Using Artificial Neural Networks For Process Planning Of Cylindrical Machined Components

Dr. H El Awady

Zagazig University

Process planning

 

Process planning

pix1

Factors affecting process plan selection

The problem Definition

Process Planning Steps

Process Planinng Steps (con.)

Factors considered for the design of the alogorithms

The Al techniques used to generate CAPP systems

Neural networks

pix2

The following factors influence one’s decision for selection of machining operations

  • Type of the feature being machined
  • Size of the feature
  • Technological attributes required of the feature (e.g. tolerance, surface finish)
  • Machining processes and their capabilities.

Capabilities of typical machining operations

 Operation Diameter (mm)  Length(mm)  Tolerance (mm)   Surface finish (microns)   Precision attribute
 Drilling

3-100

10-300  0.08-0.4  1.6-9.5   
 Reaming 10-100 30-300  0.013-0.08  0.5-2.4   
 Boring 19-300 60-900 0.025-0.13  0.4-6.3 yes 
 Grinding (internal) 25-300 75-900 0.004-0.02 0.1-1.6  yes
 Turning/Grooving 5-100 20-300 0.03-0.87  0.4-9.0   
 Grinding (external) 5-100 20-300 0.012-0.035 0.07-1.6  yes

Excerpt from Expert System rules developed for generation of operation sequences

  • IF Diameter is 3-100 mm AND Tolerance is 0.08-0.4 mm AND Surface finish is 1.6-9.5 µm THEN Drilling is recommended
  • IF Diameter is 10-100 AND Tolerance is 0.013-0.08 mm AND Surface finish is 0.5-2.4 µm THEN Reaming is recommended
  • IF Diameter is 19-300 AND Tolerance is 0.025-0.13 mm AND Surface finish is 0.4-6.3 µm THEN Boring is recommended
  • IF Diameter is 25-300 AND Tolerance is 0.004-0.02 mm AND Surface finish is 0.1-1.6 µm THEN Grinding(Internal) is recommended
  • IF Diameter is 5-100 AND Tolerance is 0.03-0.087 mm AND Surface finish is 0. 4-9 µm THEN Turning/grooving is recommended
  • IF Diameter is 5-100 AND Tolerance is 0.012-0.035 mm AND Surface finish is 0.07-1.6 µm THEN Grinding (External) is recommended.

Neural network architecture for automated selection of hole machining operations

pix4

The case study

pix3

Training error graph

pix5

Conclusions

  • The proposed CAPP methodology takes in as input the attributes of the features and automatically generates all the feasible alternatives for machining the feature in order of preference after taking their relative costs into consideration.
  • Some advantageous features of the neural network based CAPP methodology as against the knowledge based approach are:
    • Its efficient knowledge acquisition capability owing to its ability to implicitly derive the rules from sample machining cases presented to the neural network
    • Its capability to generalize beyond the original machining cases to which it is exposed during the training and face intermediate situations with reasonably good accuracy with respect to those proposed during the training
    • High processing speed once the neural network is trained.