Each idea is scoped for a two-semester project, with the tools you will need and a suggested plan. Adapt it with your supervisor.
Train a convolutional neural network to detect and measure cracks in concrete from photographs, to support faster building and bridge inspections.
Convert waste cooking oil to biodiesel and optimize catalyst amount, methanol ratio, temperature and time for yield and fuel quality.
Design a low-cost house using courtyards, shading, thermal mass and ventilation to cut indoor temperatures, and compare it with a typical local design.
Use five to ten years of published accounts to test whether models such as the Altman Z-score predict financial distress for PSX-listed firms.
Measure household or lab energy use in real time, show it on a web dashboard, and estimate which appliances are running from the load profile.
Collect water from taps, filtration plants and tube wells around Multan, count and classify microplastic particles, and compare sources.
Forecast production from historical well data, comparing classical Arps decline-curve analysis with machine-learning models.
Predict how many operating cycles an engine has left from multivariate sensor data, using the public NASA C-MAPSS dataset, and compare classical and deep-learning models.
Build a low-cost vibration logger for a motor or pump, detect imbalance, misalignment and bearing faults from frequency spectra, and classify them with machine learning.