Journal of Constructional Steel Research 57 (2001) 647659 www.elsevier.

com/locate/jcsr

Displacement measurement in studies of steel T-stub connections

S. Spyrou *, J.B. Davison
Department of Civil and Structural Engineering, The University of Shefeld, Shefeld S1 3JD, UK Received 10 May 2000; received in revised form 1 December 2000; accepted 23 January 2001

Abstract Traditionally steel and composite steel and concrete beam to column joints at the re limit state are assumed to behave in the same way as used in the design at ambient temperature, that is a joint that is assumed to be pinned at ambient temperature is also assumed to be pinned at elevated temperatures. Recent re tests on the eight-storey steel frame building at Cardington suggest that joints, which are assumed to be pinned at ambient temperature, can provide signicant levels of both strength and stiffness at elevated temperature. In steel structures, the behaviour of joints between beams and columns may be investigated by separate consideration of the various components, which comprise the joint. However, measurement of the distorted proles of components can be problematic. This paper describes the use of an image acquisition and processing technique for the measurement of deformations in steel T-stub connections, representing the tension component in a steel joint. Although the application has been developed to solve the problems associated with measuring deformations in tests conducted at high temperatures, it also offers advantages over conventional techniques at ambient temperatures. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Acquisition of images; Image processing; Components; T-stubs; Joint; Steel structures

1. Introduction The response of steel-framed structures to applied loading depends to a large degree on the behaviour of the type of connections between the columns and beams. Traditionally designers have assumed that these connections act either as pins (with

* Corresponding author. E-mail address: s.spyrou@shefeld.ac.uk (S. Spyrou).

0143-974X/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 1 4 3 - 9 7 4 X ( 0 1 ) 0 0 0 0 3 - 7

648

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

no ability to transmit moments from beam to column) or as rigid (providing perfect continuity between the connected members). Advances in analysis and developments in modern codes of practice permit designers to account for the real behaviour of steel joints where this is known. Even though experimental studies of joints conducted at many research centres around the world have provided a large bank of test data [1], the vast number of variables in connections (beam and column size, plate thickness, bolt sizes and spacing etc) often means that data for a specic connection arrangement do not exist. As a result researchers have turned their attention to predicting the behaviour of such joints. One approach, which has gained acceptance, is based on a component method in which overall joint behaviour is assumed to be the sum of the responses of various simpler components. Much work has been completed in developing this approach for beam to column connections [1]. When steel-framed structures are subjected to re, the ability to sustain load is severely impaired and the action of the joints is of particular concern. To date, data on the response of joints at high temperatures has been gathered from full-scale furnace tests [2]. The application of the component method at elevated temperatures is the subject of a recently awarded research contract [3] which aims to develop a method of predicting the behaviour of steel beam to column connections in re conditions. In the tension region of a joint, the column ange in bending, the end plate in bending and the bolts in tension are idealised as T-stub assemblies as shown in Fig. 1. T-stubs play a fundamental role in the behaviour of the joint and are used to compute the stiffness contribution of the most important joint components. For that reason, an experimental analysis on bolted T-stub assemblies has been undertaken in order to measure the forcedisplacement characteristics at elevated temperatures

Fig. 1.

T-stub idealisation.

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

649

representative of those achieved in building res. In the past, deformations have been recorded by linear voltage displacement transducers, inclinometers and, in the more distant past, by dial gauges. A problem arises when displacements at elevated temperatures are required.

2. Development of the test programme An experimental method to measure deections of a T-stub assembly, as shown in Fig. 2, by instruments which are not attached to the specimen has been devised and is described in this paper. Although the primary reason for developing the method was the need to record deformations at elevated temperatures, the method is also attractive for tests conducted at ambient temperatures and offers a number of advantages over more conventional methods. 2.1. CCD camera and digital image processing Developments in video camera and digital image processing technology have led to the possibility of producing a real-time displacement measurement from a video image [4]. A solid state CCD (Charge-Coupled Device) camera produces the video signal. The CCD is made out of many charge-coupled cells or picture elements, which are arranged, in a rectangular array. Each picture element, also called a pixel

Fig. 2.

Typical T-stub assembly used in the experiments.

650

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

of the array, converts incoming light into a charge directly proportional to the amount of light received. This charge is then clocked (shifted) from cell to cell, to be nally converted to a video signal that represents the original image at the output of the CCD. The CCD image sensors are stable and accurate devices and the resolution for the current tests is selected to be of an array of, typically 768576, individual lightsensitive elements or pixels. The signal from the camera is digitised, stored and processed by a microprocessor in order to locate the position of the optical targets within the image to sub-pixel accuracy. The above technique measures displacements by tracking two contrasting targets. However, unlike most non-contact optical systems, which measure along one axis, the video extensometer is an area scan device. Due to this it is possible to track the position of the target more precisely so that the measurement is less sensitive to target distortion, testpiece lateral movement or bending. In addition, image processing enables the testpiece shape to be monitored continuously and offers the potential for observing the development of surface features. 2.2. Furnace A purpose-built furnace, with an internal capacity of 1 m3, has been designed with viewports to accommodate three video cameras. It was convenient to use an electric fan-assisted furnace in order to avoid any ames within the eld of view of the video cameras. The fan also ensures a uniform distribution of atmosphere temperature up to 1100C. Two viewports were used to accommodate the three video cameras. One is at the front of the furnace, perpendicular to the axis of loading, facing the horizontal direction where there is movement of the anges. The other is at the top of the furnace, facing the vertical direction, in which it is believed that there will be no movement during the tests. The rst viewport accommodates two video cameras, one for accurate measurements and the other for general observation of the T-stub distortion. The other viewport accommodates another video camera, again for general observation of the T-stub distortion. There are also two opposed holes in the sides of the furnace to allow a hydraulic jack to apply force to the specimens. The image acquisition and processing system had to be carefully selected. First, the required accuracy of the deection measurement was set at 40 m without any sub-pixel division. Then the camera eld of view needed to be investigated, depending upon the position of the video cameras relative to the specimen and the type of lenses to be used; this was set at 3030 mm. Another step was to investigate the resolution of the video cameras in combination with frame grabbers. Three monochrome analogue video cameras, manufactured by JAI (model: CVM50) and a colour frame grabber, manufactured by Imagenation (model: PXC200) were selected, together giving a capture resolution of 768576 pixels. The Imagenation PXC200 frame grabber features precision video capturing hardware for applications that requires high colour accuracy. Features of the precision hardware design include; High colour accuracy

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

651

Fig. 3.

Schematic diagram of the image acquisition and processing system [5].

Image capture resolution up to full-size: 768576 (PAL and SECAM video format) Continuous, software initiated and triggered image captures Four multiplexed composite video inputs The image processing software was designed to include the change of information in an image from low to high temperatures, and the capability of controlling the sub-pixel division. Fig. 3 shows a schematic diagram of the image acquisition and processing system and Fig. 4 the arrangement for the experiments.

Fig. 4. Arrangement for the experimental work.

652

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

2.3. Pilot test to check the accuracy of the image processing software A pilot test to check the accuracy of the purpose built image processing software was performed at ambient temperature. A digital calibration device was used with accuracy in the displacement readings of 1 m. The test procedure was to drive the displacement rod and capture images at 0.50 mm steps. The device is shown in Fig. 5(a) and a typical image used for measurement is presented in Fig. 5(b). Camera One, used to capture images for measuring the displacements, was placed 850 mm away from the targets giving a eld of view of 32.7421.85 mm. In order to capture that eld of view, a 75 mm lens and a double extension tube, which effectively transforms the 75 mm lens into a 150 mm lens, were used. On the calibration device a set of targets was placed and the difference between them was measured during the image processing. The targets are the two black dots shown in Fig. 5(b). The calibration device adjusted the right-hand dot while the lefthand side dot remained stationary. The initial displacement of the two dots was measured (13.386 mm) and used to calibrate the processing method in order to give the actual displacement (in millimetres) of the targets when they start moving apart. The calibration procedure of the processing method rst involves the selection of the area of the left-hand side target as shown in Fig. 6(a). Then an edge detector lter was applied to the image in order to identify the right-hand side target, using the same selected area as shown in Fig. 6(b). The distance between the two bottom left-hand corners was calculated in pixels and transformed into millimetres using the calibration reading above (13.386 mm). According to the above information the accuracy of the software without any subpixel division was up to 42.63 m and a total of 20 images were captured. The results of the real readings compared with the readings from the image processing are shown in Fig. 7. It is clear that there is a good correlation between the displacement readings. The largest difference in the displacement readings was 0.038 mm without any

Fig. 5. Calibration device and a typical measurement image from camera one. (a) Calibration device (b) Image from camera one.

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

653

Fig. 6. Calibration of the processing system. (a) Selection of the left-hand side target. (b) Edge detection lter and identication of the second target.

Fig. 7.

Comparison of the real displacements with the image processing displacements.

sub pixel division. With a sub pixel division of two, which means linear interpolation between the pixels, the largest difference in the readings was 0.025 mm. Fig. 8 below shows the inuence of linear sub-pixel division of the images on the readings. It is clear by comparing the polygons with the origin that there is some improvement in accuracy using sub-pixel division (for the majority of the readings) but the readings do not become more accurate for a sub-pixel division above two. Another set of displacements (3, 5 and 10 mm) was measured using the same calibration device. Twelve readings were recorder and Table 1 shows the statistics out of this pilot test.

654

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

Fig. 8.

Inuence of sup-pixel division to the accuracy of the readings.

Table 1 Mean value, standard deviation and standard error with and without sub-pixel division Displacement readings Sub-pixel division Mean value (mm) Standard deviation (mm) Standard error of sample means (m) 3 mm Without 2 10 2.98 2.98 2.98 10.36 10.30 11.71 2.99 2.97 3.38 5 mm Without 4.98 8.88 2.56 2 4.99 9.05 2.61 10 4.99 11.43 3.29 10 mm Without 10.01 12.61 3.64 2 10 10.01 10.01 12.89 10.33 3.72 2.98

3. Test procedure The procedure followed during the elevated temperature tests was rst to take the specimens up to the desired temperature and then apply the load. Thermocouples were positioned around the specimens in order to monitor the temperature and as soon as the specimen reached the desired temperature a load step was applied using the control panel of the hydraulic jack. According to the geometry of each specimen, images were captured at different load steps, sometimes every 1 kN in the elastic region, 0.5 kN in the elastoplastic region and back to 1 kN in the plastic region, and sometimes every 5 kN in the elastic, 1 kN in the elastoplastic and 2.5 kN in the plastic region. For this test program a series of 12 tests was carried out using two video cameras placed in the front view port of the furnace. The rst camera captured images for accurate displacement measurements and the other for general observation of the T-

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

655

stub distortion. Typical distorted images at 570C taken from the two cameras are shown in Fig. 9. The targets used to measure displacements are shown in Fig. 9(a). They have been made using a 0.5 mm drill and the distance between them was measured (using a Vernier), before placing the specimen in the furnace. After nishing the tests, the images from camera one were processed. 3.1. Results for the T-stub arrangement at elevated temperatures and comments Test Program A involved twelve tests and these tests were divided into three categories (A1, A2 and A3) according to the geometry of the specimens. Figs. 10 and 11 with Tables 2 and 3 shows the specimens arrangement for all three categories. Table 4 shows the total tests conducted with the maximum force applied the number of images taken and the accuracy of each test at elevated temperatures. Fig. 12 shows typical results, from the image processing technique, plotted against the Force. From Fig. 12 the tests at ambient and 360C temperatures were unloaded but the last one failed by fracture of the bolts.

4. Discussion The accuracy of the image processing software for T-stub tests seems to be adequate. The use of image acquisition and processing technique has many advantages in comparison with conventional mechanical instruments and offers a wide range of testing applications and benets. Some of the benets using this technique are: 1. Direct and non-contacting measurements means that nothing need to be attached to the specimen.

Fig. 9.

Typical distorted images at 570C. (a) First camera image. (b) Second camera image.

656

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

Fig. 10.

Arrangement of the T-stub specimen.

Fig. 11.

Plan view arrangement of the T-stub specimen.

2. Multiple choice of gauge length and operating range because of the exibility in choosing the eld of view for the camera (the limitation here is the resolution of the picture which is dependent on the required accuracy of the measurements). 3. Use in connection with furnaces (elevated temperatures) or within a hostile environment as the video camera can be set some distance away from the test specimen.

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

657

Table 2 Geometrical data Test program Flange 1 (UB30516540) Flange 2 (UB30516540) A A1 A2 A3 A1 A2 A3 h 151 151 151 151 151 151 L 220 140 100 220 140 100 b 165 165 165 165 165 165 d0 16 16 20 16 16 20 e 25 45 25 25 45 25 tw 6.5 6.5 6.5 6.5 6.5 6.5 tf 10.2 10.2 10.2 10.2 10.2 10.2

Table 3 Position of the bolt holes Test program Flange 1 (UB30516540) Flange 2 (UB30516540) A Le e m Ln n Sw Bolt size M16 M16 M20 M16 M16 M20

A1 A2 A3 A1 A2 A3

165 165 165 165 165 165

25 45 25 25 45 25

54.25 34.25 54.25 54.25 34.25 54.25

220 140 100 220 140 100

40 35 25 40 35 25

140 70 50 140 70 50

Table 4 T-stub tests at elevated temperatures. Temperature (C) Maximum force (kN) Total images (Nr) Accuracy without sub-pixel division (m) 31.15 34.11 35.65 34.00 35.78 39.25 38.28 37.75 39.13 40.85 40.15 37.75

T-stub A1

T-stub A2

T-stub A3

Ambient 360C 600C Ambient 320C 460C 550C 660C 200C 420C 570C 750C

205 200 90 250 250 245 175 87.5 175 135 68 24

53 45 29 52 60 39 28 28 64 94 65 24

658

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

Fig. 12. Typical force displacement curves at elevated temperatures.

4. Entire testing procedure can be observed on a monitor and even recorded so it is possible to go back and collect more information by looking the behaviour of the specimen. 5. The advantage of processing a picture is the availability of information in two directions, so it is possible to measure displacements in the transverse direction simultaneously, if needed. 6. Permits the comparison of the deected shapes of the specimen at elevated temperatures with nite element analysis deected shapes.

5. Conclusion This paper has described the use of image acquisition and processing software in T-stub tests at elevated temperatures. This technique is particularly interesting because aside from obtaining displacement readings at high temperatures it also records the behaviour of the specimen as soon as the load is applied. This is an advantage, because by careful observation of the recorded test, it is possible to recognise the failure mechanism of the specimen. This information makes the image acquisition and processing technique, attractive for ambient temperature tensile or compressive tests for steel or concrete. The T-stub tests at elevated temperatures are part of a programme, to investigate the behaviour of a steel beam-to-column joint in re conditions. The other parts of the program are to investigate by testing the compression and shear zone in a steel joint. The nal aim is to produce MomentRotationTrust curves for steel beam-tocolumn joints at elevated temperatures which should be helpful to researchers in understanding the behaviour of steel structures under re conditions.

S. Spyrou, J.B. Davison / Journal of Constructional Steel Research 57 (2001) 647659

659

Acknowledgements The work described herein is funded by EPSRC whose support is gratefully acknowledged.

References
[1] Jaspart JP, Weynand K, Steenhuis M. The stiffness model of revised Annex J of Eurocode 3. 3rd International Workshop on Connections in Steel StructuresBehaviour, Strength and Design, Trento, Italy, 2931 May 1995, CH.53, 1996. p. 44152. [2] Leston-Jones LC, Burgess IW, Lennon T, Plank RJ. Elevated temperature momentrotation tests on steelwork connections. Proc Inst Civ Eng Struct Bldgs 1997;122:4109. [3] Engineering and Physical Sciences Research Council research grant (GR/L98619 and runs from 1 November 1998 until 31 October 2001). [4] MCenteggart I. Contacting and non-contacting extensometry for ultra high temperature testing. In: Lohr RD, Steen M, editors. Ultra high temperature mechanical testing. Cambridge: Woodhead, 1995. [5] Software picture ULR, http://www.search.corbis.com, Frame Grabber and Computer picture ULR, http://www.natinst.com, Cameras picture ULR, http://www.jai.dk/ukFindex.html