Detect the position of anomalies in petrol pipelines using data generated from a maintenance device called "Pig" that travels inside the pipeline.

• Detecting Wielding
• Visualising Metal loss and or dents

• Data is in binary format saved in - 'captured_data.bin' file, captured from 84 sensors for 50_000 milliseconds
• Pigs directly captured/stored data in binary(.bin) format

• Data varies with time i.e., Time Series data
• We can also see a repeated pattern of flux at times in all sensors.

• Normalised data for comparison purposes.
• Aggregated(sum) all sensors to amplify anamoly.
• The z-score method is effective here in comparison to the Inter-Quartile Range method as per the observations.
• Edge detection using Sobel, OpenCV.

• Approach 1: Using Various Statistical methods like InterQuartile Range, Z-score, etc.
• Approach 2: By Filtering anomalies with Sobel Derivatives, Image Processing in OpenCV.

Approach 1 | Approach 2

We can detect the position of welded iron and also visualise the metal loss.

Little Business Understanding

• Many pipes are welded together to form a pipeline.
• Pigging is a concept in pipeline maintenance that involves the use of devices known as pigs, which clean pipelines and are capable of checking pipeline conditions from inside.
• Pigs inspect pipelines by receiving magnetic flux from the walls of the pipe.
• Applications of Pigging include oil & gas, lube oil, chemical plants and hygienic applications such as pharmaceutical or food, etc.

• Each "Pig"(here) has 84 sensors in circumference that capture data every millisecond while travelling in pipes. So this becomes a time series problem where each traversal time for each independent pipe is 5_000 milliseconds.