Computer File
Penggunaan jaringan saraf tiruan backpropagation dalam otomasi sistem inspeksi visual
In globalization era, reliability and quality have become ones of the
important factors to win the competition. Therefore in most mass-production
facilities, 100% inspection of all parts and products is often attempted. As a
result, the inspection process is often the most costly stage in manufacturing.
In General inspection system is divided into two inspection categories:
quantitative inspection which is deterministic and qualitative inspection.
Quantitative inspection involves measurements of product dimensions and
compare them with a standard product, while qualitative inspections involve
human senses - mostly human vision - as a measuring instruments. Therefore
qualitative inspections are more about the measurement that is not deterministic
and tend to be subjective. Such characteristic is called sensory characteristic.
This under graduate thesis explains and examines the use of neural
networks backpropagation to solve a problem which involves the sensory
characteristic. Backpropagation neural network is chosen as a loo1 thanks to its
ability to deal with undeterministic problems. For that reason, a software called
Automated Visual Inspection System or AVIS is built.
An experiment has been done in laboratory. The case study used in this
experiment is visual justification of the shape of painted surface quality of an
otomotive component: End-tie-rod. Firstly, optimal parameter for
backpropagation neural network must be determined through some experiments
during training mode. To be able to recognize bad and good products, the network
must be trainned through input and output patterns, built from some image pixels.
The neural network has been tested by using two sets of data test. These
sets have never seen by the neural network before. One of them is data test in
which part orientation is rotated by one degree. Although the test shown satisfying
result, the neural network has one weakness that it needs a lot of training samples.
Barcode | Tipe Koleksi | Nomor Panggil | Lokasi | Status | |
---|---|---|---|---|---|
skp20686 | DIG - FTI | Skripsi | TI DAF p/97 | Perpustakaan | Tersedia namun tidak untuk dipinjamkan - Missing |
Tidak tersedia versi lain