Method and system for identifying welding defects of rotary drilling barrel

Abstract

The invention relates to the technical field of welding quality detection, in particular to a method and a system for identifying welding defects of a rotary drilling barrel, and the method comprises the following steps: in the welding process of the rotary drilling barrel, collecting welding data in real time through a sensor; data preprocessing is conducted on the collected welding data according to data types, and feature extraction is conducted on the collected welding data; after data preprocessing and feature extraction are completed, a deep learning algorithm is adopted to construct a rotary drilling barrel welding defect recognition model and train the model; and after model training is completed, inputting the data acquired in real time after data preprocessing and feature extraction is completed into the model for defect identification, and outputting an identification result. According to the method, various defects generated in the welding process of the rotary excavating drill cylinder are accurately recognized through the method of combining mixed data processing with the deep learning algorithm, and the accuracy and the automation level of recognition of the welding defects of the rotary excavating drill cylinder are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of welding quality inspection, and particularly relates to a method and system for identifying welding defects of a rotary drilling bucket. Background Art

[0002] As a key component of a rotary drilling rig, the rotary drilling bucket plays a crucial role in various foundation engineering constructions. A large number of welding processes are involved in the manufacturing process of the rotary drilling bucket, and the welding quality directly determines the strength, rigidity, and service life of the bucket. Traditional detection of welding defects in rotary drilling buckets mainly relies on manual visual inspection and some conventional non-destructive testing methods, such as ultrasonic testing, radiographic testing, etc. Manual visual inspection is inefficient and subjective, and it is easy to cause missed detection or misjudgment due to factors such as the fatigue and experience differences of the inspectors; while traditional non-destructive testing technologies, although having high accuracy, are expensive in equipment, complex in operation, long in detection cycle, and radiographic testing has certain radioactivity, making it difficult to meet the real-time detection requirements on modern high-efficiency production lines. In addition, the complex structure, large size, and diverse welding parts of the rotary drilling bucket further increase the difficulty of welding defect detection. Therefore, it is of great practical significance to develop a method for identifying welding defects of rotary drilling buckets with high efficiency, accuracy, and high degree of automation. Summary of the Invention

[0003] The present invention aims at the above-mentioned existing technical deficiencies and provides a method and system for identifying welding defects of a rotary drilling bucket.

[0004] The present invention is realized through the following technical solutions:

[0005] A method for identifying welding defects of a rotary drilling bucket is provided, and the method includes the following steps:

[0006] Step S10: During the welding process of the rotary drilling bucket, welding data is collected in real time through a sensor;

[0007] Step S20: The collected welding data is respectively preprocessed according to the data type, including filtering, image enhancement, and smoothing, and feature extraction is respectively performed, including average value, maximum and minimum values, and standard deviation;

[0008] Step S30: Normal welding data and welding data samples containing various defects during the welding process of the rotary drilling bucket are obtained, and a deep learning algorithm is used to construct and train a welding defect identification model for the rotary drilling bucket;

[0009] Step S40: After the model training is completed, the real-time collected data after data preprocessing and feature extraction is input into the model for defect identification and the identification result is output;

[0010] Among them, the sensors used in step S10 include: welding current sensor, welding voltage sensor, thermal imaging sensor and weld tracking sensor;

[0011] Among them, when constructing the rotary drilling barrel welding defect recognition model using the deep learning algorithm in step S30, the model is constructed with the TensorFlow or PyTorch deep learning framework;

[0012] Among them, the recognition results output in step S40 include having defects and no defects. When the recognition result is having defects, an audible and visual alarm is triggered, and the recognized defect type is displayed on the display screen while marking the position where the defect is located and the relevant welding data.

[0013] Preferably, in the process of rotary drilling barrel welding in step S10, the welding data is collected in real time through sensors, and the sensors used include:

[0014] Welding current sensor: Installed in the welding power supply circuit, it is used to accurately collect the real-time value of the welding current and generate a change curve according to the collected real-time value, reflecting the fluctuation of the current during the welding process and the trend of the current changing with time, and is used to judge whether there are welding defects caused by current problems;

[0015] Welding voltage sensor: Installed in the welding power supply circuit and connected to the welding electrode, it is used to collect the real-time data of the welding voltage, reflecting the change of the potential difference between the two ends of the electrode during the welding process, and is used to judge whether there are welding defects caused by voltage problems;

[0016] Thermal imaging sensor: It includes an infrared thermal imager, which is used to monitor the temperature distribution during the welding process in real time, obtain the thermal imaging diagram during the welding process, and display the temperature changes of the welding molten pool and the heat affected area in the form of an image, including the highest temperature, the lowest temperature and the temperature gradient information during the welding process, and is used to judge whether there are welding defects caused by heat conduction problems;

[0017] Weld tracking sensor: Installed within the range of the welding path, it is used to collect the position information data of the welding torch head and the actual forming data of the weld, including the width, depth and shape deviation of the weld, and is used to judge whether there are defects such as undercut, incomplete fusion and poor weld formation in the weld.

[0018] Preferably, in step S20, the collected welding data is preprocessed according to the data type, including:

[0019] Filtering and denoising: Perform filtering and denoising processing on the collected welding current and voltage data. Use a Butterworth filter to set the cut-off frequency according to the sampling frequency and noise characteristics of the data to remove the noise interference in the signal;

[0020] Image enhancement processing: Perform image enhancement processing on the thermal imaging map collected during the welding process to increase the temperature difference contrast in the thermal imaging map, and use the method of gray-scale stretching to enhance the image;

[0021] Data smoothing processing: Use the moving average method to smooth the collected position information data to eliminate the small fluctuations caused by sensor jitter or measurement errors.

[0022] Among them, the steps of enhancing the image by using the gray-scale stretching method in the image enhancement processing include:

[0023] Read the thermal imaging map: Read the thermal imaging map collected by the thermal imaging sensor into the computer program. The thermal imaging map is a gray-scale map, and the gray value of each pixel represents the temperature information at that position;

[0024] Obtain the gray-scale histogram of the heat image: Count the frequencies of each gray level in the thermal imaging map to obtain the gray-scale histogram of the image; By analyzing the histogram, understand the general situation of the temperature distribution in the image, such as the temperature range, the number of pixels in different temperature intervals, etc.;

[0025] Determine the stretching interval: After excluding the highest 3 gray values and the lowest 3 gray values in the gray-scale histogram, determine the minimum effective gray value min_val and the maximum effective gray value max_val, and set the target gray interval after stretching to [new_min, new_max], where new_min is the minimum value after stretching and new_max is the maximum value after stretching;

[0026] Gray-scale stretching calculation: For each pixel point (x, y) in the thermal imaging map, its original gray value is old_val, and its gray-scale stretching calculation formula is shown in Equation (1):

[0027]

[0028] Among them, new_val is the gray value after stretching;

[0029] Update the image pixel value: Update the calculated stretched gray value new_val to the corresponding pixel point in the thermal imaging map. After traversing all pixel points in the thermal imaging map, complete the gray-scale stretching processing of the image.

[0030] Among them, the steps of using the moving average method to smooth the collected position information data in the data smoothing processing include:

[0031] Determine the window size of the moving average: The window represents the number of data points participating in the average calculation, and set the window size to window_size;

[0032] Moving average calculation: Starting from the k-th data point, determine the position of k according to Equation (2) as shown in Equation (2):

[0033]

[0034] Among them, window_size is the set window size. Calculate the moving average for each data point, and the calculation formula is as shown in Equation (3):

[0035]

[0036] Among them, y(n) is the moving average of the n-th data point, i is the retrieval index, ranging from n - k to n + k, and x(i) is the i-th data point within the range of n - k to n + k; after traversing the collected position information data, the smoothing process of the data is completed.

[0037] Preferably, the feature extraction in step S20 includes:

[0038] Extract features from the welding current data, including the average current, current standard deviation, current peak value, current frequency, etc. These parameters can reflect the energy input stability and arc behavior characteristics during the welding process;

[0039] Extract features from the welding voltage data, including the average voltage, voltage change rate, and number of voltage spikes, etc., for analyzing the stability of the welding arc and the rationality of the welding process parameters;

[0040] Extract features from the thermal imaging map, including the highest temperature, lowest temperature, average temperature, and temperature gradient in the thermal imaging map, etc. An abnormal change in the temperature gradient indicates that there are defects during the welding process. For example, cracks will cause abnormal local heat conduction and lead to a sudden change in the temperature gradient;

[0041] Extract features from the position information data of the welding torch and the actual forming data of the weld, including geometric features such as the average weld width, weld width standard deviation, average weld height, and weld edge straightness, etc. These features are directly related to the forming quality and defect type of the weld.

[0042] Preferably, the steps of obtaining normal welding data and welding data samples containing various defects during the rotary drilling cylinder welding process in step S30, and constructing and training a rotary drilling cylinder welding defect recognition model using a deep learning algorithm include:

[0043] Dataset construction: Obtain normal welding data and welding data samples containing various defects (such as pores, slag inclusions, cracks, incomplete penetration, lack of fusion, undercut, etc.) during the rotary drilling cylinder welding process, perform the above data preprocessing and feature extraction operations on the sample data, and divide them into a training set and a validation set according to a ratio of 8:2;

[0044] Model construction: The rotary drilling cylinder welding defect recognition model includes an input layer, a convolutional layer, an average pooling layer, and a fully connected layer; the input of the input layer is the training set divided in the dataset construction step, including welding current, welding voltage, thermal imaging map, weld tracking data, and corresponding features. The number of neuron nodes in the input layer is the same as the number of features of the input data. The convolutional layer is used to extract local features of the input data through convolutional kernels. Different sizes of convolutional kernels are adopted, including 3×3 and 5×5 convolutional kernels distributed crosswise. The number of convolutional layers and convolutional kernels is set according to the data features, and ReLU is used as the activation function. The average pooling layer is set after each convolutional layer and adopts average pooling operation to reduce the data dimension, reduce the computational amount while retaining the main features of the data. The fully connected layer is set in the last layer of the network layer to integrate the feature information extracted from the previous layers. The number of neuron nodes in the fully connected layer is the same as the number of neuron nodes in the input layer.

[0045] Model training and verification: After the model is constructed, set the model parameters, adopt the Adadelta optimization algorithm, set the initial learning rate to 0.01, use the training set as the input to train the model, set the initial training epoch to 50, adopt the cross-entropy loss function as the loss function during the model training process, and use the divided validation set to verify the trained model after each round of training to verify whether the model can accurately identify the rotary drilling cylinder welding defects in the validation set. Stop the model training when the recognition accuracy of the validation set no longer rises in 5 consecutive rounds of verification, and complete the training and verification of the model.

[0046] Preferably, after the model training is completed in step S40, the real-time collected data after data preprocessing and feature extraction is input into the model for defect recognition and the recognition result is output. Record the defect recognition result and the real-time collected data. As the rotary drilling cylinder welding defect recognition progresses, the newly collected rotary drilling cylinder welding data is updated as new sample data to the training set, and the rotary drilling cylinder welding defect recognition model is retrained to continuously update the model version to adapt to the welding defect feature changes that may be caused by factors such as minor changes in the rotary drilling cylinder welding process, differences in raw materials, and equipment wear, ensuring that the model always maintains a high recognition accuracy.

[0047] In addition, to achieve the above object, the present invention also proposes a rotary drilling cylinder welding defect recognition system, and the rotary drilling cylinder welding defect recognition system includes:

[0048] Rotary drilling cylinder welding data acquisition module: used to collect welding data in real time through sensors during the rotary drilling cylinder welding process;

[0049] Rotary drilling cylinder welding data preprocessing and feature extraction module: used to perform data preprocessing on the collected welding data according to the data type, including filtering, image enhancement, and smoothing, and perform feature extraction respectively, including average value, maximum and minimum values, and standard deviation;

[0050] Rotary drilling cylinder welding defect identification model construction and training module: used to construct and train a rotary drilling cylinder welding defect identification model using a deep learning algorithm after data preprocessing and feature extraction are completed;

[0051] Rotary drilling cylinder welding defect identification module: used to input the real-time collected data after data preprocessing and feature extraction into the model for defect identification and output the identification result after the model training is completed;

[0052] The sensors used in the rotary drilling cylinder welding data acquisition module include: welding current sensor, welding voltage sensor, thermal imaging sensor, and weld tracking sensor;

[0053] When constructing a rotary drilling cylinder welding defect identification model using a deep learning algorithm in the rotary drilling cylinder welding defect identification model construction and training module, the model is constructed with the TensorFlow or PyTorch deep learning framework;

[0054] The identification results output by the rotary drilling cylinder welding defect identification module include the existence of defects and no defects. When the identification result is the existence of defects, an audible and visual alarm is triggered, and the identified defect type is displayed on the display screen while marking the location of the defect and the relevant welding data.

[0055] In addition, to achieve the above object, the present invention also proposes a rotary drilling cylinder welding defect identification device, the device includes: a memory, a processor, and a program such as a deep learning-based rotary drilling cylinder welding defect identification algorithm stored on the memory and executable on the processor, and the program such as the deep learning-based rotary drilling cylinder welding defect identification algorithm is to implement the steps of a rotary drilling cylinder welding defect identification method as described above.

[0056] In addition, to achieve the above object, the present invention also provides a computer program product, the computer program product includes a program such as a deep learning-based rotary drilling cylinder welding defect identification algorithm, and when the program such as the deep learning-based rotary drilling cylinder welding defect identification algorithm is executed by a processor, it implements a rotary drilling cylinder welding defect identification method as described above.

[0057] The advantages and effects of the present invention are:

[0058] A method and system for identifying welding defects of a rotary drilling cylinder proposed by the present invention adopt a method of combining comprehensive data processing technology with deep learning algorithms. By separately processing various different types of data and extracting features, it can quickly and accurately identify the welding defects of the rotary drilling cylinder, effectively improve the welding quality and efficiency of the rotary drilling cylinder. Through automated data processing and defect identification without manual participation, it greatly saves time and labor costs, shortens the identification time of the welding defects of the rotary drilling cylinder, and improves the detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0060] Figure 1 It is a flowchart of a method for identifying welding defects of a rotary drilling cylinder according to the present invention.

[0061] Figure 2 It is a schematic structural diagram of a system for identifying welding defects of a rotary drilling cylinder according to the present invention.

[0062] Figure 3 It is a schematic block diagram of the structure of an electronic device for identifying welding defects of a rotary drilling cylinder according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0064] The present invention provides a method for identifying welding defects of a rotary drilling cylinder, as Figure 1 shown, including the following steps:

[0065] Step S10: During the welding process of the rotary drilling cylinder, various sensors are used to collect welding-related data in real time.

[0066] Among them, the sensors used in step S10 include: a welding current sensor, a welding voltage sensor, a thermal imaging sensor, and a weld tracking sensor.

[0067] Specifically, in step S10, during the welding process of the rotary drilling cylinder, various sensors are used to collect welding data in real time. The sensors used include:

[0068] Welding current sensor: Installed in the welding power supply circuit, it is used to accurately collect the real-time value of the welding current and generate a change curve based on the collected real-time value, reflecting the current fluctuation during the welding process and the trend of current change over time, and is used to judge whether there are welding defects caused by current problems;

[0069] Welding voltage sensor: Installed in the welding power supply circuit and connected to the welding electrode, it is used to collect the real-time data of the welding voltage, reflecting the change of the potential difference at both ends of the electrode during the welding process, and is used to judge whether there are welding defects caused by voltage problems;

[0070] Thermal imaging sensor: Including an infrared thermal imager, it is used to monitor the temperature distribution during the welding process in real time, obtain the thermal imaging diagram during the welding process, and display the temperature changes in the welding molten pool and the heat affected area in the form of an image, including the highest temperature, the lowest temperature and the temperature gradient information during the welding process, and is used to judge whether there are welding defects caused by heat conduction problems;

[0071] Weld seam tracking sensor: Installed within the range of the welding path, it is used to collect the position information data of the welding torch and the actual forming data of the weld seam, including the width, depth and shape deviation of the weld seam, and is used to judge whether there are defects such as undercut, incomplete fusion and poor weld seam formation in the weld seam.

[0072] Step S20: Preprocess the collected welding data according to the data type, including filtering, image enhancement and smoothing processing, and extract features respectively, including average value, maximum and minimum values, and standard deviation.

[0073] Specifically, in step S20, the collected welding data is preprocessed according to the data type, including:

[0074] Filtering and denoising: Perform filtering and denoising processing on the collected welding current and voltage data. Use a Butterworth filter and set the cut-off frequency according to the sampling frequency and noise characteristics of the data to remove the noise interference in the signal;

[0075] Image enhancement processing: Perform image enhancement processing on the thermal imaging diagram during the welding process, increase the temperature difference contrast in the thermal imaging diagram, and use the method of gray scale stretching to enhance the image;

[0076] Data smoothing processing: Use the moving average method to smooth the collected position information data to eliminate the small fluctuations caused by sensor jitter or measurement errors.

[0077] Among them, the steps of enhancing the image by using the method of gray scale stretching in the image enhancement processing include:

[0078] Read the thermal image: Read the thermal image collected by the thermal imaging sensor into a computer program. The thermal image is a grayscale image, and the grayscale value of each pixel represents the temperature information at that position;

[0079] Obtain the grayscale histogram of the heat image: Count the frequency of each grayscale level in the thermal image to obtain the grayscale histogram of the image; Analyze the histogram to understand the general situation of the temperature distribution in the image, such as the temperature range, the number of pixels in different temperature intervals, etc.;

[0080] Determine the stretching interval: After excluding the highest 3 grayscale values and the lowest 3 grayscale values in the grayscale histogram, determine the minimum valid grayscale value min_val and the maximum valid grayscale value max_val, and set the target grayscale interval after stretching to [new_min, new_max], where new_min is the minimum value after stretching and new_max is the maximum value after stretching;

[0081] Gray-scale stretching calculation: For each pixel point (x, y) in the thermal image, its original grayscale value is old_val, and its gray-scale stretching calculation formula is shown in Equation (1):

[0082]

[0083] where new_val is the grayscale value after stretching;

[0084] Update the image pixel value: Update the calculated stretched grayscale value new_val to the corresponding pixel point in the thermal image. After traversing all pixel points in the thermal image, the gray-scale stretching process of the image is completed.

[0085] Among them, the steps of using the moving average method to smooth the collected position information data in the data smoothing process include:

[0086] Determine the window size of the moving average: The window represents the number of data points participating in the average calculation, and set the window size to window_size;

[0087] Moving average calculation: Start calculating from the kth data point, and determine the position of k according to Equation (2), as shown in Equation (2):

[0088]

[0089] where window_size is the set window size, and calculate the moving average for each data point. The calculation formula is shown in Equation (3):

[0090]

[0091] Where y(n) is the moving average of the nth data point, i is the retrieval index ranging from n - k to n + k, and x(i) is the ith data point within the range of n - k to n + k; after traversing the collected position information data, the smoothing process of the data is completed.

[0092] Specifically, the feature extraction in step S20 includes:

[0093] Extract features from the welding current data, including the average current, current standard deviation, current peak value, and current frequency, etc. These parameters can reflect the stability of energy input and the characteristics of arc behavior during the welding process;

[0094] Extract features from the welding voltage data, including the average voltage, voltage change rate, and the number of voltage spikes, etc., for analyzing the stability of the welding arc and the rationality of welding process parameters;

[0095] Extract features from the thermal imaging map, including the highest temperature, lowest temperature, average temperature, and temperature gradient in the thermal imaging map, etc. An abnormal change in the temperature gradient indicates the existence of defects during the welding process. For example, cracks will cause abnormal local heat conduction and lead to a sudden change in the temperature gradient;

[0096] Extract features from the position information data of the welding torch and the actual forming data of the weld, including geometric features such as the average weld width, weld width standard deviation, average weld height, and weld edge straightness, etc. These features are directly related to the forming quality and defect types of the weld.

[0097] Step S30: Obtain normal welding data and welding data samples containing various defects during the rotary drilling barrel welding process, and use a deep learning algorithm to construct and train a rotary drilling barrel welding defect recognition model.

[0098] Among them, when using a deep learning algorithm to construct a rotary drilling barrel welding defect recognition model in step S30, the model is constructed with the TensorFlow or PyTorch deep learning framework.

[0099] Specifically, the steps of using a deep learning algorithm to construct and train a rotary drilling barrel welding defect recognition model in step S30 include:

[0100] Dataset construction: Obtain normal welding data and welding data samples containing various defects (such as porosity, slag inclusion, crack, incomplete penetration, lack of fusion, undercut, etc.) during the rotary drilling barrel welding process, perform the above data preprocessing and feature extraction operations on the sample data, and divide them into a training set and a validation set according to a ratio of 8:2;

[0101] Model construction: The rotary drilling bucket welding defect recognition model includes an input layer, a convolutional layer, an average pooling layer, and a fully connected layer; the input of the input layer is the training set divided in the dataset construction step, including welding current, welding voltage, thermal imaging map, weld tracking data, and corresponding features. The number of neuron nodes in the input layer is the same as the number of features of the input data; the convolutional layer is used to extract local features of the input data through convolutional kernels. Different sizes of convolutional kernels are used, including cross-distributed convolutional kernels of 3×3 and 5×5. The number of convolutional layers and convolutional kernels is set according to the data features, and ReLU is used as the activation function; the average pooling layer is set after each convolutional layer and uses average pooling operations to reduce the data dimension, reduce the computational amount while retaining the main features of the data; the fully connected layer is set in the last layer of the network layer to integrate the feature information extracted from the previous layers. The number of neuron nodes in the fully connected layer is the same as the number of neuron nodes in the input layer;

[0102] Model training and verification: After the model is constructed, set the model parameters. Use the Adadelta optimization algorithm, and set the initial learning rate to 0.01. Use the training set as the input to train the model. Set the initial training epoch to 50. Use the cross-entropy loss function as the loss function during the model training process. After each round of training ends, use the divided validation set to verify the trained model to check whether the model can accurately identify the rotary drilling bucket welding defects in the validation set. Stop the model training when the recognition accuracy of the validation set no longer increases in 5 consecutive rounds of verification, and complete the training and verification of the model.

[0103] Step S40: After the model training is completed, input the real-time collected data after data preprocessing and feature extraction into the model for defect recognition and output the recognition result.

[0104] Among them, the recognition results output in step S40 include the existence of defects and no defects. When the recognition result is the existence of defects, trigger an audible and visual alarm, display the identified defect type on the display screen, and mark the location of the defect and relevant welding data at the same time.

[0105] In addition, after the model training in step S40 is completed, input the real-time collected data after data preprocessing and feature extraction into the model for defect recognition and output the recognition result, record the defect recognition result and the real-time collected data. As the rotary drilling bucket welding defect recognition progresses, use the newly collected rotary drilling bucket welding data as new sample data to update the training set, retrain the rotary drilling bucket welding defect recognition model, and continuously update the model version to adapt to the welding defect feature changes that may be caused by factors such as minor changes in the rotary drilling bucket welding process, differences in raw materials, and equipment wear, ensuring that the model always maintains a high recognition accuracy.

[0106] In addition, the present invention also provides a system for identifying welding defects of a rotary drilling cylinder. Please refer to Figure 2 , the system for identifying welding defects of a rotary drilling cylinder includes:

[0107] Welding data acquisition module for rotary drilling cylinder: used to collect welding data in real time through sensors during the welding process of the rotary drilling cylinder;

[0108] Welding data preprocessing and feature extraction module for rotary drilling cylinder: used to perform data preprocessing on the collected welding data according to the data type, including filtering, image enhancement, and smoothing, and perform feature extraction respectively, including average value, maximum and minimum values, and standard deviation;

[0109] Model construction and training module for identifying welding defects of rotary drilling cylinder: used to construct and train a model for identifying welding defects of rotary drilling cylinder using a deep learning algorithm after data preprocessing and feature extraction are completed;

[0110] Welding defect identification module for rotary drilling cylinder: used to input the real-time collected data after data preprocessing and feature extraction into the model for defect identification and output the identification result after the model training is completed;

[0111] The sensors used in the welding data acquisition module for rotary drilling cylinder include: welding current sensor, welding voltage sensor, thermal imaging sensor, and weld tracking sensor;

[0112] When constructing a model for identifying welding defects of a rotary drilling cylinder using a deep learning algorithm in the model construction and training module for identifying welding defects of a rotary drilling cylinder, the model is constructed with the TensorFlow or PyTorch deep learning framework;

[0113] The identification results output by the welding defect identification module for rotary drilling cylinder include the existence of defects and no defects. When the identification result is the existence of defects, an audible and visual alarm is triggered, and the identified defect type is displayed on the display screen while marking the location of the defect and relevant welding data.

[0114] The system for identifying welding defects of a rotary drilling cylinder provided in this application adopts the method for identifying welding defects of a rotary drilling cylinder in the above embodiment, and can solve the technical problems of low efficiency and low accuracy of the traditional method for identifying welding defects of a rotary drilling cylinder. Compared with the prior art, the beneficial effects of the system for identifying welding defects of a rotary drilling cylinder provided in this application are the same as those of the method for identifying welding defects of a rotary drilling cylinder provided in the above embodiment, and other technical features in the system for identifying welding defects of a rotary drilling cylinder are the same as those disclosed in the above embodiment method, and will not be elaborated here.

[0115] The present application provides a device for identifying welding defects of a rotary drilling bucket. The device for identifying welding defects of a rotary drilling bucket includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute a method for identifying welding defects of a rotary drilling bucket in Embodiment 1 above.

[0116] Reference is made below Figure 3 , which shows a schematic structural diagram of a device for identifying welding defects of a rotary drilling bucket suitable for implementing the embodiments of the present application. A device for identifying welding defects of a rotary drilling bucket in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions: tablet computers), PMPs (Portable Media Players), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The device for identifying welding defects of a rotary drilling bucket shown is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present application.

[0117] Figure 3As shown, a rotary drilling bucket welding defect identification device may include a processing system 1001 (such as a central processing unit, a graphics processing unit, etc.), which may perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage system 1003 into a random access memory (RAM: Random Access Memory) 1004. In the RAM 1004, various programs and data required for the operation of a rotary drilling bucket welding defect identification device are also stored. The processing system 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems may be connected to the I / O interface 1006: an input system 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output system 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage system 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication system 1009. The communication system 1009 may allow a rotary drilling bucket welding defect identification device to communicate with other devices wirelessly or wiredly to exchange data. Although a rotary drilling bucket welding defect identification device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the shown systems. More or fewer systems may be implemented or had alternatively.

[0118] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes program codes for performing the methods shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from a network through the communication system, or installed from the storage system 1003, or installed from the ROM 1002. When the computer program is executed by the processing system 1001, the above-mentioned functions defined in the methods of the embodiments disclosed in the present application are executed.

[0119] A rotary drilling bucket welding defect identification device provided by the present application adopts a rotary drilling bucket welding defect identification method in the above-mentioned embodiment, and can solve the technical problems of low efficiency and low accuracy of traditional rotary drilling bucket welding defect identification methods. Compared with the prior art, the beneficial effects of a rotary drilling bucket welding defect identification device provided by the present application are the same as those of a rotary drilling bucket welding defect identification method provided by the above-mentioned embodiment, and other technical features in the rotary drilling bucket welding defect identification device are the same as the features disclosed in the method of the previous embodiment, and will not be elaborated here.

[0120] Each part disclosed in this application can be implemented by hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in a suitable manner in any one or more embodiments or examples.

[0121] This application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the steps of a method for identifying welding defects of a rotary drilling bucket as described above.

[0122] The computer program product provided by this application can solve the technical problems of low efficiency and low accuracy of the traditional method for identifying welding defects of a rotary drilling bucket. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as those of a method for identifying welding defects of a rotary drilling bucket provided by the above embodiments, and will not be elaborated herein.

[0123] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A method for identifying welding defects of rotary drilling barrels, characterized in that: The method comprises the following steps: Step S10: During the rotary drilling barrel welding process, the welding data is collected in real time through the sensor; Step S20: performing data preprocessing on the collected welding data according to the data type, including filtering, image enhancement and smoothing, and performing feature extraction, including average value, maximum value and standard deviation; Step S30: obtaining normal welding data and welding data samples containing various defects during the rotary drilling barrel welding process, and constructing and training a rotary drilling barrel welding defect recognition model using a deep learning algorithm; Step S40: After the model training is completed, the real-time collected data after data preprocessing and feature extraction is input into the model for defect recognition and output recognition results; The sensors used in step S10 include: a welding current sensor, a welding voltage sensor, a thermal imaging sensor and a weld tracking sensor; When the deep learning algorithm is used to construct the rotary drilling barrel welding defect recognition model in step S30, the model is constructed using the TensorFlow or PyTorch deep learning framework; The recognition result output in step S40 includes defects and no defects. When the recognition result is defects, an audible and visual alarm is triggered, and the identified defect type is displayed on the display screen, while the location of the defect and related welding data are marked.

2. A method for identifying welding defects of rotary drilling barrels according to claim 1, characterized in that: In step S10, during the rotary drilling barrel welding process, the welding data is collected in real time by sensors, and the sensors used include: Welding current sensor: installed in the welding power supply circuit, used to collect the real-time value of the welding current and generate a change curve based on the collected real-time value, reflecting the fluctuation of the current during the welding process and the trend of the current change over time, and used to determine whether there are welding defects caused by current problems; Welding voltage sensor: installed in the welding power supply circuit and connected to the welding electrode, used to collect real-time data of welding voltage, reflect the change of potential difference between the two ends of the electrode during welding, and used to determine whether there are welding defects caused by voltage problems; Thermal imaging sensor: including infrared thermal imager, which is used to monitor the temperature distribution in the welding process in real time, obtain the thermal image of the welding process, and display the temperature changes of the welding pool and the heat-affected zone in the form of images, including the highest temperature, the lowest temperature and the temperature gradient information in the welding process, which is used to determine whether there are welding defects caused by heat conduction problems; Weld seam tracking sensor: Installed within the welding path, it is used to collect the position information data of the welding gun head and the actual formation data of the weld, including the width, depth and shape deviation of the weld, and is used to determine whether the weld has defects such as undercut, incomplete welding and poor weld formation.

3. A method for identifying welding defects of rotary drilling barrels according to claim 1, characterized in that: In step S20, the collected welding data is preprocessed according to the data type, including: Filtering and denoising: The collected welding current and voltage data are filtered and denoised. The Butterworth filter is used to set the cutoff frequency according to the sampling frequency and noise characteristics of the data to remove noise interference in the signal. Image enhancement processing: Perform image enhancement processing on the collected thermal images of the welding process to increase the temperature difference contrast in the thermal images and use the grayscale stretching method to enhance the image; Data smoothing: The collected location information data is smoothed using the moving average method.

4. A method for identifying welding defects of rotary drilling barrels according to claim 3, characterized in that: The step of enhancing the image by using the grayscale stretching method in the image enhancement process comprises: Reading thermal images: Read the thermal images collected by the thermal imaging sensor into the computer program. The thermal image is a grayscale image, and the grayscale value of each pixel represents the temperature information at that location. Get the grayscale histogram of the thermal image: Count the frequency of each grayscale level in the thermal image to get the grayscale histogram of the image; Determine the stretching interval: After excluding the highest 3 grayscale values ​​and the lowest 3 grayscale values ​​in the grayscale histogram, determine the minimum valid grayscale value min_val and the maximum valid grayscale value max_val, and set the target grayscale interval after stretching to [new_min, new_max], where new_min is the minimum value after stretching and new_max is the maximum value after stretching; Grayscale stretching calculation: For each pixel (x, y) in the thermal image, its original grayscale value is old_val, and its grayscale stretching calculation formula is shown in formula (1): Among them, new_val is the gray value after stretching; Update image pixel value: Update the calculated stretched grayscale value new_val to the corresponding pixel point in the thermal image, and complete the grayscale stretching processing of the image after traversing all the pixels in the thermal image.

5. A method for identifying rotary drilling barrel welding defects according to claim 3, characterized in that: The step of smoothing the collected position information data by using a moving average method in the data smoothing process includes: Determine the window size of the moving average: the window represents the number of data points involved in the average calculation, and the window size is set to window_size; Moving average calculation: Start the calculation from the kth data point and determine the position of k according to formula (2), as shown in formula (2): Wherein, window_size is the set window size, and the moving average is calculated for each data point. The calculation formula is shown in formula (3): Where y(n) is the moving average of the nth data point, i is the search index ranging from nk to n+k, and x(i) is the i-th data point in the range of nk to n+k; after traversing the collected location information data, the data is smoothed.

6. A method for identifying welding defects of rotary drilling barrels according to claim 1, characterized in that: The feature extraction in step S20 includes: Extract features from welding current data, including current mean, current standard deviation, current peak, and current frequency; Extract features from welding voltage data, including voltage average value, voltage change rate and number of voltage spikes, to analyze the stability of welding arc and the rationality of welding process parameters; Extract features from the thermal image, including the maximum temperature, minimum temperature, average temperature and temperature gradient in the thermal image; Features are extracted from the position information data of the welding gun head and the actual forming data of the weld, including the average weld width, the standard deviation of the weld width, the average weld height and the straightness of the weld edge.

7. A method for identifying welding defects of rotary drilling barrels according to claim 1, characterized in that: The steps of obtaining normal welding data and welding data samples containing various defects during the rotary drilling barrel welding process in step S30, and constructing and training a rotary drilling barrel welding defect recognition model using a deep learning algorithm include: Dataset construction: Perform the above data preprocessing and feature extraction operations on the acquired sample data, and divide it into a training set and a validation set in a ratio of 8:2; Model construction: The rotary drilling barrel welding defect recognition model includes an input layer, a convolution layer, an average pooling layer and a fully connected layer; the input of the input layer is the training set divided in the data set construction step, including welding current, welding voltage, thermal imaging image, weld tracking data and corresponding features, and the number of neuron nodes in the input layer is the same as the number of features of the input data; the convolution layer is used to extract local features of the input data through convolution kernels, and the convolution kernels are of different sizes, including 3×3 and 5×5 cross-distributed convolution kernels. The number of convolution layers and convolution kernels is set according to the data characteristics, and ReLU is used as the activation function; the average pooling layer is set after each convolution layer, and the average pooling operation is adopted; the fully connected layer is set at the last layer of the network layer, which is used to integrate the feature information extracted by the previous layers, and the number of neuron nodes in the fully connected layer is the same as the number of neuron nodes in the input layer; Model training and verification: After the model is built, the model parameters are set, the Adadelta optimization algorithm is used, the initial learning rate is set to 0.01, the training set is used as input to train the model, the initial training cycle is set to 50, and the cross entropy loss function is used as the loss function in the model training process. After each round of training, the divided verification set is used to verify the trained model to verify whether the model accurately identifies the rotary drilling barrel welding defects in the verification set. When the recognition accuracy of the verification set no longer increases in 5 consecutive rounds of verification, the model training is stopped to complete the model training and verification.

8. A rotary drilling barrel welding defect identification system, characterized in that: The rotary drilling barrel welding defect identification system comprises: Rotary drilling barrel welding data acquisition module: used to collect welding data in real time through sensors during the rotary drilling barrel welding process; The rotary drilling barrel welding data preprocessing and feature extraction module is used to preprocess the collected welding data according to the data type, including filtering, image enhancement and smoothing, and extract features, including the average value, maximum value and standard deviation. Drill barrel welding defect recognition model construction and training module: used to build and train the drill barrel welding defect recognition model using deep learning algorithm after data preprocessing and feature extraction; Rotary drilling barrel welding defect recognition module: After the model training is completed, the real-time collected data after data preprocessing and feature extraction is input into the model for defect recognition and output recognition results; The sensors used in the rotary drilling barrel welding data acquisition module include: welding current sensor, welding voltage sensor, thermal imaging sensor and weld tracking sensor; When a deep learning algorithm is used to construct a rotary drilling barrel welding defect recognition model in the rotary drilling barrel welding defect recognition model construction and training module, the model is constructed using a TensorFlow or PyTorch deep learning framework; The recognition results outputted from the rotary drilling barrel welding defect recognition module include the presence of defects and the absence of defects. When the recognition result is the presence of defects, an audible and visual alarm is triggered, and the identified defect type is displayed on the display screen, while the location of the defect and related welding data are marked.

9. A rotary drilling barrel welding defect identification device, characterized in that: The rotary drilling barrel welding defect identification device comprises: A memory, a processor, and a rotary drill barrel welding defect identification program stored in the memory and executable on the processor, wherein the rotary drill barrel welding defect identification program, when executed by the processor, implements a rotary drill barrel welding defect identification method as described in any one of claims 1 to 7.

10. A computer program product, characterized in that The computer program product includes a rotary drilling drill barrel welding defect recognition program, and when the rotary drilling drill barrel welding defect recognition program is executed by a processor, a rotary drilling drill barrel welding defect recognition method as described in any one of claims 1 to 7 is implemented.

Images