F19 – 51 Integrated Cleaning and Inspection Pig Localization Test Bench

Overview

In the oil and gas industry, it is often necessary to clean and inspect pipelines. This is done using what is known as a pipeline “pig”. A standard pig is often made from polyurethane and is meant to travel the length of the pipe and remove any debris or buildup. There also exist “Smart” pigs, which can scan the pipeline walls for corrosion or cracks. The issue with these pigs is that once a defect is found, it is often hard to distinguish where in the pipeline it occurred. Since most of these pipelines exists underground and are made of steel, it is likely that many places in the pipe are unable to receive GPS data. Therefore, a different method must exist for keeping track of the pig inside the pipeline, and that is what our solution focuses on. 

Our prototype makes use of what is known as an Inertial Measurement Unit (IMU), which contain both accelerometers and gyroscopes. By collecting acceleration and gyroscope data while the pig travels through the pipeline, we can perform a series of integrations to determine the position of the pig. However, this method alone is usually not accurate enough to give a good approximation of position. Our solution also makes use of a proximity detector, which can detect small changes in the distance to metallic object. By combining this data with our IMU data, we can perform multiple runs and using resampling techniques, are able to combine the position data across all runs in order to get a better approximation.     

The design approach for this project breaks down into three categories: test vehicle, test bench and code development. Each area had different challenges and different steps along the way to reaching the final solution. For each, the design decisions were driven by testing, development and client feedback. None of the initial designs from the beginning of the fall semester made it to the end without major adjustments.   

Test Vehicle Design  

The test vehicle went through many changes and iterations throughout the course of the project. We had originally planned to have a person drive the vehicle. We found that a human driver left to much uncertainty in the path of the vehicle. To achieve a repeatable path, we replaced the human driver with a pin and slot system, like what is found in a slot car.  

The selection of sensors, micro-controllers and battery packs where a vital part in the vehicle design. Working with the client and using information from past groups, we determined the use of an Inertial Measurement Unit (IMU) would be the most appropriate sensor to collect localization data. Proximity sensors where determined to be best option to collected data from the steel portions of the track to pair with the IMU data. The next step was determining what micro-controllers would provide the most efficient methods to store the data and drive the motors. A Raspberry Pi was selected to collect data from the IMU and was paired with an Arduino that drove the motors and proximity sensors. With the type of components selected, the next step in the design process was to design a sensor platform.     

The stock platform on the test vehicle was not large enough to support the multiple components needs for data collection. A design was developed with each sensor in mind. Our sensor platform design had the needed space for each IMU, proximity sensor, controller and battery pack, while keeping the weight equally distributed across the vehicle frame.   

Mechanical Test Bed (Track) Design   

The test track went through many conceptual changes before the final version was built. Due to limited materials, there was only one opportunity to build a full-scale test track. Initially, the whole track was going to be built from steel, but due to cost, issues with storage, transportation and fabrication that idea was quickly set aside. Working with the client, it was determined that a wooden track with the slot design to match the pin on the vehicle, with sections of steel, would be more fitting with the localization aspect being of greatest importance. A modular design was also important so that the test bed could easily be moved to different locations and arranged in different shapes.    

Code Development  

There are two distinct sections of the software in this project. First, the raw data from the sensors is collected and stored for processing. Second, there is the localization algorithm that estimates the position of the test vehicle using the raw acceleration and gyroscope data. At the start of the project, the team overcame multiple challenges when interfacing the sensors with the Raspberry Pi. A successful implementation was completed by the end of the first semester. 

The next challenge was data processing. Determining the vehicles position from raw acceleration and gyroscope data was a difficult challenge. After examining several approaches and working with advisors, the team settled on the final approach of integrating the acceleration data to obtain a position. In addition, the gyroscope data was combined in order to orient the test vehicle in 2D space. This position data was then resampled and combined across 10 runs in order to provide a better estimation.