Method and system for monitoring surface quality of part machined by numerical control machine tool

Abstract

The invention discloses a numerical control machine tool machining part surface quality monitoring method and system. The method comprises the steps that S1, real-time machining data are collected; s2, according to the real-time processing data, constructing a neural network prediction model based on LSTM; and S3, predicting the surface quality of the part according to the neural network prediction model based on the LSTM, and by adopting the technical scheme of the invention, the problems of low accuracy and poor real-time performance of surface quality monitoring of the existing traditional machined part are solved.

Description

Technical Field

[0001] The invention belongs to the technical field of mechanical processing, and in particular relates to a method and system for monitoring the surface quality of parts processed by a numerically controlled machine tool. Background Art

[0002] CNC machine tools are automated machine tools equipped with program control systems, and are widely used in aviation, automobile, mold manufacturing and other fields. With the rapid development of the manufacturing industry, the quality monitoring of its processed parts has become a focus of increasing attention. The main method for monitoring the surface quality of workpieces in traditional industrial processing is offline monitoring, which has problems such as high subjectivity, low efficiency and low precision. In addition, with the rapid development of information technologies such as mobile communications, cloud computing and the Internet of Things, the generation of a large amount of data has led to a continuous decrease in computing speed. The use of cloud computing to analyze the surface quality of workpieces also has shortcomings such as high latency and low security. Domestic and foreign scholars have proved that the combination of deep learning and edge computing can more accurately and real-time monitor the surface quality of processed parts.

[0003] At present, the surface quality monitoring of parts processed by CNC machine tools is mainly divided into image recognition and signal evaluation. The image recognition method takes the surface image of the workpiece after processing and uses deep learning algorithms such as YOLO, SSD, and R-CNN to detect scratches, pits, rust, etc. on the image. The signal evaluation method analyzes the impact on the surface quality of the workpiece through the changes in signals such as vibration, acoustic emission, and power generated during the processing. However, the above monitoring methods are all offline monitoring, and this type of monitoring method will generate a large amount of data, resulting in a decrease in the calculation rate, making it impossible to perform real-time monitoring in the actual processing environment.

[0004] The surface quality monitoring method of machined parts based on deep learning and edge computing has fast calculation speed and high accuracy, and can monitor the surface quality of parts in real time close to the machine tool. However, the existing methods combining deep learning with edge computing are almost all used in image detection. More importantly, in the actual machine tool processing process, it is difficult to collect real-time processing images, and this type of image recognition method can only identify obvious defects on the surface of the workpiece, and cannot more accurately identify subtle changes on the surface of the part. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for monitoring the surface quality of parts processed by CNC machine tools, which solves the problems of low accuracy and poor real-time performance of the existing traditional surface quality monitoring of processed parts.

[0006] To achieve the above object, the present invention adopts the following technical solution:

[0007] A method for monitoring the surface quality of parts processed by a numerical control machine tool, comprising:

[0008] Step S1, collecting real-time processing data;

[0009] Step S2: construct a LSTM-based neural network prediction model based on real-time processing data;

[0010] Step S3: predict the surface quality of the part according to the LSTM-based neural network prediction model.

[0011] Preferably, in step S1, a vibration sensor, a vibration acquisition card and a roughness meter are used to collect real-time processing data.

[0012] Preferably, the real-time processing data includes: a vibration signal of a CNC machine tool and the roughness of a part surface.

[0013] Preferably, step S2 comprises:

[0014] Step S21, the vibration signal of each processing cycle will be converted into a set of characteristic values ​​and correspond to a set of roughness parameters;

[0015] Step S22: Using 20 eigenvalues ​​of the vibration signal of each processing cycle as input data and the roughness parameter of the corresponding processing cycle as output data, the LSTM network model is trained to obtain a neural network prediction model based on LSTM.

[0016] The present invention also provides a surface quality monitoring system for parts processed by a numerically controlled machine tool, comprising:

[0017] Acquisition equipment, used to collect real-time processing data;

[0018] Edge computing devices are used to build LSTM-based neural network prediction models based on real-time processing data;

[0019] Monitoring equipment used to predict part surface quality based on a LSTM-based neural network prediction model.

[0020] Preferably, the acquisition equipment includes: a vibration sensor, a vibration acquisition card and a roughness meter.

[0021] Preferably, the real-time processing data includes: a vibration signal of a CNC machine tool and the roughness of a part surface.

[0022] Preferably, the edge computing device includes:

[0023] A first processing unit, in which the vibration signal for each machining cycle is converted into a set of characteristic values ​​and corresponds to a set of roughness parameters;

[0024] The second processing unit is used to take the 20 eigenvalues ​​of the vibration signal of each processing cycle as input data, and the roughness parameter corresponding to the processing cycle as output data, train the LSTM network model, and obtain the LSTM-based neural network prediction model.

[0025] The present invention collects real-time processing data, uses a deep learning model to predict the surface quality of parts, and displays it in real time in the monitoring system. Compared with the prior art, the present invention has the following beneficial effects:

[0026] 1. The present invention can accurately and real-time predict and display the surface quality parameters of the currently processed parts close to the machine tool end.

[0027] 2. The present invention adopts deep learning method to construct a prediction model, which can avoid the complicated steps of subsequent analysis and processing and improve the accuracy of part surface quality monitoring.

[0028] 3. By combining the deep learning model with edge computing, the present invention can quickly calculate the deep learning model at the edge, thereby improving the real-time performance of part surface quality monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0030] Figure 1 This is a flow chart of the surface quality monitoring method of CNC machine tool workpiece based on edge computing;

[0031] Figure 2 It is a schematic diagram of processing data collection;

[0032] Figure 3 It is a flowchart of format conversion and edge deployment of neural network prediction model;

[0033] Figure 4 It is the working flow chart of the monitoring equipment. DETAILED DESCRIPTION

[0034] 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.

[0035] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] Embodiment 1:

[0037] like Figure 1 As shown, an embodiment of the present invention provides a method for monitoring the surface quality of parts processed by a CNC machine tool, comprising:

[0038] Step S1, collecting real-time processing data;

[0039] Step S2: construct a LSTM-based neural network prediction model based on real-time processing data;

[0040] Step S3: predict the surface quality of the part according to the LSTM-based neural network prediction model.

[0041] As an implementation of the present invention, in step S1, a vibration sensor, a vibration acquisition card and a roughness meter are used to collect real-time processing data. The real-time processing data includes: a vibration signal of a CNC machine tool and the roughness of a part surface.

[0042] As an implementation of the embodiment of the present invention, step S2 includes:

[0043] Step S21, the vibration signal of each processing cycle will be converted into a set of characteristic values ​​and correspond to a set of roughness parameters;

[0044] Step S22: Take the 20 eigenvalues ​​of the vibration signal of each processing cycle as input data, take the roughness parameter of the corresponding processing cycle as output data, train the LSTM network model, and obtain the LSTM-based neural network prediction model. Set the model training rounds to 50 rounds, and set the weight update to be performed once every 32 samples. During the training process, 20% of the training data will be used to verify the model performance. By adjusting the model parameters, the model prediction accuracy is optimized.

[0045] As an implementation method of the present invention, in step S3, the obtained part surface quality parameter results are displayed in real time in the monitoring interface and the chart is updated in the real-time curve interface, so as to facilitate the observation of the fluctuation of the part surface quality during the continuous processing. An alarm threshold is set, and an alarm is output when the part surface quality parameter obtained by model inference exceeds the threshold, and the alarm information is saved to the historical data interface for subsequent viewing.

[0046] Embodiment 2:

[0047] The present invention also provides a surface quality monitoring system for parts processed by a numerically controlled machine tool, comprising:

[0048] Acquisition equipment, used to collect real-time processing data;

[0049] Edge computing devices are used to build LSTM-based neural network prediction models based on real-time processing data;

[0050] Monitoring equipment used to predict part surface quality based on a LSTM-based neural network prediction model.

[0051] As an implementation of the embodiment of the present invention, the acquisition equipment includes: a vibration sensor, a vibration acquisition card and a roughness meter. The real-time processing data includes: a vibration signal of a CNC machine tool and the roughness of a part surface.

[0052] Further, if Figure 2 As shown. Install a vibration sensor on the machine tool spindle, connect the vibration sensor to the acquisition card, and connect the acquisition card output port to the industrial computer, and use the edge computing device to communicate with the industrial computer. After the CNC machine tool is processed, the vibration acquisition card transmits the data to the industrial computer. Secondly, build an MQTT server and client in the edge computing device. The edge computing device and the industrial computer use TCP / IP network communication, and use the MQTT protocol to translate the data and save it to the edge computing device.

[0053] As an implementation of an embodiment of the present invention, an edge computing device includes:

[0054] A first processing unit, in which the vibration signal for each machining cycle is converted into a set of characteristic values ​​and corresponds to a set of roughness parameters;

[0055] The second processing unit is used to take the 20 eigenvalues ​​of the vibration signal of each processing cycle as input data and the roughness parameters of the corresponding processing cycle as output data, train the LSTM network model, and obtain the LSTM-based neural network prediction model; set the model training rounds to 50 rounds, and set the weight update to be performed once every 32 samples. During the training process, 20% of the training data will be used to verify the model performance. By adjusting the model parameters, the model prediction accuracy is optimized.

[0056] Furthermore, the high-precision neural network prediction model is saved and converted into the format model required by the edge development board of the edge computing device, compiled according to the model structure, and after the compilation is completed, the neural network prediction model is deployed to the development board to realize edge prediction. Figure 3As shown. After saving the high-precision neural network prediction model in a specific format. Use the model conversion tool RKNN-Toolkit2 specified by the edge development board to convert the model. First, create an RKNN object and initialize the RKNN SDK environment, and then configure the preprocessing parameters of the model. Import the trained TensorFlow LSTM model and use the build interface to build the RKNN model. After initializing the runtime environment, infer the input data through the inference interface to obtain the inference result. Finally, export the RKNN model and release the RKNN object. After obtaining the RKNN model, compile the model in the edge development board according to the input and output types of the model and the neural network structure of the model, compile the model into a structure type that can be recognized by the edge development board, and finally infer the model in the edge development board to realize the operation of the LSTM neural network model on the edge development board.

[0057] As an implementation method of the present invention, the monitoring device is used to display the surface roughness parameters of the parts in real time, set the upper limit to realize the over-threshold alarm, and realize the storage and query of the alarm information. Figure 4 As shown in the figure, after the long short-term memory network model can be successfully inferred in the edge development board, a surface quality monitoring interface based on the embedded software development tool QtCreator is designed in the edge development board, which mainly includes a monitoring interface, an alarm interface, a real-time curve interface, and a historical data interface. The surface quality parameter results of the part obtained by the edge model through real-time vibration signal inference are displayed in real time in the monitoring interface and the chart is updated in the real-time curve interface, which is convenient for observing the fluctuation of the surface quality of the part during continuous processing. Set the alarm threshold, and output an alarm when the surface quality parameter of the part obtained by model inference exceeds the threshold, and save the alarm information to the historical data interface for subsequent viewing.

[0058] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for monitoring the surface quality of parts processed by a CNC machine tool, characterized in that: include: Step S1, collecting real-time processing data; Step S2: construct a LSTM-based neural network prediction model based on real-time processing data; Step S3: predict the surface quality of the part according to the LSTM-based neural network prediction model.

2. The method for monitoring the surface quality of parts processed by a CNC machine tool according to claim 1, characterized in that: In step S1, real-time processing data is collected using a vibration sensor, a vibration acquisition card and a roughness meter.

3. The surface quality monitoring method of parts processed by CNC machine tools according to claim 2, characterized in that: The real-time processing data includes: a vibration signal of a numerically controlled machine tool and the roughness of a part surface.

4. The method for monitoring the surface quality of parts processed by a CNC machine tool as claimed in claim 3, characterized in that: Step S2 includes: Step S21, the vibration signal of each processing cycle will be converted into a set of characteristic values ​​and correspond to a set of roughness parameters; Step S22: Using 20 eigenvalues ​​of the vibration signal of each processing cycle as input data and the roughness parameter of the corresponding processing cycle as output data, the LSTM network model is trained to obtain a neural network prediction model based on LSTM.

5. A surface quality monitoring system for parts processed by CNC machine tools, characterized in that: include: Acquisition equipment, used to collect real-time processing data; Edge computing devices are used to build LSTM-based neural network prediction models based on real-time processing data; Monitoring equipment used to predict part surface quality based on a LSTM-based neural network prediction model.

6. The surface quality monitoring system for parts processed by a CNC machine tool as claimed in claim 5, characterized in that The acquisition equipment includes: vibration sensor, vibration acquisition card and roughness meter.

7. The surface quality monitoring system for parts processed by a CNC machine tool as claimed in claim 6, characterized in that The real-time processing data includes: a vibration signal of a numerically controlled machine tool and the roughness of a part surface.

8. The surface quality monitoring system for parts processed by a CNC machine tool as claimed in claim 7, characterized in that Edge computing devices include: A first processing unit, in which the vibration signal for each machining cycle is converted into a set of characteristic values ​​and corresponds to a set of roughness parameters; The second processing unit is used to take the 20 eigenvalues ​​of the vibration signal of each processing cycle as input data, and the roughness parameter corresponding to the processing cycle as output data, train the LSTM network model, and obtain the LSTM-based neural network prediction model.

Images