Automatic Identification of Drill Condition During Drilling Process in Standard Laminated Chipboard with the Use of Long Short-Term Memory (LSTM)

This paper presents a new approach to the identification of the state of drill wear and tear during the drilling process. Identification provides for three states (classes) of drill which were marked according to traffic lights classification: red - “useless” (worn out), yellow (still sharp but should be manually assessed by a specialist), green -“useful” (sharp). Red state indicates symptoms of drill worn out and determines that it should not be used during the drilling process (does not satisfy the furniture process quality). Yellow drill state should draw attention to the fact that the drill may be excessively blunt. Later on, it is recommended to assess the drill manually. When the system indicates the green state of drill, it means that the drill may still be used in the drilling process, maintaining the high quality of drill holes. The recognition of drill state is based on 5 registered signals: feed force, cutting torque, noise, vibration and acoustic emission. The hardest part in the implementation of automatic identification of drill condition is to find appropriate features, which will separate cases belonging to 3 classes (states). When input data are presented in the form of images, we can use the convolution neural network, but in that case we have signals registered as time series, hence the long-short term memory (LSTM) is applied.