Lubricating Oil Wear Particle Detection Technology Based on Telecentric Imaging and Random Forest

Fig. 5. Wear particle images and boundary images. (a)‒(c) Original collected images; (d) normal particle; (e) fatigue particle; (f) cutting particle; (g) sphere particle

Download full size

Fig. 6. Impact of number of decision trees on classification accuracy

Download full size

Fig. 7. Impact of number of features on classification accuracy

Download full size

Fig. 8. Confusion matrix of classification result

Download full size

Table 1. Features of wear particle

View table

Table 1. Features of wear particle

Parameter	Description
Area $A$
Perimeter $P$
Equivalent diameter $D$	$D = \sqrt[]{\frac{4 A}{π}}$
Aspect ratio $A_{R}$	$A_{R} = \frac{b}{a}$
Roundness $R_{o}$	$R_{o} = \frac{4 A π}{P^{2}}$
Convexity $C_{o}$	$C_{o} = \frac{P_{c}}{P}$
Solidity $S_{o}$	$S_{o} = \frac{A}{A_{c}}$
Curvature $k_{i}^{m}$	$θ_{i - 1}^{m} = a r c t a n (\frac{y_{i} - y_{i - m - 1}}{x_{i} - x_{i - m - 1}})$ ， $θ_{i + 1}^{m} = a r c t a n (\frac{y_{i + m + 1} - y_{i}}{x_{i + m + 1} - x_{i}})$ ， $k_{i}^{m} = θ_{i + 1}^{m} - θ_{i - 1}^{m}$

Table 2. Comparison of system parameters
View table
Table 2. Comparison of system parameters
System name Field size /（mm×mm） Flow channel depth /mm
Telecentric imaging system 2.1×1.8 0.2
Microscopic imaging system 0.9×0.7 0.1

Table 3. Partial feature data of four types of wear particles
View table
Table 3. Partial feature data of four types of wear particles
Category $A_{R}$ $R_{o}$ $C_{o}$ $S_{o}$ $K_{μ}$ $K_{σ}$
Normal 0.5158 0.7797 0.9857 0.9605 14.0259 21.7580
Fatigue 0.7678 0.6203 0.8975 0.8551 12.3750 34.5541
Cutting 0.4634 0.4044 0.9064 0.6259 14.5455 29.4525
Sphere 0.9354 0.9730 0.9922 0.9871 12.4137 15.0442

Table 4. Comparison of classification accuracy
View table
Table 4. Comparison of classification accuracy
Classification algorithm Classification accuracy /%
RF 93.75
SVM 87.50
KNN 76.25

Tools

Get Citation

Copy Citation Text

Zhuoran Cao, Fajie Duan, Xiao Fu, Guangyue Niu. Lubricating Oil Wear Particle Detection Technology Based on Telecentric Imaging and Random Forest[J]. Laser & Optoelectronics Progress, 2025, 62(6): 0612003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites