Design of Predictive Tool Condition Monitoring System for Micro Drilling

  • Norshah Afizi Shuaib

Student thesis: Phd


Tool Condition Monitoring (TCM) has been widely practised to identify the tool state during machining. Numerous research into TCM has been conducted especially at conventional or macro-scale level, but limited effort has been put into TCM at the micro scale, specifically identifying the ultimate tool life stage. When a micro twist drill breaks, any broken segments that are lodged inside a borehole are almost impossible to be extracted, hence leading to a possible workpiece loss. This risk sometimes urges the operator to change the tool after only a very short tool utilisation. This research deals with the design of a tool condition monitoring system that is conducted on an ultra-precision micro CNC machine at the University of Manchester. First, several machine tools and process related issues are investigated in their relation to provide a suitable environment for the micro drilling tool life tests. The aim is to obtain a sufficient tool life so that an algorithm will be generated from a tool that fails due to deterioration i.e. tool wear. The next task involves the feasibility assessment on a set of sensors that have been selected, which are hall sensors for both the machine’s main spindle and the spindle-carrying z-axis, an acoustic emission sensor and an accelerometer. Research addresses aspects such as the signal stability with respect to sensor mounting location, and the relation between hall sensor signal and the direct cutting forces measurement using a dynamometer. The next stage includes the tool life tests including data acquisition using the aforementioned sensors. In order to understand the signals’ patterns, a visual inspection of selected tools was carried out to establish relationships between signal and the actual wear condition of the tool. The assessment continues with the selection of signal features for the TCM algorithm. These are chosen based on the signal trend towards the event of tool breakage. Eventually, from a number of several signal features, four of them are carried forward into the decision process. The algorithm learning starts with the determination of signal feature ranges, which are found to be different for each tool. Due to this reason, the threshold value for a warning issued by the system is made to float, based on each individual tool’s starting value. Several other algorithm rules including the sensitivity, the warning count, and a decision for a process to stop are made thereafter. The aim is set to achieve the highest possible tool utilisation. Finally, the algorithm is tested using a set of independent tools, which reveals that the TCM system designed was able to save two out of three tested tools, at tool utilizations of 61% and 84%. Further improvements are suggested and tests against a number of drilling tools, in order to enhance the system’s capability in realising higher tool utilisations without allowing tool breakages to occur.  
Date of Award1 Aug 2019
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorSrichand Hinduja (Supervisor) & Robert Heinemann (Supervisor)

Cite this