Vashishtha, GovindChauhan, SumikaGupta, Munish KumarKorkmaz, Mehmet ErdiRoss, Nimel SwornaZimroz, RadoslawKrolczyk, Grzegorz M.2024-09-292024-09-2920240268-37681433-3015https://doi.org/10.1007/s00170-024-14336-7https://hdl.handle.net/20.500.14619/3831Surface quality, machining efficiency, and tool life are all significantly impacted by tool wear in metal cutting machining. Research priorities and areas of focus in tool wear are shifting as intelligent machining becomes the norm. Unfortunately, there are currently no acknowledged most effective ways for analyzing tool based on the energy signals specially in the machining of titanium and its alloys. In the present work, the titanium machining was performed under different lubrication conditions such as dry, minimum quantity lubrication (MQL), liquid nitrogen and hybrid, etc. Then, the spectrograms are used to transform the acquired energy data into time-frequency features. Starting with a set of randomly generated hyper parameters (HPs), the long short-term memory (LSTM) model is fine-tuned using sine cosine algorithm (SCA) with loss serving as the fitness function. The confusion matrix provides additional validation of the 98.08% classification accuracy. Additional evaluations of the suggested method's superiority include its specificity, sensitivity, F1-score, and area under the curve (AUC).eninfo:eu-repo/semantics/closedAccessTool wearMachine learningSignal processingData acquisitionEnergy signalsArticle10.1007/s00170-024-14336-72-s2.0-8520308294435737-8Q13561134WOS:001311863300018N/A