- 1 -

Feasibility Study of Infrared Detection of Defects in Green-State and Sintered PM

Compacts

Report No. PR-06 - #2

OBJECTIVE The objective of this research is the development of a defect detection apparatus capable of detecting defects in P/M compacts. In addition to the part evaluation this apparatus will be capable of detecting some changes in the manufacturing process thus allowing part makers to calibrate the manufacturing line to fabricate sound parts. This test system has the ability to provide one hundred percent quality assessment early in the P/M manufacturing process. Complementing our theoretical investigations and computational modeling we have devised an algorithm to process and analyze IR images; this algorithm is the basis of our software development. It includes:

• Real time data collection, processing, and image display, and • A simple and friendly graphical user interface (GUI) for press operators and

quality engineers.

The milestones identified by our focus group members for this fall meeting encompass:

a) Completion of a stand-alone software package that interfaces with the IR camera system with two options:

Research Team: Reinhold Ludwig (508) 831-5315 ludwig@wpi.edu Souheil Benzerrouk (508) 831-6797 souheil@wpi.edu

Focus Group Members:

Chaman Lall Metal Powder Products, Chair Richard Scott Nichols Portland Michael Krehl Sinterstahl Hannes Traxler PLANSEE Aktiengesellschaft

- 2 -

1. Simple pass/fail feedback for the machine operators. 2. Comprehensive data visualization and analysis for a more in-depth failure

study.

b) Exploration of a low-cost IR test system that consists of a basic, striped-down IR camera without expensive peripherals. This system can then be used as a test bed for the image analysis and crack detection software package. The hardware components will be funded by Dr. Chaman Lall, MPP.

c) Completion of a patent application through WPI.

APPROACH The PMRC focus group members identified four major tasks to be conducted during the Spring 2006 - Fall 2006 time period. Specifically, emphasis is placed on:

• Investigating adequate data analysis algorithms to perform advanced image analysis of the green-state compacts,

• Writing a Visual C++ program that allows high speed communication with the

camera for data collection and camera control (for synchronization and operation),

• researching various camera options and manufacturers to ultimately arrive at a low-cost IR system for the factory floor, and

• Filing for a provisional patent to protect the developed intellectual property.

ACCOMPLISHMENTS For the reporting time period, the researchers can document the following accomplishments:

• Researched various camera options and possibilities.

• Completed a preliminary software package capable of: Connecting a standard computer system to the camera system. Basic camera controls: On/Off, focus, zoom. Real time data collection. IR image display. Trigger recording events. Alarms/threshold. Spot/area selection and temperature display. Ability of loading images for data processing.

- 3 -

• Applied for a provisional patent.

FUTURE WORK For the next and final quarter we intend to:

• Finalize the software development to include image processing and real time feedback.

• Finish the theoretical study and modeling, including the flaw representation.

• Conduct further on-line testing of this new IR inspection prototype (including the

software) at a designated P/M part manufacturing facility. The testing should preferably include a long term inspection cycle (up to a week at a specified line). Sites that have been identified include:

Metal Powder Products in St Mary’s, PA. Nichols Portland, in Portland, ME.

REPORT ORGANIZATION Appendix A – Contains a paper presented at QNDE 2006: 33rd Annual Review of Progress in Quantitative Nondestructive Evaluation, Portland, Oregon, July 31-Aug 4, 2006. Appendix B – Contains a paper that presented at the International Conference on Powder Metallurgy and Particulate Materials, San Diego CA, June18-21 2006. ACKNOWLEDGEMENT We would like to extend appreciation to our focus group members for their guidance and their important input throughout this project period. In particular, we would like to thank Dr. Chaman Lall (Metal Powder Products) and Richard Scott (Nichols Portland) for their help in providing insight proved invaluable.

- 4 -

Report No. 06-# 2 APPENDIX A ACTIVE THERMOGRAPHY FOR THE DETECTION OF DEFECTS IN POWDER METALLURGY COMPACTS

Souheil Benzerrouk1,2, Reinhold Ludwig1,2 and Diran Apelian1

1Powder Metallurgy Research Center, METAL PROCESSING INSTITUTE 2Department of Electrical and Computer Engineering

Worcester Polytechnic Institute, Worcester, MA 01609

ABSTRACT. Active thermography is an established NDE technique that has become the method of choice in many industrial applications which require non-contact access to the parts under test. Unfortunately, when conducting on-line infrared (IR) inspection of powder metallic compacts, complications can arise due the generally low emissivity of metals and the thermally noisy environment typically encountered in manufacturing plants. In this paper we present results of an investigation that explores the suitability of active IR imaging of powder metallurgy compacts for the detection of surface and sub-surface defects in the pre-sinter state and in an on-line manufacturing setting to ensure complete quality assurance. Additional off-line tests can be carried out for statistical quality analyses. In this research, the IR imaging of sub-surface defects is based on a transient instrumentation approach that relies on an electric control system which synchronizes and monitors the thermal response due to an electrically generated heat source. Preliminary testing reveals that this newly developed pulsed thermography system can be employed for the detection of subsurface defects in green-state parts. Practical measurements agree well with theoretical predictions. The inspection approach being developed can be used for the testing of green-state compacts as they exit the compaction press at speeds of up to 1,000 parts per hour. Key words: Active thermography, defect detection, dynamic temperature recording. PACS: 81.70Ey, 41.20.Cv, 44.05.+e, 44.40.+a, 02.60.Cb, 02.70.Dh, 07.57.Ty, 61.43.Gt

INTRODUCTION

The applicability of active infrared (IR) imaging for the detection of surface and subsurface defects in green state powder metallurgy (P/M) compacts was confirmed through theoretical analysis, comprehensive modeling, and extensive testing.

Theoretically, we have employed a combination of theoretical and computational methods to model the idealized defect response. For the numerical simulations we combined an electromagnetic model with a heat transfer model to evaluate the thermal behavior of a part subjected to induction heating. Experimentally, we have analyzed single and multi-level green-state parts in a laboratory environment as well as in a manufacturing setting where parts are inspected as they reside on a moving conveyer belt in a production plant.

- 5 -

In this paper we show that the application of an alternating current excitation allows us to maintain uniform, and sufficient heat deposition over the part surface. Moreover, we demonstrate that the IR system is capable of collecting reliable thermal data from a process line producing complex multi-level gears at high speed. The camera system is able to focus and image at a high frame rate; this permits comprehensive data analysis both in time and space.

ANALYTICAL DEFECT MODELING

A defect can be regarded as a discontinuity in the material and can thus be modeled as either a source or a sink of energy depending on the electric energy deposition. To capture the impulsive thermal response due to such a point defect, we can employ the method of images. This concept is based on creating a source, or sink, in the domain normal to the plane of symmetry, as depicted in FIGURE 1. The source and the image are paired at the observation boundary [3]. The diffusion equation in the time domain is modified to include the impulsive point source. The corresponding Green’s function G(r,r0) is then written as

( ) ( )0

224 tt

t

GG !!!="

#

$%&

'

(

(!) *+*,

0rr (1)

where we specify via delta functions in space and time, i.e. ( )0rr !" and ( )

0tt !" , the

point source at location r0 and at time t0 with unit strength. In (1),! is the thermal diffusivity, t is time and r is the observation location. The corresponding thermal impulse response, or Green’s function solution for a semi-infinite space with a flux-free boundary, becomes in a two-dimensional Cartesian coordinate system:

( )

( ) ( )

!!!!

"

#

$$$$

%

&

++'+

'+'=

++''

'+''

22

4

)'()'(

22

4

)'()'(

3 )'()'()'()'(

1

2222

yyxx

e

yyxx

e

tG

t

yyxx

t

yyxx

((

) (2)

where (x,y) denotes an observation point in the simulated half-space, and (x’,y’) refers to the point defect location. Depicted in

- 6 -

FIGURE 2 is the impulsive temperature increase, G, as a function of time and space (x-direction) and recorded over the surface of a half-space (y=0) for the following parameters: simulated surface length: 10cm, defect location: (x,y) = (5cm,5cm). Although this Green’s function solution is applicable only for idealized half-space considerations, it serves as a good approximation to set defect resolution limits and provide instrument calibration.

FIGURE 1. Idealized sub-surface defect representation as an embedded point source.

OFFLINE TESTING

The green-state P/M gear shown in FIGURE 3 exhibits a typical crack encountered in gear manufacturing [7]. It also represents a particularly difficult to detect crack for the IR imaging though electric heating. The gear teeth cause a non-uniform heat deposition throughout the part, which in turn causes reflections. The steel powder compacts are constructed with 1.0% Cu, 0.2% C and lubricated with 0.8% wax and based on density ranges from 6.8 g/cm3 to 7.1 g/cm3. The defect seen in Figure 3 is a hairline crack (approximately 20microns in width) of 5mm length. This type of defect was chosen because it is not easily detectable if the source of heating is direct current (DC) as it requires high current density and additional electrode contacts to insure uniform current flow throughout the part [1]. An improved inspection approach utilizes an AC current excitation where the frequency of the source dictates the depth of penetration of the current. This approach ensures that the current flows on and near the surface of the compact, hence increasing the thermal signature of the defect. FIGURE 4 depicts our test arrangement; it utilizes an induction-heating unit that consists

Image

x

y

Defect

Imaged surface

of the P/M part

Convective

boundary

- 7 -

of a power supply and an induction coil suspended below the compact, rendering the technique contact-less.

FIGURE 2. Temperature rise recorded over the compact surface due to an embedded heat source.

FIGURE 3. A gear P/M compact with a surface crack situated on the tooth surface.

Figure 5 presents IR images for 2D surface and line profiles (along the dotted line). The data is collected with an IR camera positioned 50cm away and viewed from the side; it is operated at a frame rate of 30Hz. The field of view of the 240 by 320 pixel picture is 15cm by 15cm. The total line length of 10cm is subdivided into 180 points (i.e. with a point-to-point resolution of 0.5mm) whereas the thermal pixel intensity is displayed in discrete increments up to a maximum discrete level of 260K (or 460K).

0

2

4

6

8

10

-10

-5

0

5

10

0

10

20

30

40

50

60

Time (s)

Surface length (cm)

Temperature rise (K)

5

10

15

20

25

30

35

40

45

50

55

- 8 -

FIGURE 4. Contact-less test arrangement of active IR testing unit.

1 cm

0 10 20 30 40 50 60

60

80

100

120

140

160

180

200

220

240

260

Distance along profile

Pixel intensity

FIGURE 5. (a) Initial image from the IR recording of the gear part shown in FIGURE 3 and subjected to an AC current, (b) thermal profile along the dotted line.

We devised a system configuration that relies on induction heating to generate the thermal gradient which is subsequently recorded through the IR camera. This method takes advantage of the skin effect, where the frequency controls the depth of electromagnetic field penetration into the compact [11]. Another important component of our system is the signal processing of the thermal response. We can apply several processing tools to extract defect information from the image. Specifically, we have analyzed the time derivative of the thermal profiles for several pre-selected spots on the surface of the compact. In subsequent processing we

(a) (a) (a) (a) (a)

(b)

- 9 -

have extended this method to include a so-called Laplacian operator to search for surface cracks.

1 cm

0 10 20 30 40 50 60

160

170

180

190

200

210

220

230

240

250

260

Distance along profile

Pixel intensity

FIGURE 6. (a) Initial image from the IR recording of the defective gear part shown in FIGURE 3 (a), (b) thermal profile along the dotted line.

ON-LINE TESTING

Tests at various manufacturing facilities allow us to establish the stability of the inspection system and its immunity from temperature fluctuations in the plant arising from production equipment such as presses, motors, and sinter furnaces [12]. It became necessary to validate the IR system for the detection of real and commonly observed defects in the process line where our plan to extend the usability of this method requires careful analysis to detect defects regardless of material composition. Aluminum powder presents a unique challenge: it is a highly reflective material with very low emissivity (0.1 to 0.2) when compared to steel parts with high graphite content and emissivity on the order of 0.6. FIGURE 7 shows the green-state steel P/M sample. The compact is a two level gear with 13mm in height by 60mm in diameter and is typically manufactured at a rate of approximately 600 parts per hour.

A long IR image sequence of 45 seconds generates 1,350 recorded temperature points with an intensity profile depicted in FIGURE 9. As expected, as soon as a P/M compact moves past the fixed spatial sensing location, the temperature increases. FIGURE 8, shows the location of the temperature tracking point.

(a) (a) (a) (a) (a)

(b)

Crack signature

- 10 -

FIGURE 7. Picture of a green-state P/M part to be tested at a manufacturing facility.

1cm

FIGURE 8. Temperature monitoring location at a spatial point inside the compaction press.

- 11 -

300

305

310

315

320

325

330

335

0 5

10

15

20

25

30

35

40

45

Time (s)

Te

mp

era

ture

(K

)

FIGURE 9. Temperature (in K) recorded at a fixed spatial location (one spot) over time.

A detailed investigation of the data sequence reported in FIGURE 9 allows us to conduct a more thorough analysis, as depicted in FIGURE 10. As can be seen through direct visual inspection in FIGURE 10, several parts are defective. It is believed that this methodology has the potential of being a very simple, yet reliable methodology that allows us to identify defective parts in an on-line setting.

300

305

310

315

320

325

330

335

0 1 2 3 4 5 6 7 8 9

10

Time (s)

Te

mp

era

ture

(K

)

FIGURE 10. Zoomed-in temperature profile (in K) recorded at a fixed spot location.

Start defects End defects

- 12 -

CONCLUSIONS The data collected at various P/M production facilities indicates that this IR testing methodology appears to be suitable for implementation on the manufacturing floor. It poses little intrusion to the fabrication process and preliminary testing suggests that the IR imaging is relatively immune to changes in the background noise generated by the manufacturing equipment. Our data collection has the potential of detecting defects under real-time production conditions. We are confident that this system will ultimately be capable of performing a 100 percent quality assessment of green-state P/M compacts by providing on-line operator feedback. Additional off-line information about defect parameters can also be extracted for engineering purposes and for process calibration.

Besides environmental immunity, the IR testing methodology appears sufficiently robust to handle different material compositions, compaction densities, and lubricant inclusions. In particular, we were able to test aluminum parts that have detrimental testing characteristics such as high reflectivity and high cooling rate.

Presently all signal processing and data analysis is performed off-line. For a fully manufacturing-compliant system it becomes necessary to integrate these steps into a rapid data collection and processing environment for real-time feedback.

REFERENCES

1. Leuenberger, G. "Electrostatic Density Measurement in Green-State P/M Parts" PhD thesis, ECE Department, Worcester Polytechnic Institute 2003.

2. Maldag, X.P.V. "Theory and Practice of Infrared Technology for Nondestructive Testing" John Wiley & Sons Inc. 2001.

3. Morse, P.M, Feshbach, H. “Methods of Theoretical Physics” McGraw-Hill Book Company, Inc. 1953.

4. Burnay, S. G., Williams, T. L., Jones C. H. "Applications of Thermal Imaging" 10.P Publishing 1988.

5. Carslaw, H., Jaeger, J. "Conduction of Heat in Solids" Second Edition, Oxford University Press 1959.

6. Incropera, F.P., DeWitt, D.P. "Fundamentals of Heat and Mass Transfer" 4th edition, John Wiley & Sons, New York 1996.

7. German, R.M. "Powder Metallurgy Science" Metal Powder Industries Federation, Princeton, New Jersey, 1984.

8. Ringermacher, H.I., Howard, D.R. and Gilmore, R.S., "Discriminating Porosity in Composites Using Thermal Depth Imaging" CP 615, Review of Quantitative Nondestructive Evaluation, Vol. 21, ed. by Thompson and D.E Chimenti. American Institute of Physics. 2002.

9. Sun, I.G "Analysis of Quantitative Measurements of Defects by Pulsed Thermography Imaging" CP 615, Review of Quantitative Nondestructive Evaluation, Vol. 21, ed. by Thompson and D.E Chimenti. American Institute of Physics. 2002.

10. Powder Metallurgy Research Center (PMRC), Metal Processing Institute, Worcester Polytechnic Institute, fall meeting, Oct. 20, 2004.

- 13 -

11. S. Benzerrouk R. Ludwig and D. Apelian: “Electrothermal Defect Detection in Powder Metallurgy Compacts”, Proceedings of the 2005 International Conference on The Review of Progress in Quantitative Nondestructive Evaluation, Vol. 25. pp. 1201-1208, Published by the American Institute of Physics.

12. S. Benzerrouk, R. Ludwig, and D. Apelian: “Contact-less Active Infrared Imaging System for the Detection of Defects in Green-State P/M Compacts", Proceedings of the 2006 International Conference on Powder Metallurgy & Particulate Materials (PowderMet 2006 –San Diego), published by MPIF, Princeton, NJ, Part 11, pp. 25-42, 2006.

- 14 -

Report No. 06-# 2 APPENDIX B

Contact-less Active Infrared Imaging System for the Detection of Defects in Green-State P/M Compacts

Souheil Benzerrouk1,2, Reinhold Ludwig1,2 and Diran Apelian1

1Powder Metallurgy Research Center

METAL PROCESSING INSTITUTE

and 2Department of Electrical and Computer Engineering

Worcester Polytechnic Institute, Worcester, MA 01609

ABSTRACT

The applicability of active infrared (IR) imaging for the detection of surface and subsurface defects in green state powder metallurgy (P/M) compacts was confirmed through a theoretical analysis, a comprehensive modeling, and extensive testing. Theoretically, we have employed the method of images to model the idealized defect response. For the numerical simulations we combined an electromagnetic model with a heat transfer model to evaluate the thermal behavior of a part subjected to induction heating. Experimentally, we have analyzed single and multi-level green-state parts in a laboratory environment as well as in a manufacturing setting where we tested parts as they travel in the conveyer belt in a production plant. In this paper, we show that the application of alternating current excitation allows us to maintain uniform and sufficient heat deposition over the part surface. Moreover, we demonstrate that the IR system is capable of collecting reliable thermal data from a process line producing complex multi-level gears at high speed. The camera system is able to focus and image at a high frame rate, permitting comprehensive data analysis both in time and space.

- 15 -

INTRODUCTION

Active thermography is an established NDE technique that has become the method of choice in many industrial applications that require non-contact access to the parts under test [2]. However, when conducting on-line infrared inspection of powder metallic compacts, complications can arise due the generally low emissivity of the metal powders and the thermally noisy environment in manufacturing plants. In this paper we present results of an investigation that tests the suitability of active IR imaging of P/M compacts for the detection of surface and sub-surface defects in the pre-sinter state in an on-line manufacturing environment to ensure hundred percent quality assurances. Additional off-line tests are then carried out for a statistical quality analysis. Specifically, the IR imaging of sub-surface defects is based on a transient instrumentation approach that relies on an electric control system that synchronizes and monitors the thermal response due to an electrically generated heat source [10].

Preliminary testing reveals that this newly developed pulsed thermography system can be employed to detect subsurface defects in green-state parts. Practical measurements agree well with theoretical predictions. The inspection approach presently under development targets the testing of all green-state compacts as they exit the compaction press at speeds of up to 1,000 parts per hour.

ANALYTICAL DEFECT MODELING

A defect is a discontinuity in the material and can be modeled as either a source or a sink of energy depending on the electric energy deposition. To capture the impulsive thermal response, we can employ the method of images, a technique widely used in electromagnetics. This concept is based on creating a source, or sink, in the plane or space normal to the plane of symmetry [3], as depicted by FIGURE 1.

Figure 1: Idealized sub-surface defect representation as an embedded point source.

Image

x

y

Defect

Imaged surfaceof the P/M part

Convectiveboundary

- 16 -

The corresponding thermal impulse response, or Green’s function solution, can be written as follow:

( )

( ) ( )

⎥⎥⎥

⎦

⎤

⎢⎢⎢

⎣

⎡

++−+

−+−=

++−−

−+−−

22

4)'()'(

22

4)'()'(

3 )'()'()'()'(1

2222

yyxxe

yyxxe

tG

tyyxx

tyyxx

ππ

κ

(3)

Here (x,y) indicate a point in the simulated half-space and (x’,y’) denote the defect coordinates, κ is the thermal diffusivity and t is the time FIGURE 2 depicts the temperature increase (G) over time and the length in the x-direction on the surface of the part (y=0) for the following parameters:

• Part surface Length: 10cm • Defect location: (x,y)=(5cm,5cm) • Heating power: Q=8 Watts

Figure 2: Temperature rise over the compact surface due to an embedded heat source.

02

46

810

-10-5

0

510

0

10

20

30

40

50

60

Time (s)

Surface length (cm)

Tem

pera

ture

rise

(K)

5

10

15

20

25

30

35

40

45

50

55

- 17 -

OFFLINE TESTING OF COMPLEX PARTS The gears shown in FIGURE 3, present two typical difficulties encountered owing to their geometric arrangement. First, the gear teeth cause non-uniformity in the part, which in turn causes reflections. Second, the multilevel nature of the part makes it prone to corner cracks which cannot easily be detected as a result of complicated heat transfer mechanisms at the corner.

Figure 3: (a) A gear compact with a surface crack situated on the tooth surface, and (b) the same gear compact with a surface-breaking flaw at the corner

These steel powder compacts are constructed with 1.0% Cu, 0.2% C and lubricated with 0.8% wax. The density ranges from 6.8 g/cm3 to 7.1 g/cm. These defects are not easily detectable if the source of heating is direct current (DC) because it requires high current density and additional electrode contacts to insure uniform current flow throughout the part. An improved inspection approach utilizes an AC current excitation where the frequency of the source dictates the depth of penetration of the current. This approach will ensure that the current flows on and near the surface of the part, hence increasing the thermal signature of the defect to a detectable level.FIGURE 4 depicts the test arrangement; it utilizes an induction-heating unit consisting of a power supply and an induction coil suspended below the compact which renders the technique contact-less.

(a) (b)

- 18 -

Figure 4: Contact-less test arrangement of active IR testing unit.

Figure 5 depicts IR images for 2D surface and line profiles (along the dotted line). The data is collected with an IR camera positioned 50cm away (viewed from the side) and operated at a frame rate of 30Hz. The field of view of the 240 by 320 pixel picture is 15cm by 15cm. The total line length of 10cm is subdivided into 180 points (i.e. with a point-to-point resolution of 0.5mm) whereas the thermal pixel intensity is displayed in discrete increments up to a maximum discrete level of 260 (or 460K).

1 cm

0 10 20 30 40 50 6060

80

100

120

140

160

180

200

220

240

260

Dis tanc e along profile

Pix

el in

tens

ity

Figure 5: (a) Initial image from the IR recording of the gear part shown in FIGURE 3 and subjected to an AC current, (b) thermal profile along the dotted line.

(a) (b)

- 19 -

1 cm

0 10 20 30 40 50 60160

170

180

190

200

210

220

230

240

250

260

D is ta nce a lo ng p ro fi le

Pix

el in

tens

ity

Figure 6: (a) Initial image from the IR recording of the defective gear part shown in FIGURE 3 (a), (b) thermal profile along the dotted line.

1 cm

0 10 20 30 40 50 60100

120

140

160

180

200

220

240

260

Dis tance along profile

Pix

el in

tens

ity

Figure 7: (a) Initial image from the IR recording of the defective gear part shown in FIGURE 3 (b), (b) thermal profile along the dotted line. It is clear from Figure that corner defects are not easily discernable. This is also the case for the tooth defect even through an induction heating system is employed. Consequently we have to use more elaborate image processing algorithms and mathematical tools such as applying spatial and temporal derivatives. The examples listed below show the application of the second derivative operated on the thermal signature of several spots over the surface of the part as a function of time. The part is subjected to pulse heating through injecting a high frequency AC current for a short period of time (less than 2seconds).

(a) (b)

Crack signature

(a) (b)

- 20 -

Figure 8: Image from an IR recording of the defective gear part shown in FIGURE 3(b), and selected temperature tracking points.

After establishing the background, Figure depicts the profiles for a defective part with the defect type shown in FIGURE 3 (b).

Figure 9: Thermal signature of the points shown in Figure of a gear with a corner crack. To capture the fast thermal transitions, we compute the second time derivative of the curve shown in Figure 8.

1 cm

30 4

30 6

30 8

31 0

31 2

31 4

31 6

31 8

32 0

32 2

32 4

-4 6 1 6 26 36 4 6 56 6 6

T im e (s)

Tem

pera

ture

(K) S p ot1

S p ot2

S p ot3

S p ot4

S p ot5

S p ot6

S p ot7

S p ot8

S p ot9

- 21 -

-2000

-1500

-1000

-500

0

500

1000

1500

2000

-4 6 16 26 36 46 56 66

Time(s)

2nd

Der

ivat

ive

Spot1

Spot2

Spot3

Spot4

Spot5

Spot6

Spot7

Spot8

Spot9

Figure 10: Second derivative of the temperature response at the locations shown in Figure of a defective part.

Our preliminary thermal measurements on complex gears show that the heat source is of extreme importance. We have concluded that the DC current is not appropriate unless multiple probes are used to maintain uniform current density throughout the volume of the part including its surface. Furthermore, this requirement adds complexity to the instrument and makes its operation difficult to use, thereby limiting its versatility [10]. To remedy to this limitation we devised a system that relays on induction heating to generate the thermal gradient that is later recorded through the IR camera. This method takes advantage of the skin effect, where we control the depth of penetration into the part by controlling the frequency of the source. Another important component of our system is the signal processing. We apply several mathematical tools to extract defect information from the image. Specifically, we analyze the derivative of the thermal profiles of several pre-selected spots on the surface of the compact. Later, we extended this method to include a so-called Laplacian operator to search for surface cracks.

ON-LINE TESTING

Tests at various manufacturing facilities allowed us to establish the stability of our inspection system and its immunity from temperature fluctuations in the plant arising from production equipment such as presses, motors, and sinter furnaces. It became necessary to validate the IR system for the detection of real and commonly observed defects in the process line. Furthermore, our plan to extend the usability of this

- 22 -

method to detect defects regardless of material composition requires careful analysis. Aluminum powder presents a unique challenge: it is a highly reflective material with very low emissivity (0.1 to 0.2) when compared to steel parts with high graphite content where the emissivity is of the order of 0.6. FIGURE 7 shows the green-state steel P/M sample. The compact is a two level gear with 13mm in height by 60mm in diameter and is typically manufactured at a rate of approximately 600 parts per hour.

Figure 11: Picture of a green-state P/M part to be tested at a manufacturing facility.

The following IR images in Figures 12 and 13 represent 2D surface and line profiles (recorded along the dotted line) of parts that are expected to be defect-free. The images are recorded with an IR camera positioned 50cm away (viewed from the side) and operated at a frame rate of 30Hz. The field of view of the 240 by 320 pixel viewing is 15cm by 15cm. The total line length of 10cm is subdivided into 180 points (i.e. with a point-to-point resolution of 0.5mm) whereas the thermal pixel intensity is displayed in discrete increments from a baseline of 0 (or 200K) to 260 (or 460K).

1cm

0 5 1 0 1 5 2 0 2 51 0 0

1 2 0

1 4 0

1 6 0

1 8 0

2 0 0

2 2 0

2 4 0

2 6 0

D is t a n c e a lo n g p ro fi le

Pix

el In

tens

ity

Figure 12: (a) First image from the IR recording of the gear shown in FIGURE 7 at a speed of 0.13m/s, and (b) thermal profile along the dotted line.

(a) (b)

- 23 -

A long IR image sequence of 45 seconds generates 1350 recorded temperature points with an intensity profile depicted in Figure . As expected, as soon as a compact moves past the fixed spatial sensing location, the temperature increases. FIGURE 8, shows the location of the temperature tracking point.

1cm

Figure 13: Temperature monitored at the point shown.

300

305

310

315

320

325

330

335

0 5 10 15 20 25 30 35 40 45Time (s)

Tem

pera

ture

(K)

Figure 14: Temperature (in K) recorded at a fixed spatial location (one spot) over time.

Zooming into the data sequence reported in Figure allows us to conduct a more detailed analysis, as depicted in Figure .

- 24 -

300

305

310

315

320

325

330

335

0 1 2 3 4 5 6 7 8 9 10

Time (s)

Tem

pera

ture

(K)

Figure 15: Temperature (in K) recorded at a fixed spatial location (one spot) over time.

Apart from some small variations, the temperature profiles are reproducible. This is consistent with the fact that the parts are defect-free, an observation that was verified off-line. Therefore, we attribute the thermal fluctuations to instabilities in the process. The IR images in Figure 16 are taken at the same line as shown in the images of Figure 12. However, during the first 20 sec we see defective parts and later, after the process adjustment, the response of defect-free parts. Figure 16 shows an IR image of a defective gear and the associated profile along the dotted line.

1cm

0 2 4 6 8 1 0 1 2 1 4 1 6 1 8 2 01 6 0

1 7 0

1 8 0

1 9 0

2 0 0

2 1 0

2 2 0

2 3 0

2 4 0

2 5 0

2 6 0

D is ta n c e a lo n g p ro file

Pix

el In

tens

ity

It is apparent that the profile shown in Figure 16(b) differs from the profile shown in Figure 12(b); this is a key indication for the presence of a defect. For identifying significant defects similar to what is presented, a simple image subtraction would be sufficient to flag defective compacts. The statistical analysis that shows an entire 45 sec

Figure16: (a) Second image from the IR recording of the gear shown in FIGURE 7, at a speed of 0.13m/s, and (b) thermal profile along the dotted line.

(b) (a)

- 25 -

inspection duration, or 1350 frames, is shown in FIGURE 9. Defects were introduced by changing press settings during press operation. FIGURE 9 depicts the points were the process was modified.

300

305

310

315

320

325

330

3350 5 10 15 20 25 30 35 40 45

Time (s)

Tem

pera

ture

(K)

Figure 17: Temperature (in K) recorded at a fixed spatial location (one spot) over time.

A detailed investigation of the data sequence reported in FIGURE 917 allows us to conduct a more thorough analysis, as depicted in FIGURE 18.

300

305

310

315

320

325

330

335

0 1 2 3 4 5 6 7 8 9 10

Time (s)

Tem

pera

ture

(K)

Figure 18: Zoomed-in temperature (in K) recorded at a fixed spot location.

Start defects End defects

- 26 -

As can be seen by directly comparing 17 with FIGURE 18, several parts are defective. As a result, this methodology has the potential of being a very simple, yet reliable methodology that allows us to identify defective parts in an on-line setting. Additional tests that involve aluminum parts are discussed below. Figure depicts the green-state aluminum powder part. The compact is 20mm in height by 50mm in length and 15mm in width, and compacted at a high density (part parameters and material composition are proprietary to the manufacturer); it is manufactured at a rate of approximately 900 parts per hour. Aluminum compacts provide unique challenges of high emissivity and high cooling rate. They consequently require special attention with regard to viewing angles and part access in close proximity to the press. It is also important to ensure temperature equilibrium during the testing phase.

Figure 19: Aluminum powder green-state compact. Due to access restrictions we were unable to image the green-state compacts directly as they exited the press. This unfortunately precluded our ability to inspect the parts while they are at a high temperature setting. For the testing, we use the same procedure as discussed above: a fixed sensing point is selected in the process line as shown in Figure . We can then monitor its temperature behavior over time.

1cm

Figure 20: Identifying a point (cross) in the process line for thermal recording.

- 27 -

As discussed before, we can now examine defect-free and defective parts with artificially induced hairline cracks across the curved section of the compact. First, we report the statistical data for the defect-free parts in Figure .

310

312

314

316

318

320

322

324

326

0 5 10 15 20 25 30 35 40

Time (s)

Tem

pera

ture

(K)

Figure 21: Temperature (in K) recorded at a fixed spatial location (one spot) for defect-free aluminum compacts. Defective parts are next shown in Figure . In the thermal response it is difficult to observe the effects of cracks in the surface temperature profile. This is mainly due to the fact that the compacts do not reaching thermal equilibrium. However, we can still discern a baseline variation which could be indicative of process changes that may have caused the defects.

- 28 -

310

312

314

316

318

320

322

324

326

328

0 5 10 15 20 25 30 35 40

Time (s)

Tem

pera

ture

(K)

Figure 22: Temperature (in K) recorded at a fixed spatial location (one spot) for defective aluminum compacts.

CONCLUSIONS

The data collected at manufacturing facilities confirm the fact that the IR testing methodology is easy to implement in a manufacturing setting and poses little intrusion to the fabrication process. Preliminary testing suggests that the IR imaging system is relatively immune to changes in the background noise generated by the manufacturing equipment. Our very early data collection supports the fact that we can detect defects in a real time environment. Therefore, we are confident that this system will ultimately be capable of performing a 100 percent quality assessment of green-state P/M compacts by providing real-time operator feedback. Additional of-line information about defect parameters can also be extracted for engineering purposes and for process calibration. Besides fabrication immunity, the IR testing methodology appears sufficiently robust to handle different material compositions. In particular, we were able to test aluminum parts that have the special characteristics of high reflectivity and high cooling rate. Thus far, all of our signal processing and data analysis is performed off-line. For a fully manufacturing-compliant system it is important to combine these steps into a rapid data collection and processing environment to provide real-time feedback.

- 29 -

REFERENCES

13. Leuenberger, G. "Electrostatic Density Measurement in Green-State P/M Parts" PhD thesis, ECE Department, Worcester Polytechnic Institute 2003.

14. Fei, M. "Electromagnetic Inspection, Infrared Visualization and Image Processing

Techniques for Non Metallic inclusions in Molten Aluminum" Master Thesis, ECE Department, Worcester Polytechnic Institute 2002.

15. Maldag, X.P.V. "Theory and Practice of Infrared Technology for Nondestructive

Testing" John Wiley & Sons Inc. 2001.

16. Morse, P.M, Feshbach, H. “Methods of Theoretical Physics” McGraw-Hill Book Company, Inc. 1953.

17. Burnay, S. G., Williams, T. L., Jones C. H. "Applications of Thermal Imaging"

10.P Publishing 1988.

18. Carslaw, H., Jaeger, J. "Conduction of Heat in Solids" Second Edition, Oxford University Press 1959.

19. Incropera, F.P., DeWitt, D.P. "Fundamentals of Heat and Mass Transfer" 4th

edition, John Wiley & Sons, New York 1996.

20. German, R.M. "Powder Metallurgy Science" Metal Powder Industries Federation, Princeton, New Jersey, 1984.

21. Kraus, J.,D. "Electromagnetics" McGraw-Hill Book Company, Inc. 1953.

22. Bruhat,G. "Cours De Physique Général: ELECTRICITE" Septième Edition,

Masson & Cie . 1959.

23. Ringermacher, H.I., Howard, D.R. and Gilmore, R.S., "Discriminating Porosity in Composites Using Thermal Depth Imaging" CP 615, Review of Quantitative Nondestructive Evaluation, Vol. 21, ed. by Thompson and D.E Chimenti. American Institute of Physics. 2002.

24. Han, X., Favro L.D., and Thomas, R.L., "Recent Developments in Thermosonic

Crack Detection" CP 615, Review of Quantitative Nondestructive Evaluation, Vol. 21, ed. by Thompson and D.E Chimenti. American Institute of Physics. 2002.

25. Sun, I.G "Analysis of Quantitative Measurements of Defects by Pulsed

Thermography Imaging" CP 615, Review of Quantitative Nondestructive Evaluation, Vol. 21, ed. by Thompson and D.E Chimenti. American Institute of Physics. 2002.

- 30 -

26. Hausseker, H.W “Simultaneous Estimation of Optical Flow and Heat Transport in Infrared Image Sequences” IEEE Conference on Computer Vision and Pattern Recognition 2000.

27. Hermann A. Haus/James R. Melcher “Electromagnetic Fields and Energy

“Prentice-Hall Inc., New Jersey. 1989.

28. Powder Metallurgy Research Center (PMRC), Metal Processing Institute, Worcester Polytechnic Institute, fall meeting, Oct. 20, 2004.