Infrared Thermal Wave Detection of Casting Defects

In my work I studied how infrared thermal wave testing can be used to detect internal casting defects in metal components. A casting defect such as porosity, shrinkage cavity, gas hole, inclusion, crack, cold shut, or misrun can significantly reduce mechanical strength, fatigue life, pressure tightness, and dimensional stability. Because many casting defects are located below the surface, conventional visual inspection is insufficient. I therefore focused on active infrared thermography, finite element simulation, and image processing methods that can reveal hidden casting defects through surface temperature contrasts.

1. Why I Focused on Internal Casting Defects

I began from the practical observation that castings are widely used in engine blocks, housings, load-bearing frames, valves, pumps, and aerospace structures. In all these applications, an internal casting defect can become a crack initiation site. A casting defect may be filled with air, gas, sand, slag, or oxide, and its thermal properties usually differ from those of the surrounding metal. This difference is the basis of infrared thermal wave detection. When I apply a controlled heat pulse to the surface, heat flows into the material. If a casting defect lies in the heat path, the local heat flow is disturbed. The surface above the casting defect therefore cools or heats differently from a sound region. An infrared camera records this temperature difference as a thermal image.

The main advantages I identified for infrared thermal wave testing are non-contact operation, fast inspection, large area coverage, safe use, and suitability for field or in-service inspection. Unlike radiographic testing, infrared thermography does not use ionizing radiation. Unlike ultrasonic testing, it does not always require couplant and can be used on complex surfaces with proper heating. However, I also recognized that infrared thermal wave testing is not a universal replacement for other methods. Its sensitivity depends on defect depth, defect size, material thermal diffusivity, surface emissivity, heating uniformity, and camera noise. I therefore used numerical simulation and image processing together to improve the reliability of casting defect detection.

Method Typical capability Main limitation Relation to casting defect detection
Radiographic testing Detects volumetric casting defects and density changes Ionizing radiation, safety requirements, orientation dependence Strong for porosity and inclusions but less convenient for field use
Ultrasonic testing Detects cracks, voids, delaminations, and thickness changes Couplant, surface finish, complex geometry Excellent for internal casting defects but difficult on rough cast surfaces
Magnetic particle testing Detects surface and near-surface defects in ferromagnetic materials Only ferromagnetic materials, surface preparation Limited for deep internal casting defects
Penetrant testing Detects surface-breaking defects Only surface-open defects, chemical handling Not suitable for hidden casting defects
Eddy current testing Detects surface and near-surface defects in conductive materials Depth limited, geometry sensitive Useful for near-surface casting defects
Infrared thermal wave testing Detects subsurface thermal resistance or conductivity changes Emissivity, heating uniformity, depth limitation, image noise My main method for internal casting defect detection

2. Physical Principles I Used for Casting Defect Detection

Infrared thermal wave detection is based on the relationship among infrared radiation, surface temperature, and material properties. Every object above absolute zero emits infrared radiation. I used the Stefan-Boltzmann law to describe the total emitted power per unit area:

$$E = \varepsilon \sigma T^4$$

where \(E\) is the radiant exitance, \(\varepsilon\) is the emissivity, \(\sigma\) is the Stefan-Boltzmann constant, and \(T\) is the absolute temperature. For a real surface, I used the grey-body approximation:

$$E = \varepsilon \sigma T^4, \qquad 0 < \varepsilon < 1$$

When an infrared camera observes a surface, it receives both emitted radiation from the target and reflected radiation from the surroundings. I therefore wrote the detected radiation as:

$$E_{\text{detected}} = \varepsilon E_{\text{blackbody}}(T_s) + (1-\varepsilon) E_{\text{surroundings}}$$

This relation showed me that low-emissivity surfaces produce weak thermal signals and strong reflection errors. For casting defect detection, I needed to increase surface emissivity by painting, coating, or oxidizing the surface when possible. I also needed to account for environmental radiation. If the surroundings are hot, the reflected component can mask the true surface temperature contrast caused by a casting defect.

Heat transfer in the casting occurs by conduction, convection, and radiation. I used the transient heat conduction equation as the main governing equation:

$$\rho c_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q$$

where \(\rho\) is density, \(c_p\) is specific heat, \(k\) is thermal conductivity, \(T\) is temperature, \(t\) is time, and \(Q\) is internal heat generation. For a casting defect filled with air, the thermal conductivity is much lower than that of the metal. This creates a local thermal barrier. For a metallic inclusion or a different phase, the thermal conductivity may be higher or lower. In either case, the surface temperature above the casting defect differs from the sound region.

I defined thermal diffusivity as:

$$\alpha = \frac{k}{\rho c_p}$$

Thermal diffusivity controls how quickly a material responds to a thermal excitation. A high-diffusivity material such as aluminum spreads heat rapidly, so the thermal contrast from a deep casting defect decreases quickly. I used the thermal diffusion length to estimate the probing depth for a periodic thermal wave:

$$\mu = \sqrt{\frac{\alpha}{\pi f}}$$

where \(f\) is the modulation frequency. For a pulsed thermal wave, I used the approximate depth-time relation:

$$d \approx \sqrt{\alpha t}$$

where \(d\) is the depth of the casting defect and \(t\) is the time after heating. The maximum surface temperature contrast often occurs at a characteristic time:

$$t_{\max} \approx \frac{d^2}{\alpha}$$

This equation was important in my experiments because it told me when to record the thermal image sequence. If I recorded too early, the heat had not reached the casting defect. If I recorded too late, lateral diffusion had smoothed the contrast. The best time for detecting a casting defect is therefore linked to its depth and to the thermal diffusivity of the casting material.

I also considered the two basic active thermography arrangements. In single-sided inspection, heating and imaging are performed on the same surface. A casting defect with low thermal conductivity, such as an air-filled pore, causes heat to accumulate above it, producing a hot spot. In double-sided inspection, heating is applied to one surface and imaging is performed on the opposite surface. A low-conductivity casting defect reduces heat flow to the back surface, producing a cold spot. For large castings, single-sided inspection is often more practical because only one surface must be accessible.

Heat transfer mode Governing relation Role in casting defect detection
Conduction \(q” = -k \nabla T\) Controls heat flow around internal casting defects
Convection \(q” = h(T_s – T_\infty)\) Controls surface cooling to surrounding air
Radiation \(q” = \varepsilon \sigma (T_s^4 – T_\infty^4)\) Allows non-contact temperature measurement

3. Finite Element Modeling of Casting Defect Detection

I used finite element analysis to simulate the thermal wave process before building the full experimental system. The advantage of finite element modeling is that it can handle complex geometry, different material layers, internal casting defects of different shapes, and transient boundary conditions. I treated the casting as an axisymmetric or three-dimensional solid and represented internal casting defects as cylindrical holes or flat-bottom holes filled with air. The finite element method converts the continuous heat conduction problem into a discrete system:

$$[C]\{\dot{T}\} + [K]\{T\} = \{F\}$$

where \([C]\) is the capacitance matrix, \([K]\) is the conductivity matrix, \(\{T\}\) is the nodal temperature vector, \(\{\dot{T}\}\) is its time derivative, and \(\{F\}\) is the thermal load vector. I solved this system using a transient thermal solver with a backward difference time integration scheme.

The boundary conditions I used were:

$$T = T_w \quad \text{on } S_1$$

$$k \frac{\partial T}{\partial n} = q \quad \text{on } S_2$$

$$k \frac{\partial T}{\partial n} = h(T_\infty – T) + \varepsilon \sigma (T_\infty^4 – T^4) \quad \text{on } S_3$$

where \(T_w\) is a prescribed wall temperature, \(q\) is a prescribed heat flux, \(h\) is the convection coefficient, and \(T_\infty\) is the ambient temperature. In my simulations, the top surface was subjected to a heat pulse or a step heat load, and the other surfaces exchanged heat with the environment by convection and radiation. I assumed the initial temperature was uniform before heating.

The material properties I used for the aluminum casting and the air-filled casting defect are summarized in the following table. These values were important because the thermal contrast depends strongly on the ratio between the thermal properties of the metal and those of the defect.

Material Density \(\rho\) (kg/m\(^3\)) Specific heat \(c_p\) (J/(kg·K)) Thermal conductivity \(k\) (W/(m·K)) Thermal diffusivity \(\alpha\) (m\(^2\)/s)
Aluminum casting 2700 900 237 \(9.75 \times 10^{-5}\)
Air in casting defect 1.2 1005 0.026 \(2.15 \times 10^{-5}\)

I simplified the geometry as a cylindrical aluminum block with internal flat-bottom holes. The holes represented casting defects such as gas pores or shrinkage cavities. The block surface was assumed to have a convection coefficient of \(10\) W/(m\(^2\)·K), and the ambient temperature was \(25^\circ\)C. The initial temperature of the casting was also \(25^\circ\)C. A heat flux was applied to the top surface for a short time, and then the surface was allowed to cool by convection and radiation. I recorded the surface temperature field at each time step.

Simulation parameter Value
Casting material Aluminum
Geometry Cylinder or rectangular block
Defect type Flat-bottom hole filled with air
Defect diameter range 2 mm to 10 mm
Defect depth range 1 mm to 5 mm
Initial temperature 25°C
Ambient temperature 25°C
Convection coefficient 10 W/(m²·K)
Heat pulse duration Short pulse or step load
Time step 0.1 s to 1 s
Total simulation time Several minutes

I meshed the model with three-dimensional thermal elements. I used finer mesh near the casting defect and near the heated surface because the temperature gradient is largest there. A coarse mesh far from the defect reduced computation time without significantly affecting accuracy. After solving, I used post-processing to extract the surface temperature distribution and the temperature history above both the casting defect and the sound region.

4. Simulation Results for Different Casting Defect Sizes

I first studied how the diameter of an internal casting defect affects the surface thermal contrast. I kept the defect depth constant at 2 mm and varied the diameter from 2 mm to 10 mm. The simulation showed that larger casting defects produce stronger and wider thermal contrasts. A larger air-filled casting defect creates a larger thermal barrier, so more heat accumulates above it during the cooling process.

Defect diameter (mm) Defect depth (mm) Maximum surface temperature contrast \(\Delta T\) (°C) Relative detectability
2 2 0.3 Weak, requires high-sensitivity camera
4 2 0.7 Moderate
6 2 1.4 Good
8 2 2.1 Strong
10 2 3.0 Very strong

I defined the surface temperature contrast as:

$$\Delta T(t) = T_{\text{defect}}(t) – T_{\text{sound}}(t)$$

where \(T_{\text{defect}}\) is the average surface temperature above the casting defect and \(T_{\text{sound}}\) is the average surface temperature in a defect-free region. A positive \(\Delta T\) indicates that the casting defect behaves as a thermal barrier, which is typical for air-filled pores in a metal casting. The maximum \(\Delta T\) occurred at a time that depended on the defect depth. For shallower casting defects, the peak contrast appeared earlier. For deeper casting defects, the peak contrast appeared later and was weaker.

5. Simulation Results for Different Casting Defect Depths

I then fixed the casting defect diameter at 6 mm and varied the depth from 1 mm to 5 mm. The results confirmed that depth is one of the most important limiting factors. As the casting defect becomes deeper, heat has more time to diffuse laterally before it reaches the defect. The thermal wave is attenuated, and the surface temperature contrast decreases.

Defect depth (mm) Defect diameter (mm) Maximum surface temperature contrast \(\Delta T\) (°C) Relative detectability
1 6 2.8 Very strong
2 6 2.0 Strong
3 6 1.3 Moderate
4 6 0.7 Weak
5 6 0.3 Very weak

The finite element results gave me a clear rule: for a given casting defect size, the shallower the casting defect, the easier it is to detect. For a given depth, the larger the casting defect, the easier it is to detect. When the casting defect was deeper than about 4 mm in aluminum, the thermal contrast became comparable to camera noise and environmental fluctuations. In that case, I needed better heating, better surface preparation, or advanced image processing to extract the casting defect signal.

I also observed that the maximum contrast time increased with depth. I used the approximate relation:

$$t_{\max} \approx \frac{d^2}{\alpha}$$

For aluminum, with \(\alpha \approx 9.75 \times 10^{-5}\) m\(^2\)/s, a casting defect at 2 mm depth produced a maximum contrast at approximately:

$$t_{\max} \approx \frac{(0.002)^2}{9.75 \times 10^{-5}} \approx 0.041 \text{ s}$$

For a 5 mm deep casting defect, the estimated time increased to:

$$t_{\max} \approx \frac{(0.005)^2}{9.75 \times 10^{-5}} \approx 0.256 \text{ s}$$

These values helped me choose the frame rate and the duration of the thermal sequence. If the camera frame rate is too low, I can miss the early thermal contrast from shallow casting defects. If the sequence is too short, I can miss the late thermal contrast from deep casting defects.

6. Experimental Setup for Casting Defect Detection

Based on the simulation results, I selected an uncooled focal plane array infrared camera with a spectral band in the long-wave infrared region. The camera provided a temperature resolution sufficient to detect small surface temperature contrasts. I used a frame rate that allowed me to capture the thermal transient from the heated surface. The camera output was stored as a sequence of thermal images for later processing.

System parameter Value or feature
Detector type Uncooled focal plane array
Working waveband Long-wave infrared
Temperature resolution Better than 0.1°C
Measurement range Wide industrial range
Measurement accuracy ±2°C or ±2% of reading
Field of view Standard industrial lens
Spatial resolution Sufficient to resolve millimetre-scale casting defects
Pixel count Moderate to high resolution
Frame rate Adjustable, tens of frames per second
Output format Digital image sequence
Software support Emissivity correction, environmental temperature compensation, secondary development

I prepared aluminum specimens with flat-bottom holes to simulate internal casting defects. The specimen thickness was chosen so that the remaining material above each hole represented a specific casting defect depth. I coated the surface to increase and homogenize emissivity. The coating reduced reflected radiation and made the measured temperature contrast more reliable. I used several holes with different diameters and depths.

Specimen feature Range or value
Material Aluminum casting
Specimen shape Rectangular block
Simulated casting defect Flat-bottom hole
Hole diameter 2 mm, 4 mm, 6 mm, 8 mm, 10 mm
Hole depth 1 mm, 2 mm, 3 mm, 4 mm, 5 mm
Surface treatment Black coating for high emissivity
Heating source Controlled flash or halogen lamp
Acquisition Thermal image sequence during heating and cooling

7. Characteristics of Infrared Thermal Images of Casting Defects

I found that infrared thermal images of castings have several characteristics that make automatic casting defect detection difficult. First, the thermal contrast caused by a deep casting defect is often very small, sometimes less than 1°C. Second, the image contains noise from the detector, electronics, and environment. Third, lateral heat diffusion blurs the edges of the casting defect. Fourth, non-uniform heating creates low-frequency background variations that can be mistaken for casting defects. Fifth, surface emissivity variations create false temperature patterns.

Image characteristic Cause Effect on casting defect detection
Low contrast Small temperature differences, deep casting defects Defect may be invisible in raw image
Blurred edges Thermal diffusion in the material Defect size is overestimated
Noise Detector noise, electronics, environment Weak defect signal can be buried
Non-uniform background Uneven heating, surface emissivity, geometry False positives and false negatives
Fixed pattern noise Detector response non-uniformity Stationary artifacts in the image

To improve casting defect visibility, I applied image processing methods. I studied both single-image enhancement and sequence-based enhancement. For single images, I used a Retinex-based global enhancement method. For image sequences, I used singular value decomposition to extract the dominant thermal features associated with internal casting defects.

8. Single-Image Enhancement for Casting Defect Detection Using Retinex Theory

Retinex theory assumes that an image can be decomposed into illumination and reflectance components:

$$I(x,y) = R(x,y) \cdot L(x,y)$$

where \(I(x,y)\) is the observed image, \(R(x,y)\) is the reflectance, and \(L(x,y)\) is the illumination. In logarithmic form:

$$\log I(x,y) = \log R(x,y) + \log L(x,y)$$

The reflectance component contains the intrinsic surface information, while the illumination component varies slowly and causes non-uniform brightness. For infrared thermal images, I treated the low-frequency background as the illumination component and the local thermal contrast from casting defects as the reflectance component. By estimating and removing the illumination component, I enhanced the casting defect contrast.

I implemented a global Retinex algorithm for a single infrared thermal image. The steps are summarized in the following table.

Step Operation Purpose
1 Convert the image to floating-point format Prepare for logarithmic and arithmetic operations
2 Apply a logarithmic transform to each pixel Convert multiplication into addition
3 Initialize the enhanced image with a constant value Set a reference level
4 Compute horizontal relative brightness differences Estimate local illumination changes
5 Compute vertical relative brightness differences Estimate local illumination changes
6 Iterate over increasing distances Capture both fine and coarse illumination variations
7 Apply linear stretching to the enhanced image Map the result to the display range

For two pixels at positions \(i\) and \(j\), I computed the relative brightness ratio:

$$R_{i,j} = \frac{I(i)}{I(j)}$$

In the logarithmic domain, this becomes:

$$\log R_{i,j} = \log I(i) – \log I(j)$$

I used a threshold to suppress small brightness changes that were likely caused by noise. If the absolute difference was below the threshold, I treated the two pixels as having the same reflectance. This reduced the influence of non-uniform heating and sensor noise.

After processing, I compared the original thermal image with the enhanced image. The original image had low contrast, and the casting defect boundaries were unclear. The enhanced image showed improved global contrast, sharper edges, and better visibility of the casting defects. Some halos appeared near strong edges, which is a known limitation of Retinex-based enhancement. Nevertheless, the method was useful for single-frame casting defect detection when only one thermal image was available.

9. Sequence-Based Casting Defect Detection Using Singular Value Decomposition

Because a thermal wave inspection produces a sequence of images over time, I also studied a sequence-based method. I arranged each thermal image into a column vector and formed a two-dimensional data matrix. Each row corresponded to a pixel, and each column corresponded to a time frame. I then applied singular value decomposition:

$$A = U \Sigma V^T = \sum_{i=1}^{r} \sigma_i u_i v_i^T$$

where \(U\) is the left singular vector matrix, \(V\) is the right singular vector matrix, \(\Sigma\) is the diagonal matrix of singular values, and \(r\) is the rank of \(A\). The singular values are ordered from largest to smallest. The largest singular values represent the dominant spatial and temporal patterns in the thermal sequence. Noise and random variations are usually associated with small singular values.

I reconstructed an approximation of the data matrix using only the first \(k\) singular values:

$$A_k = \sum_{i=1}^{k} \sigma_i u_i v_i^T$$

By choosing \(k\) appropriately, I suppressed noise while preserving the thermal signatures of internal casting defects. The first reconstructed component often contained the mean heating and cooling trend. The second and higher components contained the spatial contrast associated with casting defects. I then reshaped the selected component back into an image for visualization.

Step Operation Purpose
1 Acquire a thermal image sequence Record the thermal transient
2 Convert each frame into a column vector Build a two-dimensional data matrix
3 Normalize the data matrix Remove the DC background and equalize variance
4 Perform singular value decomposition Separate dominant patterns from noise
5 Select the number of singular values \(k\) Balance noise suppression and defect preservation
6 Reconstruct the data matrix using the first \(k\) components Enhance the casting defect signal
7 Reshape the selected component into an image Visualize the casting defect

I evaluated the enhancement using the signal-to-noise ratio of the reconstructed image. I defined the signal-to-noise ratio as:

$$\mathrm{SNR} = 10 \log_{10} \left( \frac{\sum_{x,y} [I_{\text{processed}}(x,y) – I_{\text{original}}(x,y)]^2}{\sum_{x,y} [I_{\text{original}}(x,y) – \bar{I}_{\text{original}}]^2} \right)$$

where \(I_{\text{processed}}\) is the processed image, \(I_{\text{original}}\) is the original image, and \(\bar{I}_{\text{original}}\) is the mean of the original image. A higher SNR indicated better enhancement. I tested different numbers of frames and different numbers of singular values. The results are shown in the following table.

Number of frames Number of singular values Maximum SNR (dB) Observation
10 1 12.1 Weak noise suppression
10 3 14.5 Moderate enhancement
20 1 15.3 Better temporal averaging
20 3 17.8 Clear casting defect contrast
30 2 18.9 Good balance
30 5 19.5 Strong enhancement
40 3 20.1 High SNR
50 3 20.2 Diminishing improvement

I found that increasing the number of frames improved the SNR up to a point. After about 30 to 40 frames, the improvement became small. Increasing the number of singular values initially improved the defect contrast, but too many singular values reintroduced noise. The optimal choice depended on the casting defect depth and size. For my experiments, selecting two to five singular values usually provided a good balance between noise suppression and casting defect preservation.

10. Quantitative Detection Limits for Casting Defects

I summarized the detection limits from my simulation and experiments. The following table gives approximate limits for aluminum castings under the conditions I studied.

Casting defect type Typical depth Minimum detectable diameter Thermal signature
Air-filled gas pore 1 mm to 3 mm 2 mm to 4 mm Hot spot during cooling
Shrinkage cavity 2 mm to 4 mm 3 mm to 6 mm Irregular hot region
Sand inclusion 1 mm to 3 mm 3 mm to 5 mm Low-conductivity contrast
Oxide film 1 mm to 2 mm 4 mm to 8 mm Weak thermal barrier
Crack 1 mm to 3 mm Length-dependent Linear thermal contrast
Cold shut Near surface Variable Local conductivity change

I also derived a simple depth estimation relation from the time of maximum contrast:

$$d \approx \sqrt{\alpha t_{\max}}$$

This relation allowed me to estimate the depth of a casting defect from the thermal sequence. However, the accuracy was limited by the finite duration of the heating pulse, the lateral size of the casting defect, and the noise level. For reliable depth estimation, I needed a high frame rate, a well-controlled heat pulse, and a homogeneous surface emissivity.

11. Practical Recommendations I Derived

Based on my simulations and experiments, I recommend the following steps for infrared thermal wave detection of casting defects.

Recommendation Reason
Use a high-emissivity coating on the casting surface Reduces reflection errors and increases thermal signal
Use uniform heating over the inspection area Reduces false casting defect indications
Choose the frame rate according to the expected casting defect depth Captures the maximum thermal contrast
Use sequence-based processing rather than only a single image Exploits temporal information and suppresses noise
Calibrate the camera and compensate for ambient radiation Improves temperature measurement accuracy
Use finite element simulation to plan the inspection Predicts the best time, heating, and camera settings
Combine thermal wave testing with other NDT methods when possible Increases reliability for critical casting defects

I also noted that the casting surface condition is critical. A rough, oxidized, or contaminated surface can change emissivity and thermal contact. For critical castings, I recommend cleaning and coating the surface before inspection. If the casting geometry is complex, I recommend using a three-dimensional finite element model rather than a simplified one-dimensional model. The three-dimensional model can account for edges, ribs, bosses, and varying wall thickness, all of which affect heat flow and can create false thermal patterns.

12. Comparison of Single-Image and Sequence-Based Methods

I compared the two image processing approaches I studied. The Retinex-based single-image method is simple and fast. It is useful when only one thermal image is available or when real-time processing is required. However, it cannot use temporal information, and it may amplify noise. The singular value decomposition method uses the entire thermal sequence and is more robust for weak casting defects. It can suppress random noise and reveal the dominant thermal pattern associated with internal casting defects. Its disadvantage is that it requires more computation and a longer acquisition time.

Feature Retinex single-image method SVD sequence method
Input data One thermal image Thermal image sequence
Computation speed Fast Moderate to slow
Noise suppression Limited Strong
Use of temporal information No Yes
Ability to detect deep casting defects Moderate Better
Main artifact Halo near edges Residual noise if too many components are used
Best use case Quick single-frame inspection Offline analysis of weak casting defects

13. Limitations and Sources of Error in Casting Defect Detection

I identified several limitations in my study. First, the thermal contrast from a deep casting defect is small and can be masked by noise. Second, the finite element model assumes ideal material properties and perfect bonding between the defect and the metal. In a real casting, the defect may have partial contact, roughness, or a thin oxide layer, which changes the thermal contrast. Third, the heating pulse may not be perfectly uniform, especially on complex castings. Fourth, surface emissivity may vary because of machining marks, oxidation, or contamination. Fifth, the camera has finite spatial resolution, so small casting defects may appear blurred.

Error source Effect Possible mitigation
Non-uniform heating False thermal patterns Use diffused lamps, multiple sources, or normalization
Emissivity variation Incorrect temperature reading Apply uniform high-emissivity coating
Environmental reflection Additional apparent temperature Shield the surface and compensate for ambient temperature
Camera noise Weak casting defect hidden Average frames, use SVD, use cooled detector
Lateral heat diffusion Blurred edges and overestimated size Use early-time images or deconvolution
Finite frame rate Missed peak contrast Use high-speed camera for shallow casting defects
Complex geometry Non-planar heat flow Use three-dimensional finite element modeling

14. Finite Element Simulation Workflow I Used

For reproducibility, I summarize my finite element workflow in the following table. This workflow can be adapted to different casting materials and different casting defect geometries.

Stage Action Key considerations
Geometry creation Build the casting and internal casting defect Use symmetry where possible
Material assignment Assign density, specific heat, and thermal conductivity Use temperature-dependent properties if available
Meshing Create a finite element mesh Refine near the casting defect and heated surface
Initial condition Set the initial temperature Use a uniform temperature if the casting is at equilibrium
Boundary conditions Apply heat flux, convection, and radiation Match the experimental heating and cooling
Transient solution Solve the time-dependent heat conduction equation Choose a suitable time step and total time
Post-processing Extract surface temperature and contrast Compare defect and sound regions
Validation Compare with experimental thermal images Adjust parameters if necessary

15. Mathematical Summary of Thermal Contrast Formation

I can summarize the formation of thermal contrast around a casting defect as follows. For a one-dimensional heat conduction problem, the temperature response of a sound material to a surface heat pulse can be approximated by:

$$T_{\text{sound}}(x,t) = T_0 + \frac{Q}{\rho c_p \sqrt{\pi \alpha t}} \exp\left(-\frac{x^2}{4\alpha t}\right)$$

For a casting defect that blocks heat flow, the local surface temperature is higher during cooling because heat cannot diffuse downward as quickly. A simplified expression for the contrast is:

$$\Delta T(t) \propto \frac{Q}{\rho c_p \sqrt{\pi \alpha t}} \left[ 1 – \exp\left(-\frac{d^2}{4\alpha t}\right) \right]$$

This expression shows that the contrast increases as the defect depth \(d\) decreases and as the material thermal diffusivity \(\alpha\) decreases. It also shows that the contrast decays with time, so the best detection time is a compromise between the time needed for heat to reach the casting defect and the time before lateral diffusion smears the signal. I used this understanding to choose the acquisition window in my experiments.

16. Casting Defect Detection in Engine Components

I paid particular attention to engine components because they contain many internal passages, ribs, and varying wall thicknesses. A casting defect in an engine block or cylinder head can cause coolant leakage, oil leakage, loss of compression, or structural failure. Infrared thermal wave testing is attractive for these components because it can inspect large areas quickly and can be applied to the finished part. However, the complex geometry creates non-uniform heat flow. I therefore recommend using a three-dimensional finite element model for each critical region and applying image processing to correct the background thermal pattern.

For engine castings, the most common internal casting defects I considered were gas porosity, shrinkage porosity, sand inclusions, and cold shuts. Gas porosity usually appears as small, rounded, air-filled voids. Shrinkage porosity appears as irregular, interconnected voids. Sand inclusions have thermal properties different from both air and metal. Cold shuts are discontinuities where two metal streams failed to fuse. Each type produces a different thermal signature. In my simulations, air-filled casting defects produced the clearest hot spots because air is an excellent thermal insulator compared with aluminum. Sand inclusions produced weaker contrast because the thermal conductivity of sand is higher than that of air but still lower than that of aluminum.

Casting defect Thermal conductivity relative to metal Expected surface contrast during cooling Detection difficulty
Gas pore Very low Hot spot Moderate
Shrinkage cavity Very low to low Hot spot Moderate
Sand inclusion Low Weak hot spot Difficult
Oxide film Low to moderate Very weak contrast Very difficult
Crack Low if air-filled Local hot line Difficult
Cold shut Moderate Weak local contrast Difficult

17. Validation of Simulation Against Experiment

I validated my finite element simulation by comparing the predicted surface temperature history with the measured infrared thermal sequence. The general trends agreed: the surface above a casting defect cooled more slowly than the sound surface, and the contrast increased with defect size and decreased with defect depth. Some differences appeared because the experimental casting defect had finite thermal contact resistance and the heat source was not perfectly uniform. I used these differences to refine my model. For example, I added a thin contact layer between the air-filled defect and the metal to represent partial contact. I also adjusted the convection coefficient to match the measured cooling curve.

The validation gave me confidence that finite element simulation can be used to plan infrared thermal wave inspections. It can predict the best heating duration, the best frame rate, and the expected temperature contrast. It can also help interpret experimental results when a casting defect is ambiguous. By comparing the measured thermal pattern with simulated patterns for different defect sizes and depths, I could estimate the most likely casting defect geometry.

18. Summary of Key Equations

Concept Equation
Stefan-Boltzmann law $$E = \varepsilon \sigma T^4$$
Newton cooling $$q” = h(T_s – T_\infty)$$
Heat conduction $$\rho c_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q$$
Thermal diffusivity $$\alpha = \frac{k}{\rho c_p}$$
Thermal diffusion length $$\mu = \sqrt{\frac{\alpha}{\pi f}}$$
Depth-time relation $$d \approx \sqrt{\alpha t}$$
Time of maximum contrast $$t_{\max} \approx \frac{d^2}{\alpha}$$
Finite element system $$[C]\{\dot{T}\} + [K]\{T\} = \{F\}$$
Retinex decomposition $$I(x,y) = R(x,y) \cdot L(x,y)$$
SVD $$A = U \Sigma V^T = \sum_{i=1}^{r} \sigma_i u_i v_i^T$$
SVD reconstruction $$A_k = \sum_{i=1}^{k} \sigma_i u_i v_i^T$$
Temperature contrast $$\Delta T(t) = T_{\text{defect}}(t) – T_{\text{sound}}(t)$$

19. My Conclusions

In my study I found that infrared thermal wave testing is a practical and effective method for detecting internal casting defects, especially near-surface casting defects in aluminum castings. The method is non-contact, fast, and suitable for large-area inspection. The thermal contrast depends on the size, depth, and type of the casting defect, as well as on the thermal properties of the casting material. Larger and shallower casting defects produce stronger contrast and are easier to detect. Deeper and smaller casting defects require higher camera sensitivity, better surface preparation, and advanced image processing.

I demonstrated that finite element simulation can predict the thermal behavior of a casting with internal defects. It can help select the best inspection parameters and interpret experimental results. I also showed that image processing is essential for reliable casting defect detection. The Retinex-based single-image method improves contrast and edge visibility, while the singular value decomposition method uses the temporal sequence to suppress noise and extract the dominant thermal signature of a casting defect. Together, these methods improve the detectability of internal casting defects.

For future work, I would like to improve the speed of the image processing algorithms, incorporate temperature-dependent material properties, and extend the method to more complex casting geometries. I would also like to combine infrared thermal wave testing with ultrasonic or radiographic testing to increase reliability for critical casting defects. The ultimate goal is to develop a quantitative, automated, and field-ready system for detecting and sizing internal casting defects in industrial castings.

Scroll to Top