Monitoring Keausan Pisau Chipper Menggunakan Analisa Audio Signal Dengan Metode Expert System Pada Proses Wood Chipping
Benedictus Adityo Imanuel Kurnia, Ir. Muslim Mahardika, S.T., M.Eng., Ph.D., IPM., ASEAN Eng. ; Dr. Veronica Lestari Jauw
2025 | Tesis | S2 Teknik Mesin
Wood chipping process is one important part in pulp and paper industry. Nowadays, there are still a lot of the processes being done manually, which rely too much on operator’s expertise. One of them is on determining chipper knife wear-out. This thing can cause disparities on decision making on chipper knife replacement, which can make problem occurs in production line. This research was conducted to characterize knife sharpness and wear-out using micrograph and sound signal. From that characterization a model then developed to do monitoring on knives wear-out. Micrograph characterization was being done by taking micrograph using Scanning Electron Microscope (SEM). The micrograph shows different characteristics. Sharp knife still has sharp edge marking from knife grinding process, while worn-out knife already lost that marking, even at some parts deformation has been found, results from abrasion and friction when come in contact with logs that generates heat, also contact with other material such as sand, soil, and stone. Aside of that, this research also examines chip quality, power consumption, and productivity data, starting from the knives still sharp until worn-out. Hourly data that already collected indicating that on the fourth hour the knives already started to worn-out, while on the seventh hour the knives already worn-out. Sound signal for sharp knife, half-worn, and worn-out then collected on first, fourth, and seventh hour of chipping using microphone and then processed using LabVIEW software to get the result of time domain and frequency domain data. This frequency domain data then becomes the input for expert system model, consisting of Artificial Neural Networks (ANN) and Support Vector Machine (SVM) using Matlab software. The result was these two models can successfully identify chipper knife wear-out, with SVM accuracy was better than ANN on determining sharp knife, while ANN was better than SVM on determining half-sharp and worn-out knife. Combining these two best traits, combined model can give final accuracy of 65,9%.
Kata Kunci : pencacahan kayu, keausan, karakterisasi, gambar mikro, sinyal suara