An Image Analysis for Designing an Optimal Stirrer in Metal Matrix Composites Manufacturing

1. INTRODUCTION

As energy shortages and environmental pollution risks increase, the demand for stronger, lighter, and more environmentally friendly materials continues to grow. Aluminum alloy ceramic matrix composites are extensively used across various industries and are expected to expand further as efforts to protect the environment and reduce fossil fuel usage intensify. Stir casting, also known as the vortex process, is widely recognized as an economical method for producing metal matrix composites. However, achieving mixing homogeneity remains a significant challenge, heavily influenced by stirrer design. In this study, the authors isolated the effect of stirrer design by keeping other key variables – stirring time, stirring speed, and stirrer position – constant, thereby eliminating potential interference or interaction effects from these factors. These parameters were fixed as follows: Stirring time at 60 s, stirring speed at 240 rpm, and stirrer position at two-thirds of the crucible height. This approach allowed the exclusive evaluation of the stirrer design’s function. Fig. 1 provides a flowchart summarizing the research methodology used to achieve the study’s objectives.

Fig. 1. Diagram of the research work.

Using image processing to assess the uniform distribution of solid particles in liquids at elevated temperatures, before solidification. Accurately measuring particle distribution in high-temperature fluids presents considerable difficulties due to several key challenges. These include the requirement for sensors that can withstand high temperatures, real-time, non-destructive measurement techniques capable of providing sufficient penetration depth, and the ability to monitor dynamic changes in particle distribution amidst stirring and thermal fluctuations. Traditional methods such as microscopy are inadequate due to their destructive nature.

Potential alternatives, including high-temperature ultrasound, electromagnetic methods, and specialized optical techniques, each possess their limitations and drawbacks. As such, image processing emerges as a promising alternative deserving further exploration and evaluation.

2. RELATED WORK

Digital image processing (DIP) technology has significant potential in particle mixture analysis. Nowadays, surface analysis has many applications across various fields, and one important area is material surface analysis [1]. Proposed a model where a color histogram was implemented to analyze particle mixtures in fluidized beds. The proposed method serves as a proof of concept, demonstrating that mixing particles changes the color distribution of the image and can be used to evaluate the quality of the mixing process. The results show that particle mixing alters the histograms of images taken at different times, providing an interesting avenue for exploring automatic particle mixing techniques based on image processing. On the other hand [2], [3], proposed an approach in which texture-based image processing plays a more prominent role than standard image enhancement. This method facilitates the detection of small obstacles during surface preparation by making individual features and surface details more visible. Various algorithms and mathematical models were applied to compare different image surfaces. In addition, statistical measures were used to evaluate the performance of this approach, further validating its effectiveness in surface analysis [4] studied the effects of adding micro, nano, and hybrid particles in a 1:1 ratio of Al2O3 and TiO2 to epoxy resin on thermal conductivity. The study analyzed samples before and after immersion in 0.3 N HCl acid for 14 days, considering weight fractions of 0.02, 0.04, 0.06, and 0.08, with a thickness of 6 mm. Image analysis techniques were employed to evaluate the homogeneity of the mixtures, highlighting the practical application of DIP in material science.

An analysis and comparison of the image analysis approach with other existing methods are necessary to evaluate the homogeneity of mixing in a moving casting. The selection of an appropriate method depends on several factors. These include particle properties such as size, shape, density, and concentration; fluid properties such as viscosity, density, and surface tension; type of mixing equipment, including mixer type, design, and operating conditions; required level of accuracy and sensitivity; and available resources and expertise. This research focuses on finding non-destructive tests. Even if we assume that destructive tests are acceptable, microscopic analysis, as one of the destructive tests, has limitations. It is time-consuming and provides data only from limited sample areas. It is time-consuming and provides data only from limited sample regions [5].

3. METAL MATRIX COMPOSITES

In 2022, the market for metal matrix composites (MMCs) was valued at USD 211.3 billion. It is projected to grow significantly, reaching USD 369.3 billion by 2032, with an annual growth rate of 6.40% during the forecast period from 2023 to 2032. The increasing demand for lightweight materials in the automotive and aerospace sectors is expected to be a key driver of this market growth. Fig. 2 illustrates the projected trend for MMCs in the coming years.

Fig. 2. MMC trend.

Aluminum matrix composites (AMCs) use pure aluminum or an alloy as the matrix and are becoming more popular in industrial applications due to their outstanding mechanical and tribological properties. The properties of Al alloys might be greatly tailored by adding ceramic reinforcing particles by stir casting. AA 6063 ingot was chosen as a matrix alloy which was analyzed carefully by using an X-MET8000 Handheld XRF analyzer and the compositions of the used material are shown in Tables 1-4.

TABLE 1: Composition of the AA 6063 ingot

TABLE 2: Chemical Composition of Silicon Carbide (SiC)

TABLE 3: Properties of Silicon Carbide (SiC)

TABLE 4: Properties of (Sio₂)*

In previous work [11], the visualization procedure was performed using two stirrer designs: Ahelical stirrer with a cylindrical shaft (D4) and a helical stirrer with a helical shaft (D5), as shown in Fig. 3. Aglass jar, similar in dimensions to a graphite crucible, was filled with fluids of varying viscosities – water (low viscosity) and engine oil (high viscosity) – to simulate the stirring process of molten aluminum and SiC or silica particles. The jar had a height of 150mm and a diameter of 100mm. A1% volume fraction of real micro-sized SiC and silica particles was added to the fluid, and stirring was conducted using an SEI-WA MG-915 radial drill connected to different stirrer designs. The stirring speed was set at 240 rpm, with the stirrer positioned at 30% of the crucible’s height from the base. Images were captured immediately after removing the stirrer following a constant holding time of 60 s of stirring. The images, taken from the open top of the jar, minimized glare issues caused by the glass. This visualization experiment was repeated under consistent conditions for all five stirrer designs: Single-blade, double-blade, multi-stage stirrer, helical stirrer, and helical stirrer with a helical shaft, as shown in Figs. 4 and 5. Both stirrers (D4 and D5) were used to mix a 0.05% volume fraction of SiC particles and silica sand particles (50–80 microns in size) under two conditions: Once with one liter of water and then with one liter of engine oil. The mixing was conducted for 60 s at a stirring speed of 250rpm, with the stirrer operating within the top two-thirds of the glass jar’s height. Images were captured from the open top portion, as shown in Fig. 6, to ensure clarity and avoid glare issues. These images were then analyzed using MATLAB to evaluate the mixing process.

Fig. 3. Implemented approach.

Fig. 4. SEIWA MG-915 Radial Drill setting.

Fig. 5. Images were captured from the exposed upper section.

Fig. 6. Five different designs of stirrers, a four blades single stirrer (D1), double stages stirrer (D2), a multi-stages stirrer (D3), a helical stirrer (D4), and a helical stirrer with helical shaft (D5)

5. IMAGE PROCESSING ANALYSIS

Traditional image processing has been widely utilized to enhance the aluminum alloy casting process by automating quality control and optimizing casting methods. Various image processing techniques have been employed for aluminum alloy analysis, including X-ray CT for internal defect detection, surface inspection for identifying surface flaws, real-time monitoring for process optimization, microstructure analysis for understanding material properties, dimensional measurement for precision, and robotic vision for automated handling [12], [13]. Mixing is a critical process that directly influences the quality of final products in several industries. Ensuring homogeneous mixtures requires a thorough examination of the mixing process [14]. This study highlights an alternative, non-invasive method for analyzing particle mixing and segregation using image processing, which has shown great potential. Aproposal to utilize low-cost cameras combined with image processing has been introduced as an effective, affordable solution [15]. High-tech methods for analyzing mixing processes are often costly and demand strict safety protocols, making image processing a promising alternative.

In recent years, image analysis has gained significance due to the increasing volume of images transmitted through various media in daily life. Image processing involves applying specific operations to enhance images or extract meaningful information from them. An image is mathematically represented as a two-dimensional function (x,), where x and y are spatial coordinates, and the amplitude at any coordinate pair corresponds to the image intensity. When, the amplitude values of f are finite and discrete, the image is termed a digital image [15]. Digital images are composed of pixels, each with a specific location and value. This paper uses statistical methods to identify optimal mixtures by analyzing images captured under varying incident light conditions and with different stirrer designs. Image content analysis is an essential step for producing accurate and unbiased evaluations.

Algorithms and techniques for processing color information in three-dimensional scenes and converting them into two-dimensional images in trichromatic color models and digital color spaces demonstrate similarities between color and grayscale image processing [16], [17]. Image processing is integral to modern manufacturing, enabling the manipulation and analysis of digital images to enhance product quality, extract critical information, and optimize processes. Techniques such as object recognition, segmentation, and optimization improve efficiency, automate quality control, detect defects, and streamline diagnostics, contributing to reliable and cost-effective production systems [18].

Image analysis involves extracting valuable information from images, typically using two main approaches: Spatial domain and frequency domain methods [19].

The statistical approach belongs to the spatial domain category, assessing various functions based on system objectives. Statistical measures such as mean and standard deviation play a significant role in image analysis by providing insights into the properties of incident light and mixing quality [20]. In this study, after completing the sample mixing process, images of the mixture were captured and analyzed using digital image processing techniques. The process involves several steps: Pre-processing, image acquisition, histogram equalization, edge detection, and contour detection. These steps are illustrated in Fig. 10.

Image processing provides an affordable alternative. It is a type of signal processing that generates digital images as its output or a set of properties or parameters connected to the image [15]. This is accomplished by the analysis of the pictures obtained during the mixing process by the camera. As a result of the mixing procedure, it has been discovered that the colors in the image do change. Image analysis was divided into two sections:

5.2. Implemented Approach

After completing the sample mixing, an image of the mixture is captured and analyzed based on the principles of digital image processing. The processing stage involves several steps, which are illustrated in Fig. 7:

• Pre-processing: The mixture is prepared, and an image is captured. Aportion of the image, sized 100 × 100 pixels, is selected for further processing.

• Image Acquisition: The captured image is formatted and prepared for the subsequent processing stages.

• Histogram Equalization: Histogram equalization is applied to enhance image clarity and improve visibility for the next steps.

• Edge Detection: This step involves identifying the edges within the image using the canny operator, which provides superior edge detection compared to other methods.

• Contour Detection: The contours of shapes in the image are identified to outline the structures present.

• Statistical Analysis: Various statistical measures are applied to analyze the image properties, including:

  • a. Standard Deviation: Measures the dispersion of data points and their deviation from the arithmetic mean, indicating the spread of pixel values.

  • b. Arithmetic Mean: Represents a measure of central tendency, summarizing the data set with a single value that denotes its average distribution.

  • c. Entropy: Quantifies the randomness or texture of the input image, providing insights into its complexity and information content.

Fig. 7. Helical Stirrer helical stirrer with a cylindrical shaft (D4) and a helical stirrer with a helical shaft (D5).

After processing images as shown its the steps in Fig. 8, the performances of the stirrers were calculated according to their pixel distribution.

Fig. 8. Image process steps from original to after-edge operations.

The findings from Figs. 9 and 10 suggest that D5 consistently outperforms D4 across various conditions. This includes low and high viscosity (water or engine oil) and low and high ceramic density (Silica sand or SiC). The mixing performance of D4 is characterized by a pixel distribution ranging from 75 to 165 grayscale values, while D5 shows a more refined pixel distribution between 90 and 150 grayscale values. This difference indicates that D5 achieves a higher level of homogeneity in the mixing process, as it maintains a more consistent pixel range, suggesting better uniformity in the particle distribution.

Fig. 9. (a and b) D4 stirrer mixing performance.

Fig. 10. (a and b) D5 stirrer mixing performance.

Fig. 11 illustrates the distribution of standard deviation values for the processed images. Standard deviation is a measure of the variation or spread within a set of values. Alow standard deviation indicates that the pixel intensities are close to the average, while a high standard deviation suggests a greater deviation from the average. In all conditions, the standard deviation values for D4 are higher than those for D5. This suggests that D4 exhibits greater variation in pixel intensities, indicating a less homogeneous mixing process compared to D5, which shows more consistent pixel values and a higher degree of homogeneity.

Fig. 11. The standard deviation for D4 and D5 stirrers.

6. CONCLUSION

7. ACKNOWLEDGMENT

Sincere appreciation is extended to all individuals who provided support to this paper. Their invaluable guidance, assistance, and encouragement are deeply acknowledged and greatly appreciated.

8. DECLARATION OF CONFLICTING INTERESTS

The authors stated that there are no potential conflicts of interest related to the research, writing, or publication of this article.

9. FUNDING

The authors did not receive financial support for any aspect of the research, authorship, or publication of this article.

10. DATA AVAILABILITY

No new data were generated or analyzed in support of this research.