TOFAŞ – CNC Cutting Tool Life Prediction System

TOFAŞ possesses a largely automated production infrastructure performing high-precision manufacturing with CNC machines. Yet, despite this level of automation, deciding when cutting tools should be replaced still relied on experience and operator intuition.

When tool life is exceeded:

• Product quality decreases,
• Scrap rates increase,
• Unplanned stoppages occur,
• And production efficiency is seriously affected.

TOFAŞ approached us with the need for a system that could predict these issues in advance.

Our goal was to develop a system that could predict in advance when the cutting tools used in CNC machines would reach the end of their life. To that end, we built a structure capable of analyzing data generated during production.

Core components of the project:

• Sensor data collected throughout tool operation (vibration, temperature, power consumption, etc.) was integrated
• Algorithms performing time-series data analysis were developed in Python
• Tool wear curves were derived and measurement-based decision models were built for life prediction
• An Angular-based interface allowed maintenance teams to access these predictions in real time

The main issues we addressed with the project were:

• Product quality dropping because tool changes were done too late
• Unplanned stoppages causing inefficiency on the line
• Maintenance plans being based on experience rather than foresight

With the system we developed: • Dedicated prediction models were created for each tool • Maintenance teams now know in advance when a tool needs to be replaced • Intervention is possible before a breakdown occurs • Unplanned downtime has been minimized

Once integrated into the production line, the system delivered the following gains:

• Tool life management was digitized and became independent of personal experience
• Scrap rate decreased
• The number of unplanned stops on the CNC line dropped significantly
• Maintenance costs were saved
• Teams began making tool change decisions based on measurable data