Measurement and analysis of duty cycle of construction equipment is essential from the perspective of making decision with regards to controlling idle time and realizing productivity improvement. However, current monitoring techniques like Vehicle Health Monitoring Systems (VHMS) are either too expensive and/or are not compatible with outdated equipment fleets and equipment across different manufactures. To address these issues, we aim to develop a non-invasive technique of using a smart phone to measure the various activity modes (e.g. wheel base motion, cabin rotation and arm movement for excavator) and subsequently duty cycle of construction equipment. The smart phone is mounted inside the cabin of construction equipment to automatically capture engine vibration signatures in form of three-dimensional acceleration. Various time and frequency domain features are extracted from this raw data and are tested and classified into different equipment actions using machine learning algorithms from WEKA (Waikato Environment for Knowledge Analysis) data mining set. The classification accuracy on a random sample generated from various experiments on hydraulic excavator (CAT 330CL) was turned out to be between 72-86%. The average cycle time measurement accuracy based on predicted labels for equipment actions was around 88.5%. This result demonstrates the potential use of the proposed technique as an affordable system for automated and real time measurement of construction equipment duty cycle to facilitate detailed productivity analysis.