CN114577334B - Real-time online monitoring method and system for laser shot blasting processing state based on machine learning - Google Patents
Real-time online monitoring method and system for laser shot blasting processing state based on machine learning Download PDFInfo
- Publication number
- CN114577334B CN114577334B CN202210210094.XA CN202210210094A CN114577334B CN 114577334 B CN114577334 B CN 114577334B CN 202210210094 A CN202210210094 A CN 202210210094A CN 114577334 B CN114577334 B CN 114577334B
- Authority
- CN
- China
- Prior art keywords
- laser
- acoustic
- shot blasting
- machine learning
- processing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2132—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on discrimination criteria, e.g. discriminant analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Abstract
The invention provides a real-time online monitoring method and a real-time online monitoring system for a laser shot blasting state based on machine learning, wherein the method comprises the following steps: establishing a machine learning model; performing laser shot blasting, reading a light emitting signal fed back by a laser by a control module, triggering an acoustic sensing module to record a laser shot blasting acoustic signal, sending laser shot blasting acoustic signal data to a data processing module, and extracting multi-dimensional acoustic characteristic parameters of the acoustic signal; carrying out dimension reduction processing on the multi-dimensional acoustic characteristic parameters to obtain characteristic discrimination parameters of the multi-dimensional acoustic characteristic parameters; sending the acoustic characteristic discrimination parameters to a machine learning classification model, identifying the current processing condition, and feeding back the result to an upper computer; if the current machining state result is that machining is not finished, the steps are repeatedly executed; and if the current machining state result is that the machining is finished, finishing the operation. The sensor is flexible in installation and arrangement and low in cost; the data volume of the acoustic signals needing to be transmitted and processed is small; the machine learning method is adopted, so that the recognition speed is high, and the real-time performance is strong.
Description
Technical Field
The invention relates to the field of intelligent monitoring of machining states, in particular to a real-time online monitoring method and system for a laser shot blasting machining state based on machine learning.
Background
Laser peening is a novel process for shaping or strengthening a material by inducing transient high-pressure impact force on the surface of the material using a high-energy short pulse laser beam. At present, the modern industry has a trend towards automation and intellectualization. In order to ensure the reliability of the process and the traceability of the processing conditions in the processing process, the on-line monitoring system for the processing state of the workpiece is gradually popularized in the specific production process. However, for laser peening, it is difficult for general industrial detection methods to reflect the processing state of a workpiece in real time. Therefore, an effective shot peening state identification method needs to be established to fully utilize signal data which can be collected and measured in real time to realize monitoring and fault diagnosis of the shot peening state under a non-stop running state, and to help ensure that the quality of a processed product is more stable and reliable. It has been found experimentally that the acoustic sound produced by the interaction of the laser pulses with the material surface varies during processing in the different states of the confinement and absorption layers. Compared with the real-time measurement of the forming state of the workpiece, the acoustic signal detection of the laser shot blasting can reflect the change of the processing condition in real time through the change condition of the acoustic signal in the processing process, and has stronger feasibility.
Research into monitoring of the laser peening process has been currently conducted. Qin Lin et al (Qin Lin, cheng Li, he Weifeng. A laser shot peening acoustic monitoring technology based on data correlation, vibration and impact, 2017,36 (4): 139-143.) research the feasibility of distinguishing the machining state by extracting acoustic signal features in the laser shot peening process, and provide a signal feature analysis method based on data correlation, wherein the feature value is sensitive to the energy setting and machining state in the laser shot peening process and the distance between sensors, and the direct reason of the change of the feature value is difficult to directly judge. The response process of a water restraint layer covered on the surface of a workpiece under the action of laser impact (Zhao Mimo, a travel down, hu Yong, influence of laser shot blasting process parameters on dynamic response of the water restraint layer [ J ] aviation manufacturing technology, 2021,64 (12): 47-52, 69.) is researched by using a high-speed camera shooting mode of Zhao Mimo and the like, and the method has the advantages that the cost of the adopted high-speed camera is higher, and the processing quantity of high-frame-rate two-dimensional image data is large. Zyongkang et al proposed a scheme for completing online detection by fixing a plurality of acoustic sensors in a hard hemispherical shell and collecting and analyzing laser shot-peening acoustic signal waveforms (Zyongkang, qin Haoyong, liyang, zhou Li Chun, zhangyang, laser shock peening online detection method and device based on shock wave waveform characteristics, 2012, CN101482542B). However, the number of microphones used in this solution is large and the limitation on the installation position is large, which is not favorable for practical application. The xie pillar and the like (xie pillar, wangpeng, longjiang swimming, huwei, zhangyu beam, penqing, laser-induced breakdown spectroscopy and acoustic reflection combined online monitoring system and method, 2021, cn112255191a) disclose a monitoring system using breakdown spectroscopy and acoustic reflection combined, however, the system is complex in composition, particularly, the used spectral analysis technology needs to shoot light of plasma radiation during shot blasting processing, needs to use an ICCD camera and other equipment, is high in cost, and is easily interfered by ambient light.
Therefore, the existing laser peening online monitoring technical scheme has the following defects:
(1) In the existing scheme based on acoustic signal characteristics, acousto-optic joint acquisition is mostly adopted, and a mode of detecting optical characteristics (image recognition or spectral analysis) and acoustic emission detection in parallel is adopted to monitor the processing condition. The optical characteristic detection result is greatly influenced by the environment; the mounting position of the acoustic emission sensor is required to be close to or tightly attached to the workpiece, so that equipment is not convenient to mount, and the acoustic emission acquisition result is easily influenced by the shape of the machined workpiece.
(2) The existing state identification method adopts a data analysis means to obtain higher sampling frequency of signals, so that the acquisition rate of a sensor is higher, the data processing amount in unit time is large, and real-time identification is difficult to realize.
The invention aims to overcome the defects in the conventional technical scheme and provide a laser peening state (including the states of a constraint layer and an absorption layer) real-time online monitoring system based on machine learning.
Patent document CN113312825A (application number: cn202110678982. X) discloses a method and device for monitoring laser peening effect, in which a new resonance condition obtained based on a first laser peening strategy is compared with an initial resonance condition to obtain a first laser peening strategy meeting requirements, and the fatigue life of a part is prolonged by changing the resonance conditions such as resonance vibration frequency and amplitude of the part. Meanwhile, a real measurement value of the part to be processed after laser peening is obtained through a vibration response technology, comparison analysis is carried out on the basis of the real measurement value and vibration mode characteristics, a more accurate second laser peening strategy is obtained, a corresponding real measurement value based on the second laser peening strategy is obtained, and therefore the optimal laser peening strategy is determined through iteration processing. However, the invention adopts a data analysis means to obtain higher sampling frequency of signals, so that the acquisition rate of the sensor is higher, the data processing amount in unit time is large, and the real-time identification is difficult to realize.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a real-time online monitoring system for the laser peening processing state based on machine learning.
The invention provides a real-time online monitoring system for laser peening machining state based on machine learning, which comprises:
laser shot-peening module: the laser sends out laser pulses, and the laser pulses are guided by a control light path to act on a processing workpiece which is covered with a constraint layer and an absorption layer and is clamped by a clamp to generate a laser shot blasting sound source;
a control module: the controller is connected with the FPGA processor, the upper computer and the laser; the controller reads the light emitting condition of the laser and controls the triggering and stopping of the acoustic sensor and the data acquisition card;
the sensing module: the position and the direction of the acoustic sensor have no special limit requirements, the acoustic sensor is connected with a data acquisition card, and the data acquisition card is connected with an FPGA processor; the acoustic sensor receives the laser shot blasting acoustic signal, and the laser shot blasting acoustic signal is acquired and converted into a digital signal by a data acquisition card and transmitted to the FPGA processor;
a data processing module: the FPGA processor extracts acoustic features of the acoustic signals, judges whether the laser shot blasting state is normal or not, and the upper computer displays processing results uploaded by the FPGA processor.
According to the real-time online monitoring method for the laser peening state based on machine learning, which is provided by the invention, the real-time online monitoring system for the laser peening state based on machine learning is adopted, and the implementation comprises the following steps:
step S1: establishing a machine learning model;
step S2: performing laser shot blasting, reading a light emitting signal fed back by a laser by a control module, triggering a sensing module to record a laser shot blasting acoustic signal, sending laser shot blasting acoustic signal data to a data processing module, and extracting multi-dimensional acoustic characteristic parameters of the acoustic signal;
and step S3: carrying out dimension reduction processing on the multi-dimensional acoustic characteristic parameters to obtain characteristic discrimination parameters of the multi-dimensional acoustic characteristic parameters;
and step S4: sending the acoustic characteristic discrimination parameters to a machine learning classification model, identifying the current processing condition, and feeding back the result to an upper computer;
step S5: if the current machining state result is that machining is not finished, jumping to the step S2;
and if the current machining state result is that the machining is finished, finishing the operation.
Preferably, in the step S1:
the establishment of the machine learning model comprises the following steps:
step S1.1: respectively carrying out laser shot blasting under different conditions of normal processing and abnormal processing of a workpiece, recording sound signal samples generated in the laser shot blasting process through an acoustic sensor, and establishing a shot blasting sound signal sample library in a known processing state;
step S1.2: extracting multi-dimensional acoustic parameters and performing linear dimensionality reduction processing on each acquired pulse sound signal to obtain characteristic discrimination parameters of the pulse sound signal, and forming characteristic discrimination parameter sets of shot blasting sound signals under different processing conditions;
the acoustic parameter dimension reduction processing method is a linear supervised learning method, and a projection mode that the difference of the data in the processed similar samples is minimum and the difference of the data in the different samples is maximum is searched to be used as an adopted dimension reduction method;
step S1.3: and taking the acoustic signal characteristic discrimination parameter set under the known processing condition as a training sample, and establishing a classification model for identifying the processing state in the laser peening process based on a machine learning method.
Preferably, in the step S2:
the controller is used for reading the operating process parameters of the laser, controlling all devices of the online identification system and feeding back the identified processing state result to the host computer, and in the shot blasting process, the controller triggers and stops the acoustic sensor and the data acquisition card by reading the light emitting instruction of the laser and the light emitting signal fed back by the laser;
the acoustic sensor is used for collecting laser shot blasting acoustic signals transmitted in the air and converting the received acoustic signals into electric signals to be transmitted to the data acquisition card; the data acquisition card is responsible for converting the received analog signals into digital signals for data processing; an acoustic sensor is adopted to collect acoustic signals in the laser shot blasting process, and the installation position and direction of the acoustic sensor have no special limit requirements;
and the FPGA processor is used for extracting acoustic characteristic parameters from the received sound signals and judging whether the laser peening state is normal or not based on the machine learning classification model.
Preferably, an acoustic sensor is used as the acquisition device, and an acoustic signal generated in the laser peening process is propagated in the form of spherical waves, so that the energy distribution is uniform in all directions; meanwhile, the processing state is judged by means of various acoustic characteristics, and the dependence on the energy of the obtained acoustic signal is low.
Preferably, in the recording process, the controller reads the light emitting instruction of the laser, and determines the acquisition length of the single laser shot blasting sound signal according to the laser pulse interval, so as to ensure that the acquisition is finished before the next laser pulse is emitted.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention can detect the state states of the constraint layer and the absorption layer in the laser shot blasting process only by depending on the acoustic signal, has less requirement on equipment and lower cost;
2. the acoustic sensor is used for collecting acoustic signals in the processing process instead of the acoustic emission sensor, so that the requirement on the sampling rate is low, and the real-time performance is strong;
3. the sensor is freely arranged, does not need to be tightly attached to the surface of a material, is convenient to install, and has wider applicable scenes.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is a real-time on-line monitoring system for laser peening state;
FIG. 2 is a method for establishing a linear discriminant analysis-based dimension reduction (LDA) and Support Vector Machine (SVM) classification model;
FIG. 3 is a flow chart of the operation of a linear discriminant analysis dimension reduction (LDA) and Support Vector Machine (SVM) recognition program;
FIG. 4 is a diagram illustrating a dimension reduction result of the linear discriminant analysis method;
FIG. 5 is a diagram illustrating classification results of a support vector machine model.
In the figure:
1 is a laser;
2 is a control light path;
3 is a constraint layer;
4 is an absorption layer;
5, a clamp;
6, processing a workpiece;
7 is a host computer;
8 is a controller;
9 is an FPGA processor;
10 is an acoustic microphone;
and 11, a data acquisition card.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
Example 1:
according to the present invention, a real-time online monitoring system for laser peening processing status based on machine learning is provided, as shown in fig. 1 to 5, comprising:
laser shot peening module: the laser 1 emits laser pulses, and the laser pulses are guided by the control light path 2 to act on a processing workpiece 6 which is covered with the constraint layer 3 and the absorption layer 4 and is clamped by the clamp 5 to generate a laser shot blasting sound source;
a control module: the controller 8 is connected with the FPGA processor 9, the controller 8 is connected with the upper computer 7, and the controller 8 is connected with the laser 1; the controller 8 reads the light emitting condition of the laser 1 and controls the triggering and stopping of the acoustic sensor 10 and the data acquisition card 11;
the sensing module: the position and the direction of the acoustic sensor 10 have no special limit requirements, the acoustic sensor 10 is connected with a data acquisition card 11, and the data acquisition card 11 is connected with an FPGA processor 9; the acoustic sensor 10 receives the laser shot blasting acoustic signal, and the signal is acquired and converted into a digital signal by the data acquisition card 11 and transmitted to the FPGA processor 9;
a data processing module: the FPGA processor 9 extracts the acoustic characteristics of the acoustic signal, judges whether the laser shot peening state is normal or not, and the upper computer 7 displays the processing result uploaded by the FPGA processor 9.
According to the real-time online monitoring method for the laser peening state based on machine learning, which is provided by the invention, the real-time online monitoring system for the laser peening state based on machine learning is adopted, and the implementation comprises the following steps:
step S1: establishing a machine learning model;
step S2: performing laser shot blasting, reading a light emitting signal fed back by a laser by a control module, triggering a sensing module to record a laser shot blasting acoustic signal, sending laser shot blasting acoustic signal data to a data processing module, and extracting multi-dimensional acoustic characteristic parameters of the acoustic signal;
and step S3: carrying out dimension reduction processing on the multi-dimensional acoustic characteristic parameters to obtain characteristic discrimination parameters of the multi-dimensional acoustic characteristic parameters;
and step S4: sending the acoustic characteristic discrimination parameters to a machine learning classification model, identifying the current processing condition, and feeding back the result to an upper computer;
step S5: if the current machining state result is that machining is not finished, jumping to the step S2;
and if the current machining state result is that the machining is finished, finishing the operation.
Specifically, in the step S1:
the establishment of the machine learning model comprises the following steps:
step S1.1: respectively carrying out laser shot blasting under different conditions of normal processing and abnormal processing of a workpiece, recording sound signal samples generated in the laser shot blasting process through an acoustic sensor 10, and establishing a shot blasting sound signal sample library in a known processing state;
step S1.2: extracting multi-dimensional acoustic parameters and performing linear dimensionality reduction processing on each acquired pulse sound signal to obtain characteristic discrimination parameters of the pulse sound signals, and forming characteristic discrimination parameter sets of the shot blasting sound signals under different processing conditions;
the acoustic parameter dimension reduction processing method is a linear supervised learning method, and a projection mode that the difference of the processed data in the same type of samples is minimum and the difference of the processed data in different types of samples is maximum is searched as an adopted dimension reduction method;
step S1.3: and taking the acoustic signal characteristic discrimination parameter set under the known processing condition as a training sample, and establishing a classification model for identifying the processing state in the laser peening process based on a machine learning method.
Specifically, in the step S2:
the controller 8 is used for reading the operation process parameters of the laser, controlling all devices of the online identification system and feeding back the identified processing state result to the host computer 7, and in the shot blasting process, the controller 8 triggers and stops the acoustic sensor 10 and the data acquisition card 11 by reading the light emitting instruction of the down-sending laser 1 and the light emitting signal fed back by the laser 1;
the acoustic sensor 10 is used for collecting laser shot blasting acoustic signals transmitted in the air and converting the received acoustic signals into electric signals to be transmitted to the data acquisition card 11; the data acquisition card 11 is responsible for converting the received analog signals into digital signals for data processing; an acoustic sensor 10 is adopted to collect acoustic signals in the laser shot blasting process, and the installation position and direction of the acoustic sensor 10 have no special limit requirements;
the FPGA processor 9 is responsible for extracting acoustic characteristic parameters from the received sound signals and judging whether the laser peening state is normal based on the machine learning classification model.
Specifically, using the acoustic sensor 10 as a collecting device, the acoustic signal generated during the laser peening process propagates in the form of a spherical wave, with uniform energy distribution in each direction; meanwhile, the processing state is judged by means of various acoustic characteristics, and the dependence on the energy of the obtained acoustic signal is low.
Specifically, in the recording process, the controller 8 reads the light emitting instruction of the laser 1, and determines the acquisition length of the single laser shot blasting sound signal according to the laser pulse interval, so as to ensure that the acquisition is finished before the next laser pulse is emitted.
Example 2:
example 2 is a preferred example of example 1, and the present invention will be described in more detail.
The invention is realized by the following technical scheme, and the laser shot peening constrained layer state real-time online monitoring system based on machine learning comprises the following components: the device comprises a control module, a sensing module and a data processing module.
The control module is used for reading the operating process parameters of the laser, controlling all devices of the online identification system and feeding back the identified processing state results to the upper computer. In the shot blasting process, the control module triggers and stops the sensing module by reading the light emitting instruction of the down-sending laser 1 and the light emitting signal fed back by the laser 1.
The sensing module comprises an acoustic sensor 10 and a data acquisition card 11, wherein the acoustic sensor 10 is used for acquiring laser shot blasting acoustic signals transmitted in the air and converting the received acoustic signals into electric signals to be transmitted to the data acquisition card 11; the data acquisition card 11 is responsible for converting the received analog signals into digital signals for data processing.
The sensing module adopts the acoustic sensor 10 to collect the acoustic signal of the laser shot peening process, and the installation position and the direction of the acoustic sensor 10 have no special limit requirements.
And the data processing module is used for extracting acoustic characteristic parameters from the received sound signals and judging whether the laser peening state is normal or not based on the machine learning classification model.
The present invention uses the acoustic sensor 10 as the acquisition device by the monitoring system. The acoustic signal generated in the laser shot blasting process is transmitted in the form of spherical waves, and the energy distribution is uniform in all directions; meanwhile, the processing state is judged by analyzing various acoustic characteristics including waveform contours, frequency spectrum characteristics and the like, and the dependence degree on the energy of the obtained acoustic signal is low, so that the optional range of the arrangement of the acoustic sensor 10 on the direction, the position and the distance relative to a workpiece is wide, the position of the acoustic sensor 10 is freely arranged, and the judgment accuracy rate is not influenced.
Specifically, the establishment of the machine learning model may consist of the following steps:
s1: the laser peening processing is carried out for a plurality of times under different conditions of normal processing and abnormal processing (constraint layer missing, absorption layer missing or both missing) of the workpiece respectively, sound signal samples generated in the laser peening process are recorded through the sound sensor 10, and a peening sound signal sample library with a known processing state is established.
S2: and extracting multi-dimensional acoustic parameters of each acquired pulse sound signal and performing linear dimension reduction processing to obtain characteristic discrimination parameters of the pulse sound signal, and forming characteristic discrimination parameter sets of the shot blasting sound signals under different processing conditions.
In order to ensure the accuracy of the system for judging the processing state and eliminate the misjudgment caused by the interference of the accidental events on the acoustic signals in the processing process, the invention collects various acoustic characteristics of the laser shot blasting acoustic signals to carry out joint judgment and analysis. The method has the advantages that the method can still make accurate judgment when the individual indexes fluctuate due to external interference; the disadvantage is that the amount of data required to be transmitted and processed increases dramatically for the same length of acoustic signal. Therefore, the scheme of the invention performs linear dimensionality reduction processing on the acquired multidimensional parameters after the acquisition is finished, and greatly compresses the data volume required to be transmitted and processed on the premise of keeping most of the original information.
And the characteristic discrimination parameters in the step S2 are obtained by carrying out dimension reduction processing on the laser shot blasting multi-dimensional acoustic parameters.
The acoustic parameter dimension reduction processing method is a linear supervised learning method and is characterized in that a projection mode that the difference of the processed data in the same type of samples is minimum and the difference of the processed data in different types of samples is maximum is searched to serve as a finally adopted dimension reduction method.
S3: and taking the acoustic signal characteristic discrimination parameter set under the known processing condition as a training sample, and establishing a classification model for identifying the processing state in the laser peening process based on a machine learning method. By using the machine learning method, the difference between the shot blasting acoustic signals of different processing states can be fully utilized. Meanwhile, it is noted that the recorded laser peening audio frequency may also be different according to different processing environments such as processing equipment and indoor environment. The model can be respectively established according to different processing conditions by using the analysis method of machine learning, so that the scheme can have good adaptability under different processing conditions and processing environments, and high judgment accuracy is kept.
After the model is built, the shot blasting in the machining process can be monitored in real time through the acoustic signal, and the process comprises the following steps:
s1: developing a laser shot blasting experiment;
s2: after reading the light emitting signal fed back by the laser, the control module triggers the acoustic sensing module to record the laser shot peening acoustic signal, then sends the laser shot peening acoustic signal data to the data processing module, and extracts the multi-dimensional acoustic characteristic parameters of the acoustic signal;
s3: performing dimension reduction processing on the multi-dimensional acoustic characteristic parameters by adopting the linear supervised learning method to obtain characteristic discrimination parameters of the multi-dimensional acoustic characteristic parameters;
s4: the acoustic feature discrimination parameters are sent to the established machine learning classification model, the current processing condition is identified, and the result is fed back to the upper computer 7;
s5: and repeating the steps from S2 to S4 until the shot peening is finished.
Specifically, in the recording process, the light emitting instruction of the laser 1 is read through the control module, the acquisition length of the single laser shot blasting sound signal is determined according to the laser pulse interval, and the acquisition is guaranteed to be finished before the next laser pulse is sent out.
Example 3:
example 3 is a preferred example of example 1, and the present invention will be described in more detail.
Fig. 1 shows a real-time online monitoring system for laser peening processing state based on machine learning according to the embodiment, which includes: 1. a laser; 2. controlling the light path; 3. a constraining layer (water); 4. an absorbing layer (black tape); 5. a clamp; 6. processing a workpiece; 7. and (4) a host computer.
The real-time identification system consists of three parts:
a control module: 8. a controller;
a data processing module: 9, FPGA processor;
the sensing module: 10. an acoustic microphone; 11. a data acquisition card.
Among the components, the components 1 to 6 are laser shot peening systems, and a laser 1 emits laser pulses which are guided by a light path 2 to act on a sample 6 which is covered with a restraint layer (water) 3 and an absorption layer (black adhesive tape) 4 and is clamped by a clamp 5. The controller 8 reads the light emitting condition of the laser 1, the control components 9-11 collect and judge sound data, and the upper computer 7 of the components displays results to carry out man-machine interaction.
The components 1 to 6 are responsible for generating a laser shot blasting sound source, the acoustic sensor 10 and the data acquisition card 11 are used for acquiring acoustic signals, the FPGA processor 9 is used for extracting acoustic characteristics of the acoustic signals, and the upper computer 7 and the controller 8 are used for judging whether the laser shot blasting process is normal or not. The controller 8 controls the acoustic sensor 10 to receive the laser shot blasting acoustic signal by taking the light-emitting signal of the laser shot blasting processing system as a trigger signal, and the light-emitting signal is acquired and converted into a digital signal by the data acquisition card 11 and then transmitted to the processor 9, and finally the processing result is uploaded to the upper computer 7.
Fig. 2 shows an embodiment of the method for on-line identification of laser peening processing state based on acoustic signal characteristics according to the present invention. In the laser peening experiment, the laser peening experiment is performed using the nanosecond laser 1.
The comparative conditions set for the experiment are the case of the constraining layer 3 on the surface of the sample. Under normal processing conditions, a thin water layer is additionally coated on the absorption layer 4 on the surface of the sample to serve as the restraint layer 3, while the restraint layer 3 is not present in the comparative experiment.
The acoustic transducer 10 is used to record the audio of a continuous pulse shot at a location some distance from the point of laser action. A total of 1500 acoustic signal samples were taken for the experiment, with 750 samples each being occupied by the experimental setup with and without the constraining layer 3 present.
For the audio frequency of each shot blasting pulse, intercepting a section of acoustic signal sequence behind the maximum value of a pulse peak in each pulse audio frequency, and extracting specific acoustic characteristic parameters; and then carrying out dimension reduction processing on the acquired shot blasting acoustic signal characteristic parameters. And (4) judging parameters by extracting the characteristics of the shot blasting sound signals under different constraint layer 3 conditions.
And randomly extracting samples in each group of samples to be used as a training set, standardizing the rest samples to be used as a test set, and then performing dimension reduction treatment on the training set and the test set by using a linear discriminant analysis method. Linear Discriminant Analysis (LDA) is a dimension reduction technology for supervised learning, and the basic idea is as follows: by giving a training sample set, trying to project samples into a space with lower dimensionality, and enabling intra-class variance after projection to be minimum and inter-class variance to be maximum, information which is helpful for finishing classification is extracted, and redundant data are eliminated. The method has the advantage that the data volume can be greatly reduced on the basis of ensuring the difference between different types of data as much as possible. The results of the dimension reduction analysis are shown in FIG. 4.
Aiming at the characteristics of the shot blasting acoustic signals subjected to linear discriminant analysis and dimensionality reduction, a support vector machine model is adopted to perform identification experiments on acoustic signal samples under different water constraint layers 3. The Support Vector Machine (SVM) is a generalized linear classifier for binary classification of data in a supervised learning mode, and has the advantages that a nonlinear classification task can be processed after a kernel technique is adopted, a final decision function is determined by a small number of Support vectors, the calculation amount is small, and real-time processing is facilitated. The recognition results are shown in fig. 5. The overall recognition accuracy of the final model was 98.2%.
Those skilled in the art will appreciate that, in addition to implementing the systems, apparatus, and various modules thereof provided by the present invention in purely computer readable program code, the same procedures can be implemented entirely by logically programming method steps such that the systems, apparatus, and various modules thereof are provided in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system, the apparatus, and the modules thereof provided by the present invention may be considered as a hardware component, and the modules included in the system, the apparatus, and the modules for implementing various programs may also be considered as structures in the hardware component; modules for performing various functions may also be considered to be both software programs for performing the methods and structures within hardware components.
The foregoing description has described specific embodiments of the present invention. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (5)
1. A real-time online monitoring method based on machine learning laser shot peening processing state is characterized in that a real-time online monitoring system based on machine learning laser shot peening processing state is adopted;
the real-time online monitoring system based on machine learning laser peening state includes:
laser shot peening module: the laser (1) emits laser pulses, and the laser pulses are guided by the control light path (2) to act on a processing workpiece (6) which is covered with the restraint layer (3) and the absorption layer (4) and is clamped by the clamp (5) to generate a laser shot blasting sound source;
a control module: the controller (8) is connected with the FPGA processor (9), the controller (8) is connected with the upper computer (7), and the controller (8) is connected with the laser (1); the controller (8) reads the light emitting condition of the laser (1) and controls the triggering and stopping of the acoustic sensor (10) and the data acquisition card (11);
the sensing module: the position and the direction of the acoustic sensor (10) have no special limit requirements, the acoustic sensor (10) is connected with a data acquisition card (11), and the data acquisition card (11) is connected with an FPGA processor (9); the acoustic sensor (10) receives the laser shot blasting acoustic signal, and the laser shot blasting acoustic signal is acquired and converted into a digital signal by the data acquisition card (11) and transmitted to the FPGA processor (9);
a data processing module: the FPGA processor (9) extracts the acoustic characteristics of the acoustic signals, judges whether the laser shot blasting processing state is normal or not, and the upper computer (7) displays the processing result uploaded by the FPGA processor (9);
the real-time online monitoring method for the laser peening state based on machine learning comprises the following steps:
step S1: establishing a machine learning model;
step S2: performing laser shot blasting, reading a light emitting signal fed back by a laser by a control module, triggering a sensing module to record a laser shot blasting acoustic signal, sending laser shot blasting acoustic signal data to a data processing module, and extracting multi-dimensional acoustic characteristic parameters of the acoustic signal;
and step S3: carrying out dimension reduction processing on the multi-dimensional acoustic characteristic parameters to obtain characteristic discrimination parameters of the multi-dimensional acoustic characteristic parameters;
and step S4: sending the acoustic characteristic discrimination parameters to a machine learning classification model, identifying the current processing condition, and feeding back the result to an upper computer;
step S5: if the current machining state result is that machining is not finished, jumping to the step S2;
and if the current machining state result is that the machining is finished, finishing the operation.
2. The real-time online monitoring method for laser peening machining state based on machine learning according to claim 1, wherein in the step S1:
the establishment of the machine learning model comprises the following steps:
step S1.1: respectively carrying out laser shot blasting under different conditions of normal processing and abnormal processing of a workpiece, recording sound signal samples generated in the laser shot blasting process through an acoustic sensor (10), and establishing a shot blasting sound signal sample library in a known processing state;
step S1.2: extracting multi-dimensional acoustic parameters and performing linear dimensionality reduction processing on each acquired pulse sound signal to obtain characteristic discrimination parameters of the pulse sound signals, and forming characteristic discrimination parameter sets of the shot blasting sound signals under different processing conditions;
the acoustic parameter dimension reduction processing method is a linear supervised learning method, and a projection mode that the difference of the processed data in the same type of samples is minimum and the difference of the processed data in different types of samples is maximum is searched as an adopted dimension reduction method;
step S1.3: and taking the acoustic signal characteristic discrimination parameter set under the known processing condition as a training sample, and establishing a classification model for identifying the processing state in the laser peening process based on a machine learning method.
3. The real-time online monitoring method for laser peening machining state based on machine learning according to claim 1, wherein in the step S2:
the controller (8) is used for reading the operation process parameters of the laser, controlling all devices of the online monitoring system and feeding back the identified processing state result to the host computer (7), and in the shot blasting process, the controller (8) triggers and stops the acoustic sensor (10) and the data acquisition card (11) by reading the light emitting instruction of the down-sending laser (1) and the light emitting signal fed back by the laser (1);
the acoustic sensor (10) is used for collecting laser shot blasting acoustic signals transmitted in the air and converting the received acoustic signals into electric signals to be transmitted to the data acquisition card (11); the data acquisition card (11) is responsible for converting the received analog signals into digital signals for data processing; an acoustic sensor (10) is adopted to collect acoustic signals in the laser shot blasting process, and the installation position and direction of the acoustic sensor (10) have no special limit requirements;
the FPGA processor (9) is responsible for extracting acoustic characteristic parameters from the received sound signals and judging whether the laser peening state is normal or not based on the machine learning classification model.
4. The real-time online monitoring method for the laser peening machining state based on machine learning of claim 1, wherein:
an acoustic sensor (10) is used as a collecting device, acoustic signals generated in the laser shot blasting process are transmitted in the form of spherical waves, and the energy distribution is uniform in all directions; meanwhile, the processing state is judged by means of various acoustic characteristics, and the dependence on the energy of the obtained acoustic signal is low.
5. The real-time online monitoring method for the laser peening machining state based on machine learning of claim 1, wherein:
in the recording process, a light emitting instruction of the laser (1) is read through the controller (8), and the acquisition length of a single laser shot blasting sound signal is determined according to the laser pulse interval, so that the acquisition is finished before the next laser pulse is emitted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210210094.XA CN114577334B (en) | 2022-03-03 | 2022-03-03 | Real-time online monitoring method and system for laser shot blasting processing state based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210210094.XA CN114577334B (en) | 2022-03-03 | 2022-03-03 | Real-time online monitoring method and system for laser shot blasting processing state based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114577334A CN114577334A (en) | 2022-06-03 |
| CN114577334B true CN114577334B (en) | 2022-12-09 |
Family
ID=81772954
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210210094.XA Active CN114577334B (en) | 2022-03-03 | 2022-03-03 | Real-time online monitoring method and system for laser shot blasting processing state based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114577334B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116975938B (en) * | 2023-09-25 | 2023-11-24 | 北京谷器数据科技有限公司 | Sensor data processing method in product manufacturing process |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7816622B2 (en) * | 2007-09-28 | 2010-10-19 | General Electric Company | System and method for controlling laser shock peening |
| CN103207178A (en) * | 2013-02-04 | 2013-07-17 | 中国航空工业集团公司北京航空制造工程研究所 | Online testing method and apparatus for laser shock peening effect |
| GB201510220D0 (en) * | 2015-06-11 | 2015-07-29 | Renishaw Plc | Additive manufacturing apparatus and method |
| CN110157895A (en) * | 2019-07-09 | 2019-08-23 | 西安天瑞达光电技术股份有限公司 | A kind of laser impact intensified device and method |
| CN110438333A (en) * | 2019-09-19 | 2019-11-12 | 中国人民解放军空军工程大学 | A kind of kHz low-energy laser scanning impact reinforcing method and its device |
| CN113312825B (en) * | 2021-06-18 | 2022-03-22 | 广东工业大学 | Laser shot peening effect monitoring method and device |
| CN114115047A (en) * | 2021-11-29 | 2022-03-01 | 上海智能制造功能平台有限公司 | Pulse laser light-emitting synchronous closed-loop control system, method, terminal and medium |
-
2022
- 2022-03-03 CN CN202210210094.XA patent/CN114577334B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114577334A (en) | 2022-06-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhou et al. | Review of tool condition monitoring methods in milling processes | |
| US6384995B1 (en) | Apparatus and method for detecting defects in data storage devices | |
| CN111044621B (en) | Nondestructive testing system and method based on sound quality and acoustic characteristics | |
| US10043064B2 (en) | Method and apparatus of detecting object using event-based sensor | |
| CN114577334B (en) | Real-time online monitoring method and system for laser shot blasting processing state based on machine learning | |
| US9714884B2 (en) | Method and device for recognizing state of noise-generating machine to be investigated | |
| KR101996489B1 (en) | Working error detecting apparatus for automatic manufacturing line | |
| Hou et al. | Research on audio-visual detection method for conveyor belt longitudinal tear | |
| US20200409653A1 (en) | Intelligent Audio Analytic Apparatus (IAAA) and Method for Space System | |
| JP2012018066A (en) | Device for inspecting abnormality | |
| Emamian et al. | Robust clustering of acoustic emission signals using the Kohonen network | |
| Pan et al. | Cognitive acoustic analytics service for Internet of Things | |
| TW202221624A (en) | Stamping quality inspection system and stamping quality inspection method | |
| KR20220147918A (en) | Crack sensor device and control method thereof | |
| Ezzat et al. | Discriminative word-spotting using ordered spectro-temporal patch features. | |
| CN112781721B (en) | Intelligent noise acquisition method and intelligent noise acquisition device of mobile inspection platform | |
| CN103492234A (en) | Method and device for evaluating structure-borne sound during a collision of a vehicle | |
| US20040143398A1 (en) | Method and system for monitoring vibration and/or mechanical waves in mechanical systems | |
| CN107398612A (en) | A kind of electric spark clearance discharge condition detection apparatus and method based on cluster analysis | |
| Clocker et al. | Autonomous Sensor System for Wind Turbine Blade Collision Detection | |
| CN110346032A (en) | A kind of Φ-OTDR vibration signal end-point detecting method combined based on constant false alarm with zero-crossing rate | |
| CN104991245A (en) | Unmanned aerial vehicle early warning apparatus and early warning method thereof | |
| CN205003281U (en) | Unmanned vehicles early warning device | |
| EP2473314A1 (en) | Method for classifying a laser process and a laser material processing head using the same | |
| EP3934271A1 (en) | Apparatus, method and computer program for detecting defects |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |