3D measuring and segmentation method for hot heavy forging

Highlights

• We describe a method for measuring the dimensions of hot heavy forgings.

• A sequence of 2D laser scans generates a point cloud dataset.

• The item of interest is isolated from the background via cluster analysis among other techniques.

• The processed dataset can yield accurate three dimensional measurements.

• This process has been experimentally verified in the laboratory and the workshop.

Abstract

Non-contact, three-dimensional (3D) measurements of hot heavy forgings (HHFs) are desirable for permitting real-time process control, whereas most current methods are usually inconvenient or invalid due to the difficulty in dealing with the hot heavy forgings. This paper presents a new measuring and combined segmentation approach that employs a two-dimensional (2D) laser radar with additional rotation driven by the servomotor that scans the forgings to acquire a massive 3D point cloud dataset. From this dataset, the desired forging part is roughly distinguished from the background based on the angle and distance continuity constraints, and then refined by the curvature-based border extraction method and further segmented by the hierarchical clustering analysis method. Finally, the feature points are extracted based on the normal vector variation and fitted to convey the 3D information. This novel method has been verified by experiments both in the laboratory and the forging workshop for hot heavy forging pieces, with a dimension error of less than 2%. These results indicate that the proposed approach is more practical and convenient than current methods for real-time, on-site measurements of HHFs.

Introduction

3D measurements are of great importance for the manufacturing of crucial parts, such as crankshaft and turbine rotor, in heavy industrial fields, and are desirable in a number of hot-forming processes, such as hot forging processes. Accurate 3D measurements can significantly improve productivity and quality, reduce scrap and waste, and even permit real-time process control. However manual contact measurement equipment (such as calipers) is still often used to measure hot forgings. Thus, the inaccurate measuring results and extreme working environment (forging temperature 900–1200 °C, and strong thermal radiation) limit the further application of this method. Moreover, the direct exposure to the strong radiation of hot forgings may physically injure the operation workers [10]. The characteristics of HHFs, therefore, dictate that developing a non-contact, wide-range measurement method is of the utmost importance.

At present, the hanging-wire method is used to measure the diameter of manufactured forgings in the Japanese Steel factory [15]; alternatively, the method using two laser beams, replacing the hanging wire, is used to measure diameters in Korea [16]. At Doosan Heavy Industries, the length of a forging is measured by two lasers that load on the same rail projecting beam, and Biskup et al. [1] have demonstrated the application of terrestrial laser beam assisted method for shipbuilding. Nevertheless, only simple parameters such as diameter or length can be measured by these methods. Therefore, more powerful, accurate, and non-contact methods for the 3D measurement of high-temperature heavy forgings have attracted significant attention. Such methods include the vision method, the laser-scanning method, and the laser-ranger method.

Many researchers have applied visual techniques to 3D measurement for HHFs, and this has yielded some impressive results so far. Okamoto et al. [12] reported on a measurement system for heavy forgings based on double charge-coupled device (CCD) cameras in Japan. Dworkin and Nye [4] compared the merits and demerits of monochrome cameras, color cameras and infrared cameras through experiments, eventually recommending the use of traditional CCD cameras with infrared filters to measure forging dimensions. However, the field depth of CCD cameras is relatively shallow, which limits their application in the measurement for large heavy forgings. Moreover, only 2D information can be obtained from these measurements, and the harsh environment in forging workshop makes it difficult to calibrate the camera properly.

OG Technologies [11] developed a coordinate measurement machine named HotEye, which utilizes a laser line to scan in concert with a camera to capture the images of the lines with 3D accuracies as low as 0.1 mm. However, the size of the hot forging object is quite limited due to the small vision angle, Typically maximum dimension less than 200 mm.

One of the main difficulties in measuring the hot forgings is their high temperature, under which they radiate considerable amounts of infrared (IR) energy. This radiation causes the CCD camera to lose focus, thereby reducing the quality of the final images [4]. Therefore it is difficult to extract the feature points from the images for which the 3D profile of the hot forging is unclear due to serious radiation. Aside from this temperature-related issue, other common problems involve the difficulty of sub-pixel extraction of the featured points, time-consuming calculation of the corresponding points matching, and complex environmental influence during on-site measurement.

In response to these issues, Jia et al. [7] introduced a spectrum-selective method for improving the image quality of hot parts by filtering light of certain wavelength ranges and projecting light stripes onto hot forging surfaces. Subsequently, Liu et al. [8] proposed an improved online stereo vision method using two cameras with special low-pass filters, along with and a high-brightness digital light projector to project light stripes onto the hot forging’s surface. This method overcame the problem of the field depth of a single camera and the difficulty of matching of double-eye machine visions. However, it is only suitable for diameter measurements for cylindrical forgings and unable to perform 3D profile measurements for heavy forgings of more complex shapes.

Laser scanning is another method used for non-contact measurements of hot forgings. Compared to other methods, the primary advantages of the laser method are: (1) wide measuring range; (2) strong resistance to disturbances; and (3) the ability to measure absolute coordinates. Because laser is not generally affected by the harsh environments (e.g., high temperatures and strong vibrations), laser radar has been widely used for non-contact measurement in various industries. The laser scanning method has been employed in a measuring system to detect the length of hot forgings mentioned by Okamoto et al. [12]. Määtta et al. [9] verified the feasibility of the laser method by measuring objects that were 1400 °C and 20 m away from the radar. The LaCam Forge measurement system, developed by Rech et al. [13], also adopted laser scanning to measure the flatness of forging surfaces. Unfortunately, other dimensions, such as the diameter or 3D shape, do not fall within the grasp of this method.

In response to this particular shortcoming, two commercially available time-of-flight (TOF) lasers were used by Bokhabrine et al. [2] and the acquisition of three 3D point clouds with an angular shift of about 120° between each acquisition was required. Points pertaining to the metallic shell were extracted and 3D measurements were taken on the reconstructed shell. For this technique, it was necessary that the measured objects were circular, otherwise the algorithm of extracting points would fail. Tian et al. [14] also developed a laser measurement system based on TOF for HHFs. However, the scanning device of spherical parallel mechanism is too complex to apply in the forging workshop. Subsequently, He et al. [6] improved the 3D measurement instrument with a novel 4 DOF (Degree of Freedom) robot system instead of the two DOF spherical parallel mechanism. Also, Fu et al. [5] further developed a two-laser-scanner measuring system to obtain the outer and inner diameters of forgings with the assistance of IR temperature measuring technology. In this process, the size of the outer diameter is measured according to the laser-scanner method, while the actual size of the inner diameter is calculated by establishing the relationship between temperature and size, hence the accuracy of this hybrid method cannot be guaranteed due to the indirect measurement of inner diameter.

In our previous work [3], we presented a new 3D measuring approach based on a 2D laser range sensor and a additional servo-motor, which enables the sensor to scan forgings across different scan planes. In this way, massive amounts of 3D data that contain information of the forging, the relative supporting parts, and the distant background in the workshop are obtained in a short time. Moreover, actual 3D models of HHFs can be reconstructed by using a triangulated irregular network. Different parameters of forgings, such as lengths and diameters, can be measured based on the subsequent 3D model. However, extracting the actual forging’s profile from the complex on-site background is still one of the main difficulties resulting from the previous work, and a method for doing so has not been reported in the literature yet.

Due to the large size of the datasets yielded by this approach, the post-processing of extraction directly determines whether the dimension measurement of the 3D forging profile can be realized and meet the practical requirement of industry. Hence, research on the methodology and efficiency of the segmentation of massive datasets seems to be extremely relevant at this time. Current extraction segmentation methods for massive point cloud data measurements can be divided into three categories: the curvature-based method; the normal-based method; and other mathematical methods, such as the geometrical-theory-based method, the fuzzy-mathematics-based method, and the wavelet-based extraction method. These methods enable us to extract edge-neighborhood points by considering the geometric shape of a part. However, despite the advances made in the above literature, none of it is applicable to the on-site measurement of heavy forgings under high temperature.

Based on our prior laser ranger sensor measuring system [3], a new 3D point cloud data segmentation approach for 3D profile measurement of heavy forging under high temperature has been proposed. According to this method, first the distribution characteristics of massive laser scanning cloud points in horizontal and vertical layers, which aid the further segmentation of the heavy forging cloud point, are analyzed. Then, a combined segmentation method is proposed in detail, which mainly includes 3 steps. Furthermore, the final forging profile is achieved through this method. The accuracy of this profile has been experimentally verified both in the laboratory and in the forging workshop. The results of these experiments was compared and discussed.

The remainder of this paper is organized as follows. First, the principles of the laser measurement method and point cloud data characteristics are introduced in Section 2. In Section 3, the preliminary rough segmentation method based on angle and distance constraints is introduced. Next, the precise methods of curvature-based border extraction and hierarchical clustering analysis on border points are described in detail. Afterward, the model reconstruction and process of feature-points extraction is briefly introduced in Section 4. Experimental verification of this comprehensive method is presented and discussed in Section 5. Lastly, a brief conclusion is provided in Section 6.