acoustic vibration signal; S12, performing a signal decomposition operation on the original acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; S13, performing feature extraction operations on the acoustic vibration high-frequency component and the acoustic vibration low-frequency component respectively to obtain a feature set, wherein the feature set includes a high-frequency feature set and a low-frequency feature set; S14, inputting the feature set into a preset detection model for model training to obtain an original detection model, wherein the original detection model includes a high-frequency sub-model and a low-frequency sub-model; S15, inputting the data collected from injection molding machines and molds of different models into the original detection model for generalization learning to obtain a generalized detection model; S16, inputting the real-time acoustic vibration signal into a generalized detection model to perform mold state inference, obtaining a fault alarm type and a mold fault location, and performing injection mold fault diagnosis and mold maintenance according to the fault alarm type and the mold fault location.

[0019] In step S11, an original vibration sound signal and a real-time vibration sound signal are acquired.

[0020] It should be noted that multi-channel sensors are arranged at key positions of the mold to obtain vibration signals and sound signals. For high temperature and high pressure environments, high temperature resistant sensors are selected to ensure the stability of signal acquisition. Through signal isolation technology, the interference of environmental noise on signal transmission is reduced. The collected vibration signals and sound signals are preprocessed to remove high-frequency noise and low-frequency drift. Specifically, multi-channel sensor arrangement is the key to achieving accurate monitoring. Taking the injection mold as an example, pressure sensors, temperature sensors and acceleration sensors can be arranged at key positions such as the mold cavity surface, gate system and cooling channel. This arrangement can fully capture the changes in the internal state of the mold and provide data for subsequent analysis. In high temperature and high pressure environments, sensor selection is crucial. For example, a piezoelectric acceleration sensor with a temperature resistance of up to 300°C can be used to measure vibration, and a thermocouple can be used to measure temperature. These sensors have good high temperature resistance and can maintain stable operation under harsh working conditions. At the same time, high temperature resistant cables and protective sleeves are used to further improve system reliability.

[0021] For example, through multi-dimensional signal collection and analysis methods, real-time monitoring of mold status and fault warning can be achieved. For example, when abnormal high-frequency components are detected in the vibration signal, it indicates that tiny cracks appear on the mold surface; and abnormal popping sounds in the sound signal indicate that the material is overheated or bubbles are generated.

[0022] It should be noted that for vibration signals, wavelet transform can be used to remove high-frequency noise; for sound signals, bandpass filters can be used to extract effective frequency bands. For example, in the vibration analysis of injection molding machines, Butterworth high-pass filters can be used to remove low-frequency interference below 50Hz, highlighting the characteristic frequencies of key actions such as mold opening and closing and injection. The selection of machine learning models and feature extraction directly affect the classification effect. Taking support vector machine (SVM) as an example, it performs well in processing high-dimensional features. Time domain features such as root mean square value and peak factor, frequency domain features such as power spectral density, and time-frequency domain features such as wavelet packet energy can be extracted. These features comprehensively reflect the working status of the mold and lay the foundation for accurate classification. Through this multi-dimensional signal acquisition and analysis method, real-time monitoring of mold status and fault warning can be achieved.

[0023] In step S12, the original acoustic vibration signal is subjected to a signal decomposition operation to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component.

[0024] In one embodiment, the step of performing a signal decomposition operation on the original acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component includes: The original acoustic vibration signal is subjected to signal preprocessing and frequency band calibration to obtain a filtered acoustic vibration signal; Perform signal discrete decomposition according to the filtered acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; Among them, the high-frequency component of the acoustic vibration is used to detect tiny cracks inside the mold, and the low-frequency component of the acoustic vibration is used to analyze hydraulic system abnormalities.

[0025] It should be noted that the original signal is decomposed by wavelet transform method to obtain the acoustic vibration high-frequency component and the acoustic vibration low-frequency component.

[0026] Specifically, for the high-frequency components of the acoustic vibration, a convolutional neural network is used to extract features to determine whether there are micro-crack features. If there are micro-crack features, it is determined that there are micro-cracks inside the mold. For the low-frequency components of the acoustic vibration, a support vector machine is used for feature classification to determine whether there are abnormal features of the hydraulic system. If there are abnormal features, it is determined that there is an abnormality in the hydraulic system.

[0027] Specifically, wavelet transform is a time-frequency analysis method that can decompose signals into high-frequency and low-frequency components. In the detection of internal cracks in molds, the high-frequency component of acoustic vibration contains the tiny vibration information generated by the cracks, while the low-frequency component of acoustic vibration reflects the overall operating status of the hydraulic system.

[0028] For example, after the collected original acoustic vibration signal is transformed by wavelet, a low-frequency acoustic vibration component of 0-500Hz and a high-frequency acoustic vibration component of 500-5000Hz are obtained. In the high-frequency acoustic vibration component, if a sharp peak appears in the range of 2000-3000Hz, it indicates the presence of tiny cracks. The low-frequency acoustic vibration component reflects the abnormality of the hydraulic system. For example, if an abnormal fluctuation occurs in the range of 50-100Hz, it indicates that the hydraulic pump is faulty. Support vector machine, as a classic classification algorithm, can be used to identify such anomalies. By constructing a suitable kernel function, such as radial basis function (RBF), the dimensional feature mapping can be reduced to a high-dimensional space, thereby better distinguishing between normal and abnormal states. By integrating the high-frequency and low-frequency analysis results, a comprehensive detection report can be generated.

[0029] For example, the report points out that: suspected crack features are detected at 2500Hz, and ultrasonic testing is recommended for confirmation; at the same time, abnormal fluctuations are found at 75Hz, suspected hydraulic pump failure, and it is recommended to check the hydraulic system. Decision tree algorithms can be used for health status assessment. Taking the C4.5 algorithm as an example, a decision tree can be constructed based on indicators such as crack feature intensity and hydraulic abnormality. Each node of the tree represents a decision condition, such as "crack feature intensity>0.8" or "hydraulic abnormality>0.6". The leaf node corresponds to the final health status assessment result, such as "immediate maintenance required", "regular inspection required" or "good condition". The early warning mechanism can set multiple thresholds. For example, when the crack feature intensity reaches 0.5-0.8, it is a yellow warning, prompting enhanced monitoring; when it exceeds 0.8, it is a red warning, and it is recommended to shut down immediately for maintenance.

[0030] In step S13, feature extraction operations are performed on the acoustic vibration high-frequency component and the acoustic vibration low-frequency component respectively to obtain a feature set, wherein the feature set includes a high-frequency feature set and a low-frequency feature set.

[0031] In one implementation, the feature extraction operation is performed on the acoustic vibration high frequency component and the acoustic vibration low frequency component respectively to obtain a feature set, wherein the feature set includes a high frequency feature set and a low frequency feature set, including: Extract the time domain features of the high-frequency component and the low-frequency component of the acoustic vibration to obtain the high-frequency time domain features and the low-frequency time domain features; Extract the frequency domain features of the high-frequency component and the low-frequency component of the acoustic vibration to obtain the high-frequency frequency domain features and the low-frequency frequency domain features; The high-frequency time domain features and the high-frequency frequency domain features, as well as the low-frequency time domain features and the low-frequency frequency domain features are standardized and merged to obtain a feature set.

[0032] It should be noted that time domain features can be extracted by analyzing the amplitude and frequency components of the signal. The spectrum of the signal can be obtained through frequency domain analysis, thereby extracting high-frequency frequency domain features (such as peaks) and low-frequency frequency domain features (such as spectrum distribution). For high-frequency and low-frequency components, time domain features are calculated separately. The mean reflects the overall strength level of the signal, the variance describes the degree of fluctuation of the signal, and the peak factor indicates the impact of the signal.

[0033] For example, the high-frequency component of a die-casting mold has a mean of 0.5g, a variance of 0.02g², and a peak factor of 3.5, which indicates the presence of slight cracks.

[0034] Specifically, in frequency domain analysis, we focus on spectrum energy distribution and main frequency components. Spectrum energy distribution reflects the energy distribution of the signal in different frequency bands, while the main frequency component indicates the most significant frequency component in the signal.

[0035] For example, spectrum analysis of the low-frequency component shows a significant energy peak at 75 Hz, which suggests that there is an abnormality in the hydraulic pump.

[0036] Specifically, after merging the time domain and frequency domain features into a feature matrix, the principal component analysis (PCA) method is used for dimensionality reduction. For example, the original feature matrix contains 20 features. After PCA dimensionality reduction, only 5 principal components will be retained, and these principal components can explain 95% of the variance of the original data. After dimensionality reduction, it is necessary to determine whether the feature redundancy is lower than the preset threshold. Excessive feature redundancy will lead to overfitting and affect the generalization ability of the model. Assuming that the redundancy threshold is set to 0.1, if the correlation coefficients between the features after dimensionality reduction are all less than 0.1, it is considered that the feature redundancy is low and can be used as the final feature set. This method of feature extraction and dimensionality reduction is not only applicable to die-casting molds, but can also be extended to the condition monitoring of other industrial equipment. Through similar feature extraction and dimensionality reduction processes, the abnormal state of the equipment can be effectively identified and the accuracy of predictive maintenance can be improved. Through dimensionality reduction processing, the computational complexity can be reduced and the efficiency of real-time monitoring can be improved.

[0037] In step S14, the feature set is input into a preset detection model for model training to obtain an original detection model, wherein the original detection model includes a high-frequency sub-model and a low-frequency sub-model.

[0038] In one embodiment, the feature set is input into a preset detection model for model training to obtain an original detection model, wherein the original detection model includes a high-frequency sub-model and a low-frequency sub-model, including: Pre-build and train high-frequency detection models and low-frequency detection models using historical data; Inputting the high-frequency feature set into the high-frequency detection model to perform feature dimension verification to obtain a high-frequency sub-model; Inputting the low-frequency feature set into the low-frequency detection model to perform feature dimension verification to obtain a low-frequency sub-model; Specifically, historical fault data is obtained, fault features are extracted, and a fault sample library is generated. Data enhancement technology is used to expand the features of samples in the fault sample library to generate a simulated fault sample set. Sample features are extracted from the simulated fault sample set, and a feature matrix is ​​constructed for model training. According to the feature matrix, a support vector machine model is trained to determine whether the sample is an early fault or a potential fault. If the sample features match the preset fault features, it is marked as an early fault sample. The random forest algorithm is used to optimize the features of the marked sample set to improve the model recognition ability. The optimized features are trained twice using the gradient boosting tree algorithm to obtain the final high-frequency detection model and low-frequency detection model.

[0039] Specifically, the high-frequency feature set is input into the high-frequency detection model for feature dimension verification, and the low-frequency feature set is input into the low-frequency detection model for feature dimension verification. By aligning the dimensions of the high-frequency feature set with the features in the high-frequency detection model and fine-tuning the model, a high-frequency sub-model is obtained, and the same is true for the low-frequency sub-model.

[0040] For example, taking an injection molding machine as an example, fault features can be extracted from equipment operation logs, maintenance records, and sensor data. For example, abnormal fluctuations in injection pressure, unstable mold temperature, or a sudden drop in screw speed of an injection molding machine may be signs of potential faults. Data enhancement technology can effectively expand the sample library and improve the generalization ability of the model. For injection molding machine fault samples, simulation samples can be generated by adding noise, rotation, scaling, and other methods. For example, a small perturbation is made to the injection pressure curve to simulate pressure fluctuations of different degrees; or the mold temperature data is stretched on the time axis to simulate temperature changes at different speeds. For injection molding machines, time domain features (such as mean, variance), frequency domain features (such as main frequency components), and statistical features (such as skewness, kurtosis), etc. can be extracted. The matrix composed of these features can fully characterize the operating status of the equipment. The support vector machine model is suitable for early fault identification in binary classification problems. By projecting the extracted features into a high-dimensional space, the support vector machine can effectively distinguish between normal operating conditions and potential fault conditions. For example, the injection pressure, holding time, and cooling time of the injection molding machine can be used to train the model to identify early faults that may cause product quality problems.

[0041] In step S15, the data collected from injection molding machines and molds of different models are input into the original detection model for generalization learning to obtain a generalized detection model.

[0042] In one embodiment, the data collected from injection molding machines and molds of different models are input into the original detection model for generalization learning to obtain a generalized detection model, including: Sensors transmit detection data from different injection molding machines and molds to computers; The computer inputs the detection data into the original detection model for sample detection, and fine-tunes the model parameters to obtain a generalized detection model.

[0043] Specifically, after obtaining the inspection data of the injection molding machine and the mold, data preprocessing is first performed. The original data may have noise or inconsistencies. For example, parameters such as temperature and pressure may have abnormal values ​​or inconsistent units. Then the data is normalized to make the data more concise and accurate. In the sample detection stage, the fault data and good data are divided into training sets, validation sets, and test sets. The data is tested through the model, and the fault is identified through the set fault database. If a new type of fault is detected, the fault feature is extracted and saved in the fault database. The parameters of the model are adjusted to ensure timely identification when similar faults occur, thereby obtaining a generalized detection model.

[0044] It should be noted that a small amount of sample data is collected from the target device in view of the model and brand differences of the target injection molding machine and mold. The model is fine-tuned using the sample data to adjust the model parameters to adapt to the working condition changes and equipment differences of the target device. Specifically, the fine-tuned model is applied to the target injection molding machine and mold through the transfer learning method to determine whether the generalization ability of the model meets the preset threshold. If the generalization ability of the model does not reach the preset threshold, the sample data is re-collected and the model parameters are further fine-tuned until the generalization ability of the model meets the requirements. According to the fine-tuned model, the prediction results that adapt to the working condition changes and equipment differences of the target device are generated, and the final parameter configuration of the model is determined. The model with the final parameter configuration is applied to the target injection molding machine and mold to obtain the prediction results and determine whether the adaptability of the model is stable. If the adaptability of the model is stable, the model is deployed to the target device for real-time monitoring and prediction of the operating status of the injection molding machine and mold.

[0045] For example, for a new model of injection molding machine, you can start with a pre-trained model and use a small amount of sample data from the target device for fine-tuning. Specifically, the vibration, pressure, and temperature data of the injection molding machine under different working conditions can be collected as fine-tuning samples. By adjusting the parameters of the last few layers of the model, it can be adapted to the characteristics of the new device. During the fine-tuning process, the gradient descent method can be used to gradually optimize the model parameters. For example, for high-frequency vibration data, the convolution kernel size and number of layers in the convolutional neural network can be adjusted to better capture the vibration characteristics of the new device. For low-frequency pressure data, the kernel function parameters of the support vector machine can be adjusted to adapt to the pressure change law of the new device.

[0046] It is worth noting that the cross-validation method can be used to judge the generalization ability of the model. Exemplarily, the collected sample data is divided into 5 parts, 4 parts are used as training sets each time, and 1 part is used as a validation set, and the cycle is repeated 5 times to obtain the average accuracy. If the average accuracy reaches the preset threshold (such as 95%), it is considered that the generalization ability of the model meets the requirements. If the threshold is not reached, it is necessary to re-collect more sample data or adjust the model structure, such as increasing the number of hidden layers or changing the activation function. When generating prediction results that are adapted to the target equipment, changes in the operating conditions of the equipment can be considered. For example, for an injection molding machine that produces different plastic products, different prediction models can be set for different product types. By analyzing the changes in equipment parameters during the production of different products, the input features and output results of the model are adjusted to adapt to different production needs.

[0047] In step S16, the real-time acoustic vibration signal is input into a generalized detection model to perform mold state reasoning, obtain the fault alarm type and the mold fault location, and perform injection mold fault diagnosis and mold maintenance based on the fault alarm type and the mold fault location.

[0048] In one embodiment, the real-time acoustic vibration signal is input into a generalized detection model to perform mold state reasoning, obtain a fault alarm type and a mold fault location, and perform injection mold fault diagnosis and mold maintenance according to the fault alarm type and the mold fault location, including: Obtain real-time acoustic and vibration signals of the mold through edge devices; Preprocessing the real-time acoustic vibration signal to obtain a preprocessed acoustic vibration signal; The preprocessed acoustic vibration signal is input into the generalized detection model to perform mold state inference. If the output result of the generalized detection model is a normal state, the detection continues. If the output result is an abnormal state, the time series analysis method is used to process the abnormal signal and output the fault alarm type and mold fault location.

[0049] It should be noted that the mold vibration signal and sound signal are obtained through the edge device, and the signal preprocessing method is used to denoise and extract features of the vibration signal and sound signal to obtain the preprocessed signal data. The preprocessed signal data is input into the pre-trained machine learning model, and the mold state is judged by model reasoning. If the model output is a normal state, the signal continues to be collected; if the model output is an abnormal state, the early warning mechanism is triggered. According to the abnormal state results output by the model, combined with the signal feature analysis, the fault type and location information are determined, and a fault report is generated and stored in the database of the edge device.

[0050] Specifically, the generalized detection model performs mold state inference as follows: the system collects acoustic and vibration signals during normal operation for a long time (such as working condition data of more than 1,000 hours), extracts characteristic parameters including vibration amplitude, frequency distribution, energy trend, etc., forms a mathematical model representing the "normal state", and establishes a fault database by extracting fault features. When a detection signal is input into the generalized detection model, the generalized detection model first compares the characteristic vector of the current signal with the trained "normal state": if more than 90% of the characteristic parameters fall within the normal range (such as vibration amplitude <5mm / s, specific frequency components are missing), it is judged as a normal state. If a signal containing fault features appears, the early warning mechanism is triggered.

[0051] For example, the first level warning (immediate shutdown): a serious defect that causes mold collapse (such as a cavity crack) is detected. The second level warning (planned maintenance): the wear trend of key components is found, indicating the preventive maintenance window. The third level warning (process optimization): the vibration pattern reflects abnormal injection molding parameters (such as insufficient holding time), and the production process needs to be adjusted.

[0052] Specifically, the acquisition of mold vibration signals and sound signals is the basis of fault diagnosis. The edge device collects vibration and sound data of the mold during operation through high-precision sensors. These original signals contain a lot of noise and interference. In order to improve the signal quality, wavelet transform and other methods are used for denoising, and time domain and frequency domain features are extracted at the same time, such as root mean square value, peak factor, etc. The preprocessed signal data is input into the machine learning model for state judgment. The model support vector machine or deep neural network is trained with a large amount of historical data. The model output is "normal" or "abnormal". If it is judged to be abnormal, the early warning mechanism is triggered. For example, when a mold is operating normally, the vibration amplitude is in the range of 0.1 to 0.3 mm / s. If the vibration amplitude is detected to suddenly rise to 0.5 mm / s, the model will judge it as an abnormal state. In the abnormal state, the system will further analyze the signal characteristics to determine the fault type and location. This process involves spectrum analysis, envelope analysis and other technologies.

[0053] The periodicity and trend characteristics of abnormal signals are extracted through time series analysis methods to obtain signal feature descriptions; By describing the signal characteristics and combining with a preset fault type library, a matching fault type is obtained; Fault diagnosis is performed according to the matching fault type to obtain the fault alarm type and the mold fault location.

[0054] It should be noted that the time series analysis method is used to further process the abnormal signal, extract the periodic characteristics and trend characteristics in the signal, and obtain a detailed feature description of the fault signal. According to the detailed feature description of the fault signal, combined with the preset fault type library, the fault type is matched and the fault diagnosis result is generated, and the fault type and location information are output. The fault diagnosis result is sent to the remote monitoring platform through the communication module of the edge device, and the local early warning mechanism is triggered at the same time, and the early warning information is generated and displayed on the device interface. According to the fault diagnosis results and early warning information, the mold status database is updated to record the time of the fault, the fault type and the location information, and provide data support for subsequent maintenance.

[0055] Specifically, time series analysis methods are used to deeply explore the characteristics of abnormal signals. Through autocorrelation analysis, Fourier transform and other means, the periodicity and trend characteristics of the signal can be extracted.

[0056] For example, the vibration signal of a mold presents a periodic peak every 10 minutes, suggesting a periodic failure of a component. These detailed feature descriptions help to diagnose the cause of the fault more accurately. The fault type matching process relies on a preset fault type library. The library contains various possible fault modes and their corresponding feature descriptions. The system compares the extracted signal features with the patterns in the library to find the best matching fault type. For example, if a high-frequency vibration signal and a metal collision sound are detected, it may be diagnosed as loose parts inside the mold. The diagnosis results are sent to the remote monitoring platform through the communication module of the edge device, and an early warning is triggered locally. The early warning information includes the fault type, location, severity, etc., so that maintenance personnel can respond quickly. Finally, the system updates the mold status database and records the detailed information of the fault. These data are not only used for current maintenance work, but also for long-term analysis, such as predictive maintenance and equipment life assessment. By continuously accumulating and analyzing these data, the fault diagnosis model can be continuously optimized to improve the accuracy and reliability of the system.

[0057] Specifically, the high-frequency signal and low-frequency signal data in the real-time diagnosis results reflect the changes in key parameters during the injection molding process. High-frequency signals are related to rapidly changing physical quantities such as injection pressure and mold vibration, while low-frequency signals represent relatively slow processes such as hydraulic systems and temperature changes. During the production process of the injection molding machine, a high-frequency signal exceeding the preset threshold may mean that the injection pressure is too high, which can cause mold damage or product quality problems. For example, when the high-frequency signal is detected to exceed 2000Hz and the injection pressure amplitude exceeds the preset pressure threshold, the system automatically calculates the required pressure adjustment value. Assuming that the current injection pressure is 80MPa, the system calculates the adjustment value that needs to be reduced by 5MPa based on the degree of signal abnormality, and then performs the injection pressure reduction operation to adjust the pressure to 75MPa. A low-frequency signal exceeding the preset threshold indicates that the hydraulic system is abnormal. For example, when the low-frequency signal is in the range of 0.5Hz-5Hz and the hydraulic flow amplitude exceeds the preset flow threshold of 2ml / s, the system analyzes the working status of the hydraulic system. This involves checking parameters such as hydraulic oil temperature, flow and pressure. If the analysis results show that the hydraulic oil temperature is too high, the system will generate an adjustment instruction to reduce the output power of the hydraulic pump or increase the cooling water flow.

[0058] The following describes the working process of the present invention using a common scenario as an example. Figure 2 , which is Figure 1 Schematic diagram of the working scenario of the method.

[0059] Step 1: The injection molding equipment is started and the mold is prepared to start injection molding. The computer retrieves the original sound and vibration signal from the database and transmits it to the signal decomposition device. Through the sensors arranged in the injection molding equipment and mold accessories, the real-time sound and vibration signal is obtained and transmitted to the control system for use; Step 2: Preprocess the original acoustic vibration signal through the signal decomposition device, and perform signal decomposition to obtain the acoustic vibration high-frequency component and the acoustic vibration low-frequency component, and transmit the two components to the feature extraction device for use; Step 3: extracting time domain features and frequency domain features of the acoustic vibration high-frequency component and the acoustic vibration low-frequency component respectively through a feature extraction device, and combining the features to obtain a high-frequency feature set and a low-frequency feature set; Step 4: Input the high-frequency feature set and the low-frequency feature set into the trained detection model for feature dimension benchmarking, fine-tune the detection model to make the model more suitable for the injection molding machine to be put into use, and obtain a high-frequency sub-model and a low-frequency sub-model; Step 5: Apply the fine-tuned model to the target injection molding machine and mold through the transfer learning method to determine whether the generalization ability of the model meets the preset threshold. If the generalization ability of the model does not reach the preset threshold, re-collect sample data and further fine-tune the model parameters until the generalization ability of the model meets the requirements; Step 6: Use time series analysis method to further process abnormal signals, extract periodic features and trend features in the signal, and obtain detailed feature description of the fault signal. According to the detailed feature description of the fault signal, combined with the preset fault type library, match the fault type and generate fault diagnosis results, output the fault alarm type and mold fault location, and then perform fault diagnosis and maintenance of the injection mold.

[0060] In summary, the present invention discloses a fault diagnosis and predictive maintenance method for an intelligent injection mold, comprising obtaining an original acoustic vibration signal and a real-time acoustic vibration signal; performing a signal decomposition operation on the original acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; performing a feature extraction operation on the acoustic vibration high-frequency component and the acoustic vibration low-frequency component respectively to obtain a feature set, wherein the feature set includes a high-frequency feature set and a low-frequency feature set; The feature set is input into a preset detection model for model training to obtain an original detection model, wherein the original detection model includes a high-frequency sub-model and a low-frequency sub-model; data collected from injection molding machines and molds of different models are input into the original detection model for generalized learning to obtain a generalized detection model; the real-time acoustic vibration signal is input into the generalized detection model for mold state reasoning to obtain the fault alarm type and the mold fault location, and fault diagnosis and mold maintenance of the injection mold are performed according to the fault alarm type and the mold fault location.

[0061] By arranging high temperature resistant sensors at key positions of the mold, vibration and sound signals are collected, and the signals are decomposed into high-frequency and low-frequency components. The high-frequency component is used to detect tiny cracks inside the mold, and the low-frequency component is used to analyze the abnormality of the hydraulic system. The time domain and frequency domain features are extracted, and the principal component analysis is used for dimensionality reduction. The high-frequency and low-frequency sub-models are trained separately. Through transfer learning and data enhancement technology, the generalization ability and early fault identification ability of the model are improved. The present invention deploys the model to the edge computing device, diagnoses the mold status in real time, and dynamically adjusts the injection molding process parameters according to the diagnosis results. The present invention realizes intelligent monitoring and fault warning of injection molds, improves production efficiency and product quality.

[0062] Reference Figure 2 The second embodiment of the present invention provides a fault diagnosis and predictive maintenance device for an intelligent injection mold, comprising: A data acquisition module, used to acquire original acoustic vibration signals and real-time acoustic vibration signals; A signal decomposition module, used for performing a signal decomposition operation on the original acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; A feature extraction module performs feature extraction operations on the acoustic vibration high-frequency component and the acoustic vibration low-frequency component respectively to obtain a feature set, wherein the feature set includes a high-frequency feature set and a low-frequency feature set; A feature conversion module, which inputs the feature set into a preset detection model for model training to obtain an original detection model, wherein the original detection model includes a high-frequency sub-model and a low-frequency sub-model; A generalized learning module, which inputs the data collected from different types of injection molding machines and molds into the original detection model for generalized learning to obtain a generalized detection model; The mold detection module is used to input the real-time acoustic vibration signal into the generalized detection model to perform mold state reasoning, obtain the fault alarm type and the mold fault location, and perform injection mold fault diagnosis and mold maintenance according to the fault alarm type and the mold fault location.

[0063] It should be noted that the fault diagnosis and predictive maintenance device for an intelligent injection mold provided in an embodiment of the present invention is used to execute all the process steps of the fault diagnosis and predictive maintenance method for an intelligent injection mold in the above-mentioned embodiment. The working principles and beneficial effects of the two correspond one to one, and therefore will not be repeated here.

[0064] The embodiment of the present invention also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a fault diagnosis and predictive maintenance method for an intelligent injection mold. When the processor executes the computer program, the steps in the above-mentioned embodiments of the fault diagnosis and predictive maintenance method for an intelligent injection mold are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of each module / unit in the above-mentioned device embodiments are implemented, such as the data acquisition module.

[0065] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0066] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0067] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0068] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0069] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0070] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0071] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A fault diagnosis and predictive maintenance method for an intelligent injection mold, characterized in that: include: Obtaining original acoustic vibration signals and real-time acoustic vibration signals; Decomposing the original acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; Perform feature extraction operations on the acoustic vibration high-frequency component and the acoustic vibration low-frequency component respectively to obtain a feature set, wherein the feature set includes a high-frequency feature set and a low-frequency feature set; Inputting the feature set into a preset detection model for model training to obtain an original detection model, wherein the original detection model includes a high-frequency sub-model and a low-frequency sub-model; Inputting the data collected from injection molding machines and molds of different models into the original detection model for generalization learning to obtain a generalized detection model; The real-time acoustic vibration signal is input into a generalized detection model to perform mold state reasoning, obtain the fault alarm type and the mold fault location, and perform injection mold fault diagnosis and mold maintenance according to the fault alarm type and the mold fault location.

2. The fault diagnosis and predictive maintenance method of the intelligent injection mold according to claim 1 is characterized in that: The original acoustic vibration signal is subjected to a signal decomposition operation to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component, including: The original acoustic vibration signal is subjected to signal preprocessing and frequency band calibration to obtain a filtered acoustic vibration signal; Perform signal discrete decomposition according to the filtered acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; Among them, the high-frequency component of the acoustic vibration is used to detect tiny cracks inside the mold, and the low-frequency component of the acoustic vibration is used to analyze hydraulic system abnormalities.

3. The fault diagnosis and predictive maintenance method of the intelligent injection mold according to claim 1 is characterized in that: The high-frequency component of the acoustic vibration and the low-frequency component of the acoustic vibration are subjected to feature extraction operations respectively to obtain a feature set, including: Extract the time domain features of the high-frequency component and the low-frequency component of the acoustic vibration to obtain the high-frequency time domain features and the low-frequency time domain features; Extract the frequency domain features of the high-frequency component and the low-frequency component of the acoustic vibration to obtain the high-frequency frequency domain features and the low-frequency frequency domain features; The high-frequency time domain features and the high-frequency frequency domain features, as well as the low-frequency time domain features and the low-frequency frequency domain features are standardized and merged to obtain a feature set.

4. The fault diagnosis and predictive maintenance method of the intelligent injection mold according to claim 1 is characterized in that: The step of inputting the feature set into a preset detection model for model training to obtain an original detection model includes: Pre-build and train high-frequency detection models and low-frequency detection models using historical data; Inputting the high-frequency feature set into the high-frequency detection model to perform feature dimension verification to obtain a high-frequency sub-model; Inputting the low-frequency feature set into the low-frequency detection model to perform feature dimension verification to obtain a low-frequency sub-model; The high-frequency sub-model and the low-frequency sub-model are loaded into the memory to obtain the original detection model.

5. The fault diagnosis and predictive maintenance method of the intelligent injection mold according to claim 1 is characterized in that: The data collected from different types of injection molding machines and molds are input into the original detection model for generalization learning to obtain a generalized detection model, including: Sensors transmit detection data from different injection molding machines and molds to computers; The computer inputs the detection data into the original detection model for sample detection, and fine-tunes the model parameters to obtain a generalized detection model.

6. The fault diagnosis and predictive maintenance method of the intelligent injection mold according to claim 1 is characterized in that: The step of inputting the real-time acoustic vibration signal into a generalized detection model to perform mold state inference and obtain the fault alarm type and mold fault location includes: Obtain real-time acoustic and vibration signals of the mold through edge devices; Preprocessing the real-time acoustic vibration signal to obtain a preprocessed acoustic vibration signal; The preprocessed acoustic vibration signal is input into the generalized detection model to perform mold state inference. If the output result of the generalized detection model is a normal state, the detection continues. If the output result is an abnormal state, the time series analysis method is used to process the abnormal signal and output the fault alarm type and mold fault location.

7. The fault diagnosis and predictive maintenance method of the intelligent injection mold according to claim 6 is characterized in that: If the output result is an abnormal state, the time series analysis method is used to process the abnormal signal and output the fault alarm type and mold fault location, including: The periodicity and trend characteristics of abnormal signals are extracted through time series analysis methods to obtain signal feature descriptions; By describing the signal characteristics and combining with a preset fault type library, a matching fault type is obtained; Fault diagnosis is performed according to the matching fault type to obtain the fault alarm type and the mold fault location.

8. A fault diagnosis and predictive maintenance device for an intelligent injection mold, characterized in that: include: A data acquisition module, used to acquire original acoustic vibration signals and real-time acoustic vibration signals; A signal decomposition module, used for performing a signal decomposition operation on the original acoustic vibration signal to obtain an acoustic vibration high-frequency component and an acoustic vibration low-frequency component; A feature extraction module, used to perform feature extraction operations on the acoustic vibration high-frequency component and the acoustic vibration low-frequency component respectively to obtain a feature set, wherein the feature set includes a high-frequency feature set and a low-frequency feature set; A feature conversion module, used for inputting the feature set into a preset detection model for model training to obtain an original detection model, wherein the original detection model includes a high-frequency sub-model and a low-frequency sub-model; A generalized learning module, used for inputting data collected from injection molding machines and molds of different models into the original detection model for generalized learning to obtain a generalized detection model; The mold detection module is used to input the real-time acoustic vibration signal into the generalized detection model to perform mold state reasoning, obtain the fault alarm type and the mold fault location, and perform injection mold fault diagnosis and mold maintenance according to the fault alarm type and the mold fault location.

9. An electronic device, characterized in that: It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the fault diagnosis and predictive maintenance method of the intelligent injection mold as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the fault diagnosis and predictive maintenance method for the intelligent injection mold according to any one of claims 1 to 7.

Images