Advances in slabs defects inspection with Conoscopic Holography. Ignacio Alvarez, J.M. Enguita (UniOvi) J. Marina, R.

Transcription

1 Advances in slabs defects inspection with Conoscopic Holography Ignacio Alvarez, J.M. Enguita (UniOvi) J. Marina, R. García (DSIplus)

2 Advances in slabs defects inspection with C.H. Contents: Surface inspection in Continous Casting A new technology: Conoscopic Holography Cracks detection in Continuous Casting Improvements in C.H. sensors Conclusions

3 Surface inspection in hot steel products using C.H. Contents: Surface inspection in Continous Casting A new technology: Conoscopic Holography Cracks detection in Continuous Casting Improvements in C.H. sensors Conclusions

4 Surface inspection in harsh environments Goals of on-line surface inspection: Detect defects Measure shapes Classify types of defects Provide on-line results to operators Give a general quality value for each piece Generate databases for claims or statistics Help to improve process knowledge Perform real-time automatic decisions (only in easy cases and controllable conditions)

5 Surface inspection in harsh environments Slabs in continuous casting. Surface inspection conditions: Heat: target at ºC in the inspection point. Sensors must be located at a safe distance for cooling. Scale covers partially the surface. Sensors technology must be able to distinguish scale from the steel surface. Bottom side inspection. Space is limited in the bottom side. Falling objects (scale) and dust. Defects to detect. Cracks and longitudinal marks must be detected. Other changes in the surface aspect are not defects. Automatic classification. Return of investment is maximized if the system is able to classify automatically a slab: repair (defects) / hot charge (no defects). Selected technology must be robust to classify in changing surface conditions.

6 Cracks detection in continuous casting On-line detection of defects at 800ºC has an important economical impact for the plant: Without automatic inspection Slabs marked as suspicious, depending on steel grade and process parameters, must always go to manual review. Slabs not marked as suspicious are never reviewed. Process knowledge is limited and can not be applied on-line. With automatic inspection Only slabs with defects go to manual repair. Personnel, storage and energy costs are saved. Surface defects, that may rarely appear in non suspiciuos slabs, are detected and repaired. Process knowledege is improved, and can be applied almost in real-time. Other possible problems affecting the quality (roll jams that make longitudinal marks in the surface) are not detected. Other problems are detected on-line and can be corrected as soon as possible.

7 Surface inspection in harsh environments Machine vision is the usual technology for surface inspection, but has several drawbacks when used in complex environments: Common problems in surface inspection in harsh environments Defects may only be smooth variations of the surface. Formations that are not defects may appear in the surface. Material and environment conditions are variable, and often out of control. Machine vision drawbacks If the illumination of the scene does not generate light intensity variations for the defects, nothing is detected. If these formations generate similar light intensity variations, they could only be discriminated by smart classifiers. The algorithm s robustness is reduced as the number of scenarios rises.

8 Surface inspection in harsh environments Distance sensors: the alternative They provide new information: the topography of the surface. Typical configurations: Laser triangulation Fringes projection triangulation But Are usually slower Are difficult to setup and calibrate in complex enviroment Are more affected by the harsh conditions: heat, dust, smoke Are more expensive

9 Surface inspection in harsh environments Slabs surface inspection. Available technologies: Machine vision: Hardware is simple and cheap. Information returned by the sensors is poor for automatic detection. Changing surface conditions generate difficulties for parameters setup. EMATS, ultrasound, : Can detect small sub-surface defects not revealed in the surface. Sensors must be located very close to the surface. Scale and surface variations affect largely the detection. Optical triangulation (laser, fringes projection): Hardware is simple and non-expensive. Sensors generate distance information. Space is limited in the bottom side. Air turbulences affect the detection. Changing surface conditions generate difficulties for parameters setup. Conoscopic holography profile sensors: Sensors generate distance information. Easy setup and high robustness in harsh conditions. Hardware is expensive.

10 Surface inspection in hot steel products using C.H. Contents: Surface inspection in Continous Casting A new technology: Conoscopic Holography Cracks detection in Continuous Casting Improvements in C.H. sensors Conclusions

11 The technology: Conoscopic Holography Optimet s Long Stand-off Conoline : Profilometric conoscopic holography in phase configuration Technical specifications for inspection of hot product: Long stand-off for heat protection (>= 1000 mm). High sensitive profilometric measurement: resolution under 0.1 mm. Robust against dust, heat and air turbulences. Almost collinear: easy setup in the upper and the lower face. Easy calibration, reduced maintenance

12 The technology: Conoscopic Holography Conoscopic holography principle:. Light emitting point Circular polarizers Uniaxial crystal Light valve Detector Incoherent light interferometry Based on double refraction property of uniaxial crystals Ordinary ray travels at constant speed. Extraordinary ray speed depends on angle of incidence r o Interferometric pattern: Gabor Zone Lens, easily captured by a standard CCD. r e Interferogram Fringes frequency indicate distance of the emitting point

13 The technology: Conoscopic Holography Typical configuration: co-linear punctual sensor (Conoprobe) Laser No problems due to occlusions Measured point Mirror Conoscopic module Beam splitter Objective & filters Linescan CCD Up to 85º inclined surfaces can be measured A simple change of lens modifies range and resolution Fringe pattern for close point Surface to measure Fringe pattern for far point All types of surfaces can be measured with the same sensor

14 2 mm The technology: Conoscopic Holography Long stand-off profilometer: Small triangulation 1150 mm stand-off, 300 mm aperture, 0.1 mm resolution Distance recovery from phase information 0.3 Conoscopic sensor Interferogram Frequency Phase -0.3 Laser line Depression in the surface Unwrapped phase of all the rows mm Distance map from detrended unwrapped phase

15 The technology: Conoscopic Holography A comparison in hot steel: C.H. long stand-off profilometer with 700 mm stand-off (red line). Laser triangulation profilometer with 500 mm stand-off and 500 mm triangulation aperture.

16 Surface inspection in hot steel products using C.H. Contents: Surface inspection in Continuous Casting A new technology: Conoscopic Holography Cracks detection in Continuous Casting Improvements in C.H. sensors Conclusions

17 Cracks detection in continous casting Upper side inspection: 1 linescan camera and 6 conoscopic sensors for 1800 mm width aperture Sensors can move in the vertical direction to adapt to different slab thickness Lower side inspection: 1 linescan camera and 6 conoscopic sensors with mirrors for easy installation Windows and shield for protection from dust and falling objects

18 Cracks detection in continuous casting Industrial system setup Conoscopic sensors Vertical motor Thermal protecting shield Linescan camera Cover with windows (opened) Conoscopic sensors

19 Cracks detection in continuous casting Industrial system on-line Scanned line Carriot on-line/ off-line movement

20 Cracks detection in continous casting Main data available from each sensor: Each sensor provides a distance profile for each acquisition. The advance of the slab in the strand gives the 3rd coordinate. The 3D topography map is processed by specialized machine vision algorithms for depressions detection. Longitudinal depressions with a particular shape are considered defects.

21 Cracks detection results Different configurations of defects All of them must be detected, but different severities must be assigned Long deep crack Long narrow crack Short cracks Jammed-roll mark

22 z (mm) z (mm) Cracks detection results Typical distance profiles: x (mm) x (mm) The main problems: True surface recovery (elimination of scale) Robust 2D phase unwrapping

23 Processing steps The different conditions (dust, scale, heat, reflectivity) are robustly managed by the sensor and the software. Auxiliary information from the sensors gives robustness to the detection: Luminosity Signal to noise ratio

24 Processing steps Using this additional information, non-defects do not give false detections Depressions betwen scale formations Problems due to deburring Problems due Lack of planarity to dust and particles

25 Results On-line inspection system based on conoscopic holography long stand-off profilometers: Working in harsh conditions (heat, dust, falling objects) since January No modifications in the plant throughput. High reliability (>95% usability). High productivity (more than 6,000,000 tons of steel inspected so far), 24h/24. High economic impact for the plant.

26 Surface inspection in hot steel products using C.H. Contents: Surface inspection in Continuous Casting A new technology: Conoscopic Holography Cracks detection in Continuous Casting Improvements in C.H. sensors Conclusions

27 Improvements in the sensors 2 nd generation sensors available in 1 st Q 2011: Improved speed. Improved lateral resolution. Blue laser best fitted for hot object. Spec. Current sensor New sensor Improvement Speed 60 profiles/sec prof./sec Longitudinal resolution x 5 Lateral resolution 640 pixels/320 mm = 0.5 mm 1280 pixels/320 mm= 0.25 mm Lateral resolution x 2 Depth resolution 0.1 mm <0.1 mm Depth resolution improved in hot conditions. Light source Red laser 660 nm 100 mw Blue laser 405 nm 200 mw Improved performance in hot conditions. High quality luminosity image Camera Analog Digital GigE Increased distance to processing computers Processor Pentium M 1.5 GHz (32 bit) Pentium IV Quad Core 2.66 GHz (64 bit) Maximum processing speed from 90 profiles/sec to 500 profiles/sec

28 Improvements in the sensors 2 nd generation sensors available in 1 st Q 2011: Target GigE link up to 300 fps 1400x256x8 Sensor Box 100 Mbps Ethernet link Mini-ITX computer Embedded Linux-64 bit Up to 500 fps processing Surface Processing and visualisation

29 Improvements in the sensors 2 nd generation sensors available in 1 st Q 2011: Laser line Target Processing software Sensor box

30 Improvements in the sensors Hot sample tests: The sample is heat in an oven up to 1100ºC A heat protected sensor with cold mirror is used for scans Sample steady, sensor moving in a charriot Before heating, chariot out Sample Oven Thermal shield After heating, chariot in Sample Window Sensor Mirror Charriot Advance

31 Improvements in the sensors Hot sample results: A sample with a natural defect is heat to 1000ºC Movement of the chariot= 10 m/min LS-Conoline LS-Conoline blue V2.0

32 120 mm Improvements in the sensors Hot sample results: The graylevel luminosity image is more meaningful in the new sensor LS-Conoline LS-Conoline blue V mm

33 30 m Improvements in the sensors Under research: 2 KHz sensor with 4000 points/line Results already obtained for roughness measurements at 7 KHz Muestra 513: Interferograma a procesar m Fase desplegada Frecuencia

34 Surface inspection in hot steel products using C.H. Contents: Surface inspection in Continuous Casting A new technology: Conoscopic Holography Cracks detection in Continous Casting Improvements in C.H. sensors Conclusions

35 Conclusions Conoscopic holography: Non-contact technology for surface inspection in harsh environments: Long stand-off for heat protection Compact sensor (no triangulation) Robust against heat, dust and changing surface Profilometric measurements: On-line surface defects detection Surface shape measurement Surface inspection in Continuous Casting: Proven reliability working on-line since 2007 in Continous Casting plant. New sensor improvements: New sensors improve speed, accuracy and robustness of the detection.

36 On-line surface defects detection in C.C. hot slabs Thank you for your attention Contact information: Ignacio Alvarez University of Oviedo Jorge Marina DSIplus S.L.