Mechanical action fault monitoring method based on microcontroller and convolutional neural network

Abstract

The invention discloses a mechanical action fault monitoring method based on a microcontroller and a convolutional neural network, and the method specifically comprises the steps: 1) obtaining mechanical fault action data from a sensor, and sorting the data into a one-dimensional time sequence with three features; 2) preprocessing the data set, including filling a sequence to unify length, and converting a label into one-hot code; (3) a mechanical action fault classification model is constructed, the model is composed of a plurality of convolution layers, pooling layers, flattening layers and full connection layers, and finally a classification result is output through a Softmax layer; 4) training the model by using a TensorF low framework, and adopting an Adam optimizer and a classification cross entropy loss function; and 5) deploying a trained model on the microcontroller to realize real-time mechanical fault monitoring. According to the invention, the deep learning model is combined with the microcontroller, the mechanical fault action can be efficiently identified, the method is suitable for various industrial application scenes, and the fault detection capability and maintenance efficiency of mechanical equipment are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of identifying mechanical motion faults through deep learning methods based on microcontrollers, and in particular to a mechanical motion fault monitoring method based on a microcontroller and a convolutional neural network. Background Art

[0002] Mechanical fault monitoring is of great significance in modern industry. Rotating machinery is the core part of various mechanical equipment and is widely used in manufacturing, transportation, energy and other fields. As the equipment continues to operate, the occurrence of mechanical failures may lead to reduced production efficiency, equipment damage and even safety accidents. Therefore, it is particularly important to monitor and diagnose mechanical failures in a timely and accurate manner.

[0003] At present, mechanical fault monitoring technologies mainly include methods based on vibration analysis, temperature monitoring and sound recognition. However, traditional monitoring methods often rely on manual experience or simple signal processing, and cannot fully and real-timely identify the type and extent of faults. With the development of deep learning technology, fault monitoring methods based on convolutional neural networks (CNN) have gradually emerged. CNN can effectively extract features from time series data and improve the accuracy of fault identification.

[0004] However, existing deep learning-based fault monitoring systems usually require powerful computing resources and are difficult to deploy on microcontrollers with limited resources. This limits their application in actual industrial environments. Therefore, developing a mechanical fault monitoring method suitable for real-time operation on microcontrollers can effectively improve the maintenance efficiency and safety of mechanical equipment, which is an important direction of current research.

[0005] The present invention is proposed based on the above background, combining convolutional neural network and microcontroller technology, aiming to achieve efficient mechanical action fault monitoring, improve the real-time and accuracy of fault detection, and adapt to the needs of modern industry for equipment monitoring. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of existing mechanical fault monitoring methods and propose a mechanical action fault monitoring method based on convolutional neural network (CNN) and microcontroller to improve the fault detection capability and maintenance efficiency of mechanical equipment, especially to achieve real-time monitoring in complex industrial environments.

[0007] To achieve the above object, the technical solution provided by the present invention is: a mechanical action fault monitoring method based on a convolutional neural network and a microcontroller, comprising the following steps:

[0008] 1) Obtain mechanical fault action data from sensors and organize the data into a one-dimensional time series;

[0009] 2) Preprocess the dataset, including padding the sequences to uniform length and converting the labels to one-hot encoding;

[0010] 3) Constructing a mechanical action fault classification model, the specific structure of the model is as follows: Further, in step 3), the convolutional neural network includes the following contents:

[0011] Input layer module: This module receives one-dimensional time series data with a shape of (150,3) to ensure that the format of the input data meets the model requirements;

[0012] The first convolutional layer: This layer contains 30 convolution kernels of size 3, uses the ReLU activation function, extracts features from the input data, and the output shape is (None, 50, 30);

[0013] The second convolutional layer: This layer contains 15 convolution kernels of size 3. The ReLU activation function is used to further extract features. The output shape is (None, 16, 15).

[0014] Max pooling layer: A pooling layer with a pool size of 3 is connected after the second convolutional layer to reduce the dimension of the data and the output shape is (None, 5, 15);

[0015] Flattening layer: This layer flattens the pooled multidimensional data into a one-dimensional vector, and the output shape is (None, 75);

[0016] Fully connected layer: The number of nodes in this layer is the number of classifications, which is used to further process the extracted features, and the output shape is (None, num_classes);

[0017] Dropout layer: Add a Dropout layer after the fully connected layer to prevent overfitting;

[0018] Softmax layer: This layer is responsible for converting the final output into a probability distribution to facilitate the classification of various types of mechanical action failures.

[0019] 4) Use the TensorFlow framework to train the model, using the Adam optimizer and categorical cross entropy loss function;

[0020] 5) Deploy the trained model on a microcontroller to achieve real-time mechanical fault monitoring using the NNoM framework. This step includes:

[0021] 5.1) Convert the trained model into a binary format suitable for microcontrollers so that it can run in resource-limited environments;

[0022] 5.2) Run the corresponding program on the microcontroller to receive data from the vibration sensor in real time and input it into the deployed fault monitoring model for reasoning.

[0023] Further, in step 1), the obtained mechanical fault action data set comes from sensors of multiple industrial equipment;

[0024] Further, in step 2), the preprocessing operations include denoising, standardizing the data, and performing data enhancement to improve the robustness of the model;

[0025] Further, in step 5), the microcontroller uses TensorFlow to perform model inference to ensure efficient operation in a resource-constrained environment.

[0026] By combining the deep learning model with the microcontroller, the present invention can efficiently identify mechanical fault actions, is suitable for various industrial application scenarios, and significantly improves the fault detection capability and maintenance efficiency of mechanical equipment.

[0027] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0028] 1. The present invention can monitor mechanical movements in real time, identify and warn of mechanical failures in a timely manner without replacing traditional mechanical equipment as a whole, thereby significantly reducing equipment downtime and improving production efficiency.

[0029] 2. The present invention can process mechanical fault action data directly on a microcontroller without uploading the data to the cloud for analysis, thus saving bandwidth resources and ensuring real-time performance. It is particularly suitable for industrial application scenarios with high requirements on response time.

[0030] 3. The present invention can accurately identify mechanical failures under different working environments, including high temperature, strong vibration and noise, overcomes the limitation of traditional methods' reliance on clear data, improves the reliability of the equipment, and reduces the risk of false alarms and missed alarms.

[0031] 4. The present invention can identify complex mechanical action failure modes, especially those in transitional states, thereby ensuring the safe operation of the mechanical system, improving the fault tolerance of the overall system, and enhancing the stability of the production line.

[0032] 5. The present invention uses a deep learning model to run on a microcontroller with limited resources, which reduces the company's cost of hardware and technology investment, while improving the accuracy and response speed of fault detection, making it more suitable for the needs of small and medium-sized enterprises.

[0033] 6. Through the study and analysis of historical data, the model of the present invention can be continuously optimized to improve the effect of subsequent monitoring, help realize intelligent maintenance, and reduce manual intervention.

[0034] 7. The implementation scheme of the present invention provides effective support for intelligent manufacturing, promotes the development of Industry 4.0, and helps enterprises achieve automation and intelligent transformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings are used to provide further understanding of the technical solution of the present application and constitute a part of the specification. Together with the examples of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0036] Figure 1 The present invention is a flowchart of a specific method of a mechanical motion fault monitoring method based on a microcontroller and a convolutional neural network.

[0037] Figure 2 This is the structure diagram of the convolutional neural network.

[0038] Figure 3 This is a structural diagram of the real-time mechanical fault monitoring method based on the NNoM framework. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application 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 application and are not used to limit the present application.

[0040] The following is a further description and explanation of the specific implementation details of the present invention in conjunction with the accompanying drawings of the specification.

[0041] This embodiment discloses a mechanical action fault monitoring method based on a microcontroller and a convolutional neural network, and the specific steps are as follows:

[0042] 1) Use vibration sensors to collect mechanical motion data in a specified area and organize the data into a one-dimensional time series. This data can reflect the dynamic state of the mechanical equipment and help with subsequent fault analysis;

[0043] 2) Preprocessing the mechanical motion data set obtained in step 1), specifically including the following steps:

[0044] 2.1) Pad the sequence to a uniform length to ensure consistent data format for input to the model, and convert the label to one-hot encoding for subsequent classification processing;

[0045] 3) Build a mechanical action fault classification model, which consists of multiple convolutional layers, pooling layers, flattening layers and fully connected layers, and finally outputs the classification results through the Softmax layer. The specific network structure is as follows:

[0046] 3.1) The first convolutional layer: 30 filters, kernel size 3, stride 3 are set to extract low-level features in the data;

[0047] 3.2) Second convolutional layer: 15 filters are set, and the convolution kernel size is 3 and the stride is 3 to further extract deeper feature information;

[0048] 3.3) Maximum pooling layer: The pool size is set to 3 and the stride is 3, which is used to reduce the dimension of the feature map and enhance the model's anti-noise ability;

[0049] 4) Use the TensorFlow framework to train the model, using the Adam optimizer and categorical cross entropy loss function to improve training efficiency and model accuracy;

[0050] 5) Deploy the trained model on a microcontroller to achieve real-time mechanical fault monitoring using the NNoM framework. This step includes:

[0051] 5.1) Convert the trained model into a binary format suitable for microcontrollers so that it can run in resource-limited environments;

[0052] 5.2) Run the corresponding program on the microcontroller to receive data from the vibration sensor in real time and input it into the deployed fault monitoring model for reasoning;

[0053] 5.3) After the model outputs the results, the identified mechanical action status and possible fault types are displayed through indicator lights or display screens.

[0054] In summary, the present invention realizes real-time mechanical action fault monitoring based on a microcontroller, which can effectively improve the fault detection capability and maintenance efficiency of the equipment and has a wide range of practical application value. The method can be widely used in the fields of industrial automation, equipment maintenance, etc., providing strong support for intelligent manufacturing.

[0055] The above-described embodiments are only preferred implementations of the present invention and do not limit the scope of implementation of the present invention. Any changes made according to the shape and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A mechanical motion fault monitoring method based on a microcontroller and a convolutional neural network, characterized in that: The following steps are involved: 1) Use vibration sensors to collect mechanical motion data in a specified area and organize the data into a one-dimensional time series; 2) Preprocessing the mechanical motion dataset obtained in step 1), including sequence filling and label conversion, as follows: 2.1) Pad the sequence to a uniform length and convert the label to one-hot encoding; 3) Construct a mechanical action fault classification model, which consists of multiple convolutional layers, pooling layers, flattening layers and fully connected layers, and finally outputs the classification results through the Softmax layer; the parameters of the convolutional layer are as follows: 3.1) First convolutional layer: 30 filters, kernel size 3, stride 3; 3.2) Second convolutional layer: 15 filters, kernel size 3, stride 3; 3.3) Maximum pooling layer with pool size of 3 and stride of 3; 4) Use the TensorFlow framework to train the model, using the Adam optimizer and categorical cross entropy loss function; 5) Deploy the trained model on a microcontroller to achieve real-time mechanical fault monitoring using the NNoM framework.

2. The mechanical motion fault monitoring method based on microcontroller and convolutional neural network according to claim 1 is characterized in that: In step 1), the collected mechanical motion data is time series data detected by the sensor in real time.

3. The mechanical motion fault monitoring method based on microcontroller and convolutional neural network according to claim 1 is characterized in that: In step 4), early stopping and model checkpointing are used as callback functions during the training process to optimize the training effect of the model.

4. The mechanical motion fault monitoring method based on microcontroller and convolutional neural network according to claim 1 is characterized in that: In step 5), the trained model is converted into a binary format suitable for microcontrollers using the NNoM framework.

5. The mechanical motion fault monitoring method according to claim 4, characterized in that: The model deployed on the microcontroller is able to perform real-time inference and identify different types of mechanical motion faults.

Images