KR20030016029A - Welding flaw detecting method of Spiral Welding Pipe - Google Patents

Abstract

PURPOSE: A welding defect detecting method for a spiral welding pipe is provided to realize the objective detection of welding defects fully and safely by forming an artificial intellectual detecting system. CONSTITUTION: In a welding defect detecting method for a spiral welding pipe, welding defects are detected by an ultrasonic test while controlling the position of an ultrasonic wave sensor by a CCD camera, kinds of the welding defects are recognized by a backward propagation neural network.

Description

Welding flaw detecting method of spiral welding pipe

◆ Technical Field

Defects present in the welds of spiral welded pipes always include problems of shortening the life of the product or deteriorating the safety of the welded product.

There are many non-destructive testing methods for detecting weld defects, but it is a method for artificial intelligence for the automation of inspection and the recognition of the detected weld defects using ultrasonic flaw detection.

◆ Prior art in the field

Prior art

Non-destructive testing is mainly used for the detection of welding defects. In order to detect welding defects in real time, various techniques such as radiation, eddy current, and ultrasound have been attempted.

Problems of the Prior Art

① The method of directly measuring the shape of the molten part using radiation should receive a signal from the opposite side from which the source is radiated. Therefore, it should be accessible from both sides of the welding member during real-time detection, and there is a restriction to synchronize the source and signal side. Follow.

② In the method of using eddy current, the probe should be close to the hot welding part, and it is difficult to make accurate measurement due to the influence of welding current.

③ The method of using ultrasonic wave is a possibility because only one side of the welding part needs to be accessible and is less affected by heat, light and electromagnetic field generated during welding. However, in the case of using ultrasonic waves, there are many limitations due to the difficulty of keeping the ultrasonic sensor closely adhered to the surface of the base material by the supply of the flaw detection solution and the characteristic change of the ultrasonic waves with temperature.

As another method, the research of ultrasonic generation using laser and non-contact reception using EMAT is continuously conducted, but this method is also affected by surface roughness, and the method using EMAT has low sensitivity, low signal-to-noise ratio, There is a problem to be very close to.

The non-destructive testing technique used to detect weld defects in spiral welded pipes has been used to detect radiation. However, the radiation detection method is not only difficult to detect in real time, but also has difficulty in inspection due to radiation exposure and film development time.

Ultrasonic inspection instead of radiation is used for real-time detection and full inspection of welding defects inherent in spiral welded pipes, and the position control of ultrasonic sensors for detection and recognition of weld defects by following welding lines of rotating spiral welded pipes. And the development of an intelligent defect assessment program through the analysis of the ultrasonic signal should be parallel.

1 is a block diagram of a weld defect inspection algorithm of a spiral welded pipe.

Back propagation is used to recognize the type of defect detected by the weld of the spiral weld pipe and to control the position of the ultrasonic sensor. As a preprocessing for the back propagation neural network, signal processing is used for defect type recognition and image processing is used for position control of an ultrasonic sensor.

2 is a schematic diagram for the automation of inspection. 355.6-406.4mm in diameter, 6-24mm thick, 1000mm working section, max.3.8m / min working speed, pipe weight is designed based on 230kg max.

3 is an overall configuration diagram for the automation of the spiral weld pipe and welding defect recognition.

The position control of the ultrasonic sensor is imaged by CCD camera and DSP board, and the image processing program is used to extract feature variables and feed back to back propagation neural network.

Recognition of welding defects receives RF waves from the ultrasonic flaw detector with a PC-based board, and the received signal extracts the characteristic parameters of the welding defects through a signal processing program.

-For position control of ultrasonic sensor, CCD camera is used as visual sensor and DSP board is used for image data processing to analyze shape and direction of welding line. As an analysis method, the patterning processing and the pattern recognition technique of the characteristic image which can represent the acquired image signal are applied.

The pattern recognition technique uses a backpropagation neural network with excellent performance for pattern classification.

-The analysis of the ultrasonic signal removes the noise signal through the signal processing of the defect signal obtained from the ultrasonic flaw detector and extracts the feature signal that can represent the defect signal from the preprocessed ultrasonic signal. The extracted feature signal uses the backpropagation neural network used in the position control of the ultrasonic sensor.

-It can be configured as a system by using the back propagation neural network together in the position control of the ultrasonic sensor and the analysis of the ultrasonic signal.

Compared with the conventional radiation detection method, it solves the problem of radiation exposure, which is a disadvantage of radiation, and the entire inspection due to film development time, and it is possible to detect the welding defect in the ultrasonic detection method. By constructing, objective detection is possible. In addition, by in-line the inspection process to the production process of the product, it is possible to expect a full inspection of the product and the economic derivative effect.

Claims (4)

A method for detecting and recognizing a weld defect present on a spiral weld pipe, wherein the weld defect detection is performed by ultrasonic flaw detection (UT) and the position control of the ultrasonic sensor is controlled by a CCD camera. The method of claim 1, wherein the back propagation neural network is used to recognize the weld defect type of the srilar welded pipe. The method of claim 1, wherein the back propagation neural network is used to control the position of the ultrasonic sensor. Method 2 and 3, using a PC-based oscilloscope and DSP board to simultaneously detect welding defects and recognize types of srilar welded pipes and position control of ultrasonic sensors with a single PC.