09-11-2016, 12:55 PM
1467705423-kjb.docx (Size: 145.12 KB / Downloads: 4)
ABSTRACT
This study aims at developing a non-contact method for measuring the grinding wheel loading and thereby determining the optimum dressing intervals. With the aid of a machine vision system, this paper presents a systematic process for measuring the wheel loading. The images of the grinding wheel had been taken using a digital camera. These images had been transferred to the computer and are processed for determining the percentage of loading. The image toolbox of MATLAB had been used for image processing. Global thresholding technique had been used to differentiate the loaded portion of the wheel from rest of the wheel. Experiments were conducted on Mild Steel and HCHCR Steel (High Carbon High Chromium Steel). Experimental results are presented which show the ability of using machine vision system in the online monitoring of the grinding wheel loading.
INTRODUCTION
Grinding is an abrasive machining process that uses a grinding wheel as the cutting tool. It can produce very fine finishes and very accurate dimensions. Grinding is one of the final machining processes that determine the surface quality of machined products. After a long period cycles times of the grinding process, removed chips may stick in the space between abrasive grains or weld on the top of cutting edges, which is called as wheel loading. Factors such as wheel loading and wheel wear contribute to the deterioration of the working surface and also its cutting capacity. When the loading and wear is severe, dressing of grinding wheel has to be carried out in order to bring the wheel to its best state. Determining the timing to dress the grinding wheel is extremely important in order to prevent flaws in products. Nowadays in industrial applications experience of the skilled operator plays a major role in determining the dressing interval. Therefore monitoring the grinding process for wheel loading and wheel wear is critical to ensure the surface quality of the machined component as well as the efficiency of the grinding process.
Much research concerning the monitoring of grinding process has been conducted using AE[1,2], hydrodynamic pressure [3], neural network and fuzzy logic [4], Eddy currents and laser methods[5], CCD [6].
Z.Feng et al. [7] uses the image processing toolbox of MATLAB to measure the wear of the grinding wheel. K.C.Fan et al. [8] measures the wear of grinding wheel using binarisation technique for image processing and the image is analyzed using a derived edge detection algorithm to determine the spatial coordinates of the grinding wheel edge curvature. Ste´phane Lachance et al. [9] applies a region growing method in image processing to measure the wear flat area of the grinding wheel.
Advances in the computer vision technology have led to the investigation of its application in the monitoring of the grinding process. Visual information has the advantage, that can be interpreted very easily and due to its high information content is the first choice to investigate typical surface forms, which cannot be extracted from indirect measurement signals.
Most of the previous work in the condition monitoring of grinding process has been concentrated on wheel wear. This study aims at determining the wheel loading through capturing the image of the grinding wheel surface and by analysing the image through various image processing technique.
DIGITAL IMAGE PROCESSING
Image processing has been carried out using Matlab software. Global Thresholding has been utilized for this purpose. By studying the optical properties of removed chips as well as the good wheel, a suitable threshold value of gray scale level has been selected. By setting this threshold value, all those pixels whose gray scale level is greater than that particular threshold is turned into white, while the remaining pixels is turned into black. Since the reflectivity of the loaded portion is higher, the white pixel corresponds to the loaded portion. By counting the number of white pixels and dividing it with total number of pixels, will gives the percentage of wheel loading.
EXPERIMENTAL SETUP
The experimental setup consists of a digital camera, tripod, grinding wheel and laptop as shown in figure 1. The digital camera is placed in coaxial with the wheel surface in order to make ensure that the incident light is perpendicular to the surface of the grinding wheel. The digital camera used for the experimental purpose is Kodak Easyshare CD14. It has 8.2megapixel and 3x optical zoom. A tripod is used to adjust and fix the position of digital camera. Tripod with digital camera is fixed on a proper location so that images of the grinding wheel surface can be obtained without any interruption.
1. EXPERIMENTS
Fig.1 Block Diagram of Experimental Setup
Experiments were conducted on both Mild Steel and HCHCR (High Carbon High Chromium) Steel. Before putting into operation, grinding wheel is dressed in order to bring the wheel to the best condition. Grinding wheel image is taken at this point of time. Speed of the grinding wheel is set at 2500rpm, feed at 0.06mm and depth at 0.1mm. Grinding wheel is put into operation for 1 hour. 12 images were taken for every five minute interval for mild steel specimen and 10 images were taken for HCHCR (High Carbon High Chromium) Steel at certain intervals of time. The captured image is transferred to laptop and is processed using Matlab software. By using global thresholding technique,
a binary image is created with loaded portion in white pixels and rest of background in black pixel. By analyzing the optical properties of fully dressed wheel and chips a threshold value of 200 is selected for a particular grinding wheel. Gray scale level value above 200 represents the loaded portion and is converted into white pixels whereas those below 200 shows the portion of abrasive grains on the wheel and is converted to black pixels.
1. RESULTS AND DISCUSSIONS
The actual image and processed image of fully dressed wheel and loaded wheel at the end of 1 hour of operation on Mild Steel is as shown in figure 2. Actual image obtained is color image and it had been converted to gray scale image. The white spots on the processed image indicate the percentage of loading. There are no white spots on the processed image of fully dressed wheel (Figure 2.a). This indicates that wheel is in good condition or no loading of wheel had taken place. As the grinding operation continues percentage of loading goes on increasing. Processed image of wheel at the end of 1hour of operation (Fig 2.d) had large number of white spots which indicates that a certain percentage of loading had occurred.
Percentage of loading at the end of every five minutes of operation on Mild Steel is shown in table 1. Percentage of loading as given on table 1 indicates that percentage of loading increases with time. Since the carbon content associated with Mild Steel is very low even after 1 hour of operation the percentage of loading obtained is only 0.243%.
INFERENCE
Experiment conducted on mild steel specimen shows that percentage of loading had some positive relationship with time. Since the carbon content associated with mild steel specimen is low and also the speed and feed given on the grinding machine is low, only a very small percentage of loading occurs. In order to obtain more amount of loading a specimen with high carbon content is used (HCHCR Steel).
2. CONCLUSION
This study shows the feasibility of using machine vision system for determining the loading of the grinding wheel. The following conclusion can be drawn from the study.
1. The high reflectivity of chips which get accumulated within the abrasive grains helps to differentiate the loaded portion from rest of the wheel, by setting a suitable high threshold value. By analyzing the histogram of good wheel and also by analyzing the optical properties of specimen a threshold value of 200 is selected for these experiments.
2. The optimum dressing time can be determined by measuring the wheel loading at certain intervals of time. The optimum dressing intervals can be determined by relating the surface finish with percentage of loading.
3. Experiment conducted on Mild steel and HCHCR steel shows that percentage of loading increases with time.
3. SCOPE FOR FUTURE WORK
Instead of manually capturing the image and transferring it to the computer fully automate the image capturing and analyzing process. Also there is a possibility of relating percentage of loading with surface finish using Artificial Neural Network.