CN-122007743-A - Welding workstation and automatic welding method for ship BK (BK) piece

CN122007743ACN 122007743 ACN122007743 ACN 122007743ACN-122007743-A

Abstract

The invention relates to a welding workstation and an automatic welding method for ship BK pieces, and belongs to the technical field of welding equipment. The method comprises the steps of identifying and extracting the outline of an uncertain incoming workpiece for welding a BK (stock-work) piece by a workpiece outline identification method, automatically matching a main board and a rib plate in one BK piece by a main rib plate matching method, performing character positioning by using deep learning OCR, performing feature extraction by a convolutional neural network to obtain character information of the main board and the rib plate, completing matching of the main rib plate, accurately positioning the geometric center of the workpiece by an accurate position grabbing method, realizing accurate position grabbing by matching with a robot programming system, and performing feature extraction and matching by the accurate assembly method through the deep convolutional neural network to complete assembly line positioning, so that accurate assembly is realized. According to the invention, the raw material workpiece for assembling the BK piece is judged and identified through the 3D camera, so that the identification stability is improved, the method is suitable for diversified site construction environments, the flow stability is ensured, and the hidden danger of pollution to the operation environment is reduced.

Inventors

WANG YONG
MA JIAXING
Niu Shilin

Assignees

唐山开元自动焊接装备有限公司

Dates

Publication Date: 20260512
Application Date: 20260331

Claims (7)

1. The automatic welding method for the ship BK piece is characterized by comprising a workpiece outline identification method, a main reinforcement matching method, a precise position grabbing method and a precise assembling method; the workpiece outline identification method is used for identifying and extracting the outline of the uncertain incoming workpiece for welding the BK piece; The main gusset matching method automatically matches the main gusset and the gusset in one BK piece, performs character positioning by using deep learning OCR based on a character recognition principle, and performs feature extraction by using a convolutional neural network to obtain character information of the main gusset and the gusset, so as to complete the matching of the main gusset; the accurate position grabbing method avoids collision in the falling and carrying processes when grabbing the workpiece, accurately positions the geometric center of the workpiece based on the geometric center extraction principle, and is matched with a robot programming system to realize accurate position grabbing; The accurate assembly method aims at the main board and the rib plates of the BK piece to be assembled before welding, the rib plates are placed on the assembly line of the main board, the characteristic extraction and the characteristic matching are carried out through the deep convolutional neural network based on the assembly line positioning principle, the positioning of the assembly line is completed, and then the accurate assembly is realized.
2. The automatic welding method for the ship BK piece according to claim 1, wherein the accurate position grabbing method comprises image acquisition and preprocessing, dynamic threshold segmentation, 8-neighborhood connected domain marking, stroke encoding and boundary tracking and geometric feature calculation; The image acquisition and preprocessing are carried out by using an identification camera to acquire a depth image or a gray image of a workpiece, wherein an original image is marked as I (x, y), wherein (x, y) is a pixel coordinate, I (x, y) is a gray value or a depth value, and smoothing filtering is carried out on the image to inhibit noise: I’(x,y)=Gaussian(I(x,y),σ) wherein sigma is the gaussian standard deviation; The dynamic threshold segmentation adopts a dynamic threshold method to separate a workpiece area from a background, adapts to illumination change and background non-uniformity, and is divided into a global dynamic threshold and a local self-adaptive threshold; The global dynamic threshold value is that the maximum inter-class variance determines an optimal threshold value T: T=argmax[ω 0 (t)ω 1 (t)(μ 0 (t)-μ 1 (t)) 2 ] Wherein omega 0 ,ω 1 is the pixel duty ratio of the background and the target, and mu 0 ,μ 1 is the corresponding gray average value; The local adaptive threshold value is determined by adding an offset C to each pixel (x, y) with reference to the gray scale statistic in the neighborhood window W: T(x,y)=mean(I’(x’,y’)|(x’,y’)∈W)+C according to the selected method, a binary image B (x, y) is obtained: B (x, y) =1, I' (x, y) > T (x, y), foreground B (x, y) =0, otherwise, background.
3. The automatic welding method for the ship BK piece according to claim 2, wherein the 8-neighborhood connected domain marking is based on an 8-neighborhood connected criterion, foreground pixels in a binary image are marked, different workpieces are distinguished, and a two-pass scanning algorithm is adopted: Scanning for the first time, namely scanning line by line, checking marks of the left, upper right and upper right neighborhood of each foreground pixel, if no marks exist, giving a new label, otherwise, taking the minimum label and recording an equivalence relation; And in the second scanning pass, merging the equivalent labels, and endowing each connected domain with a unique identifier L epsilon {1,2,.. N }, so as to obtain a marked image B L (x,y),B L (x, y) =0 to represent the background.
4. The method for automatically welding BK pieces of a ship according to claim 2, wherein the run-length encoding and the boundary tracking are performed by performing run-length encoding on the marked image, representing each row of continuous foreground pixels as a tuple of (row number, start column, end column), compressing data and facilitating boundary extraction, and extracting an outer contour by using Moore boundary tracking algorithm for each connected domain k: The first foreground pixel of the region is found in raster order as the starting point P 0 , starting from P 0 , searching 8 neighbors clockwise, finding the next boundary point, repeating until the starting point is returned, and obtaining the ordered contour point set C k ={P 0 ,P 1 ,…,P m-1 .
5. The method for automatically welding BK pieces of a ship according to claim 2, wherein the geometric feature calculation obtains the geometric feature of each workpiece according to a contour point set or a regional pixel set for identification, positioning or measurement.
6. The automatic welding method for the ship BK piece according to claim 1 or 2, wherein the precise position grabbing method grabs the workpiece by a carrying robot, and the geometric center of the irregularly-shaped workpiece is determined by the following steps: geometric center [ ] , ) The calculation formula: , ; Where A k is the area pixel sum and R k is the set of all pixel coordinates (x, y) belonging to the kth workpiece.
7. A welding workstation for ship BK pieces is characterized by comprising a carrying robot system, a welding robot system, a visual identification system, a working room, a dust removing system, a logistics system and an auxiliary system, wherein the auxiliary system at least comprises a centralized control console and a control cabinet, the carrying robot system, the welding robot system, the auxiliary system and the visual identification system are arranged in the working room, the working room is further provided with the dust removing system and the logistics system, the carrying robot system is used for grabbing and carrying workpieces for welding BK pieces, the welding robot system is used for welding the workpieces to assemble the BK pieces, the visual identification system is used for completing workpiece grabbing and carrying identification of the carrying robot system and is carried out through the centralized control console of the auxiliary system, the dust removing system is used for absorbing smoke dust and particles generated in the welding process, and the logistics system is used for logistics of raw material workpieces and finished BK pieces.

Description

Welding workstation and automatic welding method for ship BK (BK) piece Technical Field The invention relates to a welding workstation and an automatic welding method for ship BK pieces, and belongs to the technical field of welding equipment. Background In the hull structure in the field of ship manufacturing, a BK member (fully called Bracket, a toggle plate, a shipyard is often called as BK for short) is a key and widely applied combined toggle plate, the BK member is usually formed by welding a plurality of steel plates, the main plate and the rib plates comprise main plates and rib plates, the main plates and the rib plates formed by assembling and welding the BK member can be in regular shapes or irregular shapes, the BK member has an integrated structure and is mainly used for key parts with larger stress and higher structural integrity requirements in the hull so as to enhance local strength and overall stability. However, the welding process of current BK parts still faces a series of challenges, with the reliance on significant human involvement being particularly prominent. Firstly, the BK piece is complex in structure, the shapes and the sizes of the main plate and the rib plates are different, the positions of welding seams are changeable, the automatic equipment is difficult to accurately identify and weld, the automatic equipment is mainly finished by relying on personal experience and judgment of welders, secondly, the welding process involves continuous and tight matching of multiple working procedures, and a great deal of skilled technical workers are required to carry out intensive manual operation and monitoring from workpiece identification, positioning and assembly to welding. The method is highly dependent on a manual mode, not only causes low production efficiency and high manufacturing cost, but also causes that the welding quality and consistency are easy to be fluctuated by factors such as personnel skill level, fatigue degree and the like, and is difficult to realize stability and controllability, thereby forming potential influence on the long-term safety and reliability of a key structure of the ship body. Through searching, the Chinese patent application with publication number CN119734023A discloses a single-plate single-rib small-assembly unmanned welding device and an unmanned welding method, the device comprises a feeding robot, a welding robot, a visual identification device, a welding workbench, a feeding tray and a control system, and a two-dimensional code is arranged on the workpiece to be welded, and the single-plate single-rib small-assembly unmanned assembly and welding are realized through a mode of combining visual identification with mechanical loading. However, the device needs to scan the two-dimensional code on the workpiece to guide the operation, but in actual production, most enterprises adopt a mode of printing characters instead of the two-dimensional code, the two-dimensional code is easy to scratch and dirty when the workpiece is stacked and transported, so that the recognition failure is caused, the stability of the flow is affected, a large amount of smoke dust is generated in the welding process, a corresponding dust removing device is not configured in the system, the hidden danger of operation environment pollution exists, in addition, a visual system is easy to be interfered by an external light source, the recognition stability is poor, and the device is difficult to adapt to different areas and diversified site construction environments, so that the actual technological requirements of unmanned welding of the current BK piece are not completely met. Disclosure of Invention The invention provides a welding workstation and an automatic welding method for ship BK pieces, which are used for judging and identifying raw material workpieces for assembling the BK pieces through a 3D camera, improving the identification stability, adapting to different areas and diversified site construction environments, ensuring the flow stability, treating a large amount of smoke dust generated in the welding process, reducing the hidden danger of pollution of the operation environment, completely meeting the actual technological requirements of unmanned welding of the current BK pieces and solving the technical problems existing in the prior art. The technical scheme of the invention is as follows: An automatic welding method for a ship BK piece comprises a workpiece outer contour recognition method, a main reinforcement matching method, a precise position grabbing method and a precise assembly method; the workpiece outline identification method is used for identifying and extracting the outline of the uncertain incoming workpiece for welding the BK piece; The main gusset matching method automatically matches the main gusset and the gusset in one BK piece, performs character positioning by using deep learning OCR based on a character recognition principle, and perfo