Novel, High-Precision, On-Machine Approach for Measuring Cup Grinding Wheel Wear Using a Moveable Laser Displacement Sensor

Abstract

This study developed a movable, high-precision laser measurement system for assessing wear on cup grinding wheels. The proposed setup employs a Keyence CL-P070 laser sensor with a resolution of 0.25 μm and has a simple installation process that supports flexible deployment on multiple workstations. Unlike traditional static configurations, the compact design requires minimal adjustment and enables versatile positioning across operational environments. An automated measurement procedure was developed that can capture changes in the grinding wheel profile between grinding cycles. The experimental results indicate that the proposed system has high repeatability and accuracy in detecting the subtle progression of wear in cup grinding wheels. The proposed approach provides a user-friendly on-machine measurement solution that can improve quality control and operational efficiency in industrial grinding processes.

1. Introduction

Smart manufacturing has transformed industrial production, enabling enhanced automation, real-time monitoring, and data-driven decision making. Additionally, advances in artificial intelligence, sensor technologies, and big data analytics have facilitated the widespread adoption of intelligent manufacturing systems. Among these intelligent manufacturing systems, automated optical inspection and prognostics and health management technologies have become integral components in quality control and predictive maintenance, reducing downtime and improving manufacturing efficiency. Automated optical inspection systems employ computer vision and deep learning algorithms to detect defects with high accuracy, whereas prognostics and health management enable the proactive maintenance of equipment by predicting failures on the basis of historical and real-time sensor data. These technologies contribute to Industry 4.0, in which cyber–physical systems and smart sensors seamlessly combine to optimize production processes.

Advances in tool condition monitoring (TCM) and remaining useful life (RUL) prediction are critical to the machine tool industry. TCM involves the real-time tracking of tool wear and degradation to maintain machine accuracy and prevent unexpected failures, whereas RUL estimation is a method of predicting the remaining lifespan of cutting tools, enabling the optimization of tool replacement strategies. Studies have explored various TCM and RUL approaches, such as multisensor fusion techniques, machine learning algorithms, and signal processing methods. For example, Maier et al. [1] developed a self-optimizing grinding system utilizing Gaussian process regression and Bayesian optimization to achieve favorable tool longevity and efficiency. Luo et al. [2] investigated the influences of processing parameters on the wear characteristics of metal-bonded diamond grinding wheels through a combination of tool-evaluating instruments and surface morphology analysis. Furthermore, Bi et al. [3] employed acoustic emission (AE) signals to classify the states of a grinding wheel’s wear, demonstrating the feasibility of noninvasive monitoring techniques. These advancements have laid the groundwork for integrating intelligent monitoring solutions into production lines, reducing the costs associated with tool failures and unplanned downtime. However, although considerable progress has been made regarding the monitoring of conventional cutting tools and grinding wheels, research specific to cup grinding wheels has been lacking, and further investigation of their wear characteristics and measurement methodologies is warranted.

Grinding machines have long been integral to the fabrication of optical, semiconductor, and precision mechanical components. The study of grinding wheel wear monitoring has been a focal point of research for more than half a century, evolving from early direct observational methods to contemporary multisensor fusion and machine learning approaches.

One of the earliest systematic studies on grinding wheel wear was conducted by Malkin and Cook [4], who laid the groundwork for understanding fracture-induced wear mechanisms in grinding wheels. Their work provided fundamental insights into the material removal and degradation processes at the abrasive grain level. Subsequently, Sfantsikopoulos and Noble [5] explored the role of electrochemical grinding in the conditioning of diamond-grit cup wheels. Their findings highlighted how mechanical and electrochemical conditioning can improve grinding performance and extend a wheel’s life, particularly in electrochemical-assisted grinding applications.

Traditional direct measurement methods rely heavily on scanning electron–optical microscopy, or laser measuring device for surface characterization. Pan et al. [6] investigated the use of a laser displacement sensor to measure and analyze the height distribution of the grinding wheel surface, enabling the reconstruction of its 3D surface topography. Ito et al. [7] applied infrared thermometry to assess in situ grinding temperature variation and indirectly estimated wheel wear by correlating surface temperature increases with wheel deterioration. Nasr and Davoodi [8] employed scanning electron microscopy to evaluate subsurface damage and surface roughness in optical glass grinding and proposed a quantitative link between uncut chip thickness and the depth of subsurface damage.

Advances in sensor technology have led to increased use of indirect measurement techniques, particularly AE monitoring. In AE-based methods, high-frequency stress waves generated during the grinding process that reflect the condition of the wheel are captured. Herman and Krzos [9] analyzed vitrified bond structures in cubic boron nitride cBN grinding wheels and established a correlation between radial wear and grinding efficiency. Additionally, Bi et al. [3] developed an AE-based monitoring system that utilizes convolutional neural networks (CNNs) and long short-term memory models to accurately classify grinding wheel wear states.

Vibration analysis is another approach to assessing wheel degradation. Duan and Cao [10] integrated vibration, AE, and force sensors in a fusion framework employing machine learning models to achieve high monitoring accuracy. Their approach demonstrated that combining AE energy features with grinding force ratios substantially improved the reliability of wear predictions, especially in aerospace-grade material grinding.

Advanced sensor fusion techniques have recently enabled real-time and highly precise wheel wear monitoring. For example, Zhuo et al. [11] introduced a cloud-edge-device collaborative monitoring system utilizing AE, vibration, and power sensors to enhance predictive maintenance in the high-speed cylindrical grinding of bearing rings. Their work highlighted the importance of edge computing, which reduces data transmission delays and enables real-time decision making. They further improved the wear assessment accuracy by integrating Dempster–Shafer evidence theory into their models.

Kacalak and Lipiński [12] developed a diagnostic framework on the basis of surface topography metrology, where this framework employed optical profilometry and power consumption monitoring to quantify grinding wheel wear. Their research emphasized the stochastic nature of wear progression, and they proposed a statistical feature extraction-based classification model suitable for high-precision grinding applications.

Machine learning-based monitoring has revolutionized grinding wheel wear prediction, enabling the development of adaptive control strategies. For example, Lee and Jwo [13] pioneered an artificial intelligence-based sound monitoring system in which deep learning is used to classify wheel condition. Their CNN-based classifier accurately predicted wheel wear states, demonstrating that low-cost microphones can be used instead of expensive AE sensors.

Extending the use of artificial intelligence, Maier et al. [1] proposed a self-optimizing grinding system based on Gaussian process regression and constrained Bayesian optimization. In this approach, the grinding parameters are dynamically adjusted on the basis of the real-time wear, reducing experimental costs and maintaining surface quality. Additionally, Pazmiño et al. [14] employed discrete element method simulations to model a grinding wheel’s volumetric wear. Their work predicted grinding ratios by simulating grain detachment mechanisms, providing insights into abrasive grain retention and wheel lifespan.

In addition to developing and assessing monitoring methods, studies have evaluated the effects of a grinding wheel’s composition on its wear. For example, Luo et al. [2] investigated metal-bonded diamond wheels used for sapphire wafer grinding, categorizing the wear process into four distinct phases: slow wear, rapid wear, stable wear, and blunt wear. Their findings emphasized the necessity of periodic dressing to restore grinding efficiency.

Herman and Krzos [9] analyzed the role of vitrified bond structures in cBN wheels, demonstrating that glass–ceramic bonding agents could enhance the wheels’ longevity. Moreover, studies such as that of Solhtalab et al. [15] examined the cup grinding of optical glass, and those authors proposed a finite element model coupled with smooth particle hydrodynamics for predicting subsurface damage progression. Their simulation approach validated theoretical subsurface damage models and confirmed that minimizing the uncut chip thickness effectively reduces subsurface damage depth.

Electroplated and hybrid grinding wheels have also been extensively researched. Vidal et al. [16] conducted an in-depth study on electroplated cBN wheels, focusing on wear mechanisms and conditioning techniques for the creep-feed grinding of aeronautical alloys. Their findings demonstrated that SiC-based dressing effectively improved the cutting efficiency but accelerated wheel wear under high loads. Furthermore, Sfantsikopoulos and Noble [5] provided early insights into diamond-wheel electrochemical dressing, a subject increasingly studied because of the ability of this process to minimize grit retention loss while enhancing wheel sharpness.

Despite the substantial advances made in grinding wheel wear monitoring, research on the monitoring of cup grinding wheels has been lacking. The unique geometric characteristics and operational dynamics of cup wheels pose challenges and mean that specialized measurement strategies, different from those for conventional cylindrical and flat grinding configurations, are required. Most studies have focused on general grinding wheel wear assessment; few have investigated the unique wear mechanisms of cup wheels. Therefore, further research is required to develop methodologies that accurately capture wear progression in cup grinding wheels and thereby address scientific and industrial needs.

The definitions of wear used in research on cup grinding wheel monitoring have been inconsistent, with approaches ranging from abrasive grain loss area analysis to volumetric wear rate estimation, and this inconsistency has led to difficulties in standardizing wear assessment methodologies. Moreover, monitoring systems predominantly employ ordinal labels, such as binary classifications indicating the presence or absence of wear, rather than continual and quantitative metrics that accurately represent wear volume. This lack of precise wear quantification hinders the development of predictive maintenance strategies and limits the integration of intelligent monitoring systems into industrial applications.

Addressing this gap, the present study considered the volumetric wear rate of cup grinding wheels and proposed a novel, on-machine measurement system capable of rapidly and precisely assessing wear in these wheels. High measurement accuracy was achieved using a comprehensive wear evaluation framework, and a foundation for integrating RUL prediction and TCM technologies into cup grinding wheel applications was established. The proposed methodology provides a scalable and adaptive approach to real-time wear monitoring that achieves highly favorable precision and efficiency. This study contributes to the advancement of intelligent grinding processes by enabling data-driven decision making, improving process sustainability, and extending the operational lifespan of grinding tools.

2. Materials and Methods

2.1. Experimental Setup

The experimental setup was implemented on a high-precision intelligent five-axis grinding machine, the Tongtai GT-630 (Tongtai Machine & Tool Co., Ltd., Kaohsiung, Taiwan). The grinding wheel used in the experiment was a diamond abrasive cup grinding wheel, selected for its high durability and precision in machining applications. The measurement system utilized a Keyence CL-P070 high-precision laser displacement sensor paired with a dedicated CL-3000 controller. A standard laptop computer was used for data acquisition and processing. Figure 1 illustrates the grinding machine and installed sensor system.

Figure 2 depicts the GT-630 machine with the CL-P070 sensor, used for the in situ measurement of the grinding wheel profile. By controlling the movement of the machine spindle, this study acquired sectional profiles of the grinding wheel’s cutting edge at eight angular positions. The Keyence CL-P070 technical specifications used to ensure measurement accuracy are presented in

Table 1. Specifications of the Keyence CL-P070 laser displacement sensor.

Head Type		70 mm Head
Model	Head	CL-P070
	Optical Unit	CL-P070N
Type		Spot type
Reference Distance		70 mm
Standard Precision Measurement Range	Measurement Range	±10 mm
	Linearity	±2.2 μm
High-Precision Measurement Range	Measurement Range	±3.0 mm
	Linearity	±2.0 μm
Resolution		0.25 μm
Spot Diameter		φ50 μm

. The resolution of the CL-P070 was 0.25 μm, which was sufficient for the present study because the smallest cup grinding wheel used in optical applications had a diameter of 10 mm. This resolution conferred the ability to detect even minute changes in the wheel profile with high fidelity.

The depth of the grinding wheel cutting edge had to exceed 1 mm, meaning a wide measurement range. Hence, the standard measurement range mode was selected for data acquisition. In this mode, the linearity of the sensor is ±2.2 μm, indicating that the maximum deviation between the recorded displacement values and the actual distances does not exceed 2.2 μm under ideal, interference-free conditions.

The GT-630 machine’s typical machining parameters were adopted for the experiment to ensure consistency with and relevance to industrial applications. The grinding process was conducted with a spindle rotational speed of 2500 rpm, a C-axis rotational speed of 7 rpm, and a feed rate of 0.4 mm/min. The laser displacement sensor was set to a scanning frequency of 1024 Hz, ensuring high-resolution measurement data. The spindle feed rate was adjusted to 50 mm/min during the on-machine profile measurement to ensure stable and accurate scanning.

Initially, the profile of a new, unused grinding wheel was recorded. The wheel was then used for grinding, and after each grinding operation, the machine was immediately stopped, and an on-machine profile measurement was conducted to capture the degree of wear. This process was repeated until the grinding wheel was worn beyond usable limits. To validate the robustness and repeatability of the proposed measurement approach, in the 74th machining cycle, the laser displacement sensor was reinstalled and recalibrated, enabling a comparative analysis to assess the completeness and reliability of the methodology.

In this experiment, the grinding wheel reached its end of life at 124 machining cycles, which indicated the feasibility of the proposed measurement approach for tracking progressive wear under realistic machining conditions. This experimental validation step ensured the reliability of the measurement methodology.

2.2. Methods

This study quantified grinding wheel wear by analyzing the cross-sectional area after each machining cycle. Theoretically, the greater the change in the grinding wheel profile, the greater the influence of wear on workpiece quality. Therefore, if the laser displacement sensor data can accurately be processed to quantify the difference in profile before versus after each cycle, the wear state of the grinding wheel can be evaluated.

The experimental setup described in Section 2.1 enabled the acquisition of high-resolution grinding wheel profile data, which were recorded in Excel. Although the data in the dataset were automatically labeled GO (good) or NG (no good), the data marked as GO could not be directly used for analysis for the following reasons:

• The grinding wheel profile was a continuous variable, whereas the sensor captured discrete data points. Hence, the first profile point labeled GO was unlikely to correspond precisely to the boundary of the grinding wheel profile.
• An exact overlap between the first GO-labeled profile point in any two consecutive datasets was unlikely. An inexact overlap would lead to subsequent profile points also failing to align at the same locations on the grinding wheel profile.

To address these concerns, a profile extraction mechanism was developed for aligning and standardizing the starting reference point of each acquired profile. This step is crucial to minimizing overlay errors that could compromise the accuracy of wear assessments.

Figure 3a, showing the raw data obtained from the laser displacement sensor, indicates the grinding wheel profile at eight angular positions. Figure 3b illustrates an extracted profile for a single angular position. To ensure consistency, the precalibration profiles (before the 74th cycle) and the profiles of selected grinding cycles (the 16th, 30th, 38th, 52nd, and 68th cycles) were manually aligned to establish a reference framework, illustrated in Figure 4.

This study proposed three profile extraction methods, detailed in Section 2.3, on the basis of the characteristics of the grinding wheel profiles and to accommodate real-world constraints. Two key evaluation metrics were defined:

• Cross-sectional area reduction: This measure approximates the volumetric wear and is used to evaluate material loss;
• Trend in the cross-sectional area reduction: As machining progresses, the degree of reduction monotonically increases, reflecting the physical behavior of progressive grinding wheel wear.

2.3. Methods of Profile Extraction

2.3.1. Full Interpolation Approach

To determine the cross-sectional area of the grinding wheel’s cutting edge, an interpolator function is required to define the profile trend. Nevertheless, because of variation in sample point alignment, the start and end points of the cutting edge must be identified and normalized before processing can be performed.

The workflow of the full interpolation approach is illustrated in Figure 5. In the proposed approach, interpolation functions f₁ and f₂ are employed to detect the start and end points within their respective regions. The complete profile is then processed using an interpolator f, which generates equidistant profile points on the basis of the identified start and end points. This method ensures a comprehensive perspective of the grinding wheel profile by considering the entire shape rather than isolated segments.

2.3.2. Three-Section Interpolation Approach

In this approach, three interpolators (f₁, f₂, and f₃) are introduced to independently model the front 2 mm, middle 2 mm, and rear 1 mm of the grinding edge shown in Figure 6. This method is a valid approach because the cutting edge of the grinding wheel exhibits more pronounced variation at the points of contact than it does elsewhere.

In the full interpolation approach, inaccuracies may arise when determining the start and end points, especially if the cutting-edge turning points are incorrectly interpreted or if extrapolation methods introduce bias. Because the grinding wheel cutting edge transition occurs at a depth of approximately 1 mm, a threshold of 0.8 mm was selected as a segmentation boundary to align the start and end points.

Once the start and end points have been determined, f₁ is used to compute the start point, f₃ is used to calculate the end point, and f₁, f₂, and f₃ are employed to generate equidistant profile points. The three segments are connected to reconstruct the final grinding wheel profile.

2.3.3. Semiautomated Three-Section Interpolation Approach

Further investigation revealed that the previous approaches had limitations when applied to profiles acquired after baseline realignment (after the 74th cycle). A semiautomated three-section interpolation approach that retains the three-section structure while incorporating automated adjustments to enhance accuracy is introduced to address these challenges. The key improvements are as follows:

• Automatic edge point detection through slope-based searching: The turning points of the cutting edge are automatically identified on the basis of local slope variation, meaning that manual intervention is unnecessary.
• Statistical refinement of segmentation ratios: The front, middle, and rear sections are proportionally optimized (in a 26:41:13 ratio) using empirical data, which ensures the accurate partitioning of the grinding edge.
• Automated realignment of profiles: A root mean square comparison of each cycle’s front and rear profile sections with the initial reference profile is used to dynamically adjust the start and end points, minimizing deviations due to installation errors or machine drift.
• Smoothing function for contact point refinement: A dedicated smoothing function is applied to the transition zone between f₁ and f₂ and the transition zone between f₂ and f₃ to ensure a continual and realistic grinding wheel profile.

The workflow of the semiautomated three-section interpolation approach is illustrated in Figure 7. In this approach, the cutting-edge turning points are first identified through slope detection. Subsequently, the 26:41:13 ratio is used to design the three-section interpolators (f₁, f₂, and f₃). Automatic calibration is conducted by matching the current contour with the closest corresponding section in the initial profile. The final step involves calculating the profile coordinates and applying a smoothing function to enhance continuity.

By incorporating these improvements, this method reduces the influence of installation and operational errors, enhancing the precision of grinding wheel profile extraction for wear analysis.

3. Results and Discussion

This study conducted 123 grinding operations using a cup grinding wheel; the wheel’s profile was measured before and after each operation, resulting in 124 profile data points. In tests of the 74th data point, the measurement reference point was modified, with a more stringent test condition introduced to validate the robustness of the proposed profile extraction approaches. This section compares the results of using the full interpolation, three-section interpolation, and semiautomated three-section interpolation approaches.

To assess the effectiveness of the extraction methods, the semiautomated three-section interpolation approach was excluded from the initial comparison because it was specifically designed to accommodate changes in the measurement reference introduced in the 74th data point. Therefore, the full and three-section interpolation approaches were first compared, and differences in their profile extractions were examined.

Figure 8 presents the extraction results of the two methods. Figure 8a compares the 0th and 67th profiles obtained using the full interpolation approach, whereas Figure 8b displays an overlay of five profiles extracted using the same method. Figure 8c,d illustrate the results of using the three-section interpolation approach, depicting the results of analyzing the 0th, 16th, 33rd, 50th, and 67th profiles. A visual inspection revealed no significant differences between the results obtained using the two approaches. Consequently, cross-sectional area integration was conducted to quantify wear trends and assess the extraction performance more rigorously.

An examination of the extracted profiles beyond the 74th measurement revealed that the full and the three-section interpolation approaches exhibited poor performance. Additionally, before the 74th measurement, inconsistencies were frequently found between the extracted profiles and the original 0th profile boundary, leading to misalignment. To address these challenges, the semiautomated three-section interpolation approach was introduced. Because the data depicted in Figure 8 demonstrated that the full interpolation and three-section interpolation approaches yielded similar results, subsequent analyses focused on comparing the three-section interpolation approach with the semiautomated three-section interpolation approach.

Figure 9 presents a comparison of the two methods, showing the profiles extracted at the 0th, 16th, 32nd, 48th, 64th, 80th, 96th, and 112th cycles. Figure 9a presents the results obtained using the three-section interpolation approach, whereas Figure 9b displays those obtained using the semiautomated three-section interpolation approach. Compared with the results presented in Figure 8, the data depicted in Figure 9 show finer details and more clearly present discrepancies.

Large differences were discovered when evaluating the 48th and 64th cycle profiles obtained using three-section interpolation. Specifically, the 48th profile boundary failed to align with the 0th profile, and the profile recession behavior deviated from what was expected given the progressive reduction trend. A similar inconsistency was found in the 64th profile, further indicating the inadequacy of the three-section approach. The semiautomated three-section interpolation approach mitigated these problems, exhibiting superior ability to maintain profile alignment and retain recession trends. Additionally, for the 80th, 96th, and 112th profiles, the semiautomated approach resulted in more consistent boundary alignment and wear progression trends.

The preceding analyses were based on graphical evaluations of the extracted profiles. Nevertheless, a quantitative assessment was necessary to comprehensively compare the three approaches.

Table 2. Difference in area integrals for the three approaches. (Unit: mm²).

Compare No.	Full Interpolation Approach	Three-Section Interpolation Approach	Semiautomated Three-Section Interpolation Approach
No.0–No.16	−0.014778	0.002896	0.06408
No.0–No.32	−0.027222	0.000469	0.128057
No.0–No.48	−0.051128	0.013378	0.125873
No.0–No.64	0.110539	0.146211	0.238452
No.0–No.80	−0.785251	−0.646887	0.345205
No.0–No.96	−0.647605	−0.499334	0.433036
No.0–No.112	−0.686698	−0.598129	0.390867

presents the computed cross-sectional areas for the 0th, 16th, 32nd, 48th, 64th, 80th, 96th, and 112th profiles, where these areas numerically represent the volume of the grinding wheel’s wear.

Table 3. Trends in the difference in area integrals for the three approaches. (Unit: mm²).

Compare No.	Full Interpolation Approach	Three-Section Interpolation Approach	Semiautomated Three-Section Interpolation Approach
No.16–No.0	−0.014778	0.002896	0.06408
No.32–No.16	−0.012444	−0.002427	0.063977
No.48–No.32	−0.023906	0.012909	−0.002184
No.64–No.48	−0.051128	0.132833	0.112579
No.80–No.64	−0.895790	−0.793098	0.106753
No.96–No.80	0.137646	0.147553	0.087831
No.112–No.96	−0.039093	−0.098795	−0.042169

presents an additional quantification of the cross-sectional area reduction trend, listing the calculated difference in area between successive profiles and providing insights into the wear trend.

An analysis of the values presented in Table 2 and Table 3 uncovered notable discrepancies between the semiautomated approach and the other two methods, particularly for the 80th, 96th, and 112th profiles. After the 80th profile, both the full and three-section interpolation approaches yielded profiles that exhibited substantial deviations, suggesting an inability to adapt to changes in reference point calibration (Table 2). By contrast, the semiautomated approach yielded consistent profiles, indicating the robustness of this approach when handling reference point shifts.

The results presented in Table 3 provide a more intuitive representation of trend stability. The full interpolation approach frequently yielded negative trend values, which are inconsistent with the physical nature of progressive grinding wheel wear. Excluding anomalies such as the 80th to 64th profile comparisons resulted in a substantial improvement in trend stability for both the three-section and semiautomated three-section interpolation approaches. Additionally, when negative trends were detected, the semiautomated approach demonstrated superior robustness, yielding values closer to 0, confirming the approach’s effectiveness in ensuring progressive wear measurement consistency.

Figure 10 and Figure 11 visually support the numerical findings by illustrating the cumulative and incremental differences in area integrals across all evaluated profiles. As shown in Figure 10, the semiautomated three-section interpolation approach consistently generated positive and gradually increasing area differences, aligning with the expected behavior of progressive wear. In contrast, the full and three-section interpolation approaches exhibited irregular and, at times, abrupt deviations—most notably between the 64th and 80th profiles—indicating reduced adaptability under changing wear conditions. Figure 11 further emphasizes the local stability of the semiautomated approach, with smoother transitions and fewer oscillations in incremental area differences. These visual trends corroborate the quantitative results from Table 2 and Table 3, affirming the semiautomated approach’s superior robustness and its suitability for capturing both global and local wear progression accurately.

Both the graphical validation through profile overlays and numerical verification using cross-sectional area computations verified that the semiautomated three-section interpolation approach outperformed the other two methods in terms of precision and robustness. This method provides an adaptive and physically consistent approach to grinding wheel wear extraction that is superior for real-world applications.

4. Conclusions

This study developed a portable on-machine measurement system for assessing the wear of cup grinding wheels and addressed the quantification and definition of wear for this category of grinding tool. Additionally, several profile extraction approaches were proposed and systematically compared to evaluate their performance in processing grinding wheel contour data.

Through a comprehensive evaluation involving graphical profile overlay comparisons, cross-sectional area analysis, and cross-sectional area trend analysis, this study concluded that the proposed semiautomated three-section interpolation approach exhibits the highest accuracy and robustness of the three tested models. The results indicate that this method is highly adaptive, enabling straightforward implementation across machine tools, cup grinding wheels, and cases in which sensor misalignment occurs due to human error. A simple horizontal calibration is sufficient to restore measurement accuracy, making this approach practical for diverse industrial applications.

The two evaluation metrics introduced in this study—the cross-sectional area and cross-sectional area reduction rate—can serve as standardized quantitative indicators for future research in similar domains. By improving the efficiency, accessibility, and accuracy of cup grinding wheel wear measurement, this study advances TCM systems for applications in the optical and semiconductor industries. Overall, this study addresses a research gap regarding the definition and measurement of cup grinding wheel wear and lays a foundation for further improvements in measurement methodologies. However, several areas for improvement remain. Future research should focus on integrating the continually evolving measurement approach with the Keyence API to develop a human–machine interface. Additionally, a three-dimensional interactive grinding wheel model could be constructed by synthesizing the cross-sectional profiles acquired from eight angular positions to provide a more intuitive and detailed visualization of wear progression. These enhancements could strengthen the applicability of the proposed system in smart manufacturing environments.