IOT implementation for remote management.
Sector : Energy
A major global oil firm needed to improve its monitoring capabilities in a country where its staff and resources are vulnerable. The goal was to prevent the theft of oil while introducing the ability to rapidly pinpoint leaks and minimize human intervention. Additionally, many of its oil wells are located in challenging areas, trapped thousands of meters underground. The company was looking for new technology to address all of these issues.
Internet of things (IoT) technology was proposed to connect, control devices and collect data. To help the operations run more efficiently, the client turned to Smart Field technology installed thousands of sensors on its equipment, such as valves and pumps. It captured data on temperature, pressure, and other measurements, and sent it out to control centres. Engineers read the measurements and monitored production in real-time so they can optimize each individual process.
Automation using Cloud, Mobile and integrated Sensors
Sector : Energy
A global industrial automation company was tackling the complexities of the oil industry with a solution that helps monitor every step of the petroleum supply chain. A single pump failing in an offshore rig can halt operations and cost $100,000 to $300,000 a day in lost production. With sensors, software and the cloud, these disparate assets are difficult to manage and monitor, resulting in huge inefficiencies.
To avoid those losses, the company outfitted its pumps’ electrical variable speed drives to Microsoft’s Azure cloud so they could be monitored in real-time, providing readings for pressure, temperature, flow rates, and other measurements to engineers. Sensors and other assets became part of a Connected Enterprise, powered at its core by a rich flow of data.
The company is also working to make gas pumps smart by installing cloud gateway appliances at each station.
Data collected through these pieces of equipment can now be securely sent to a cloud platform and fed into a dashboard that can be easily viewed on a PC and mobile platforms.
Machine learning techniques delivers 82% jump in forecast accuracy
Sector : Telecom
When issues are raised by customers regarding telecoms services, their details are logged by a service team that arranges engineer visits. The service team also takes into account a seven-day weather forecast as this may influence the number of tickets raised or ability to conduct visits. These forecasts are often inaccurate, causing huge inefficiencies and impacting customer satisfaction.
We used a machine learning framework to analyse the weather model for telecom tickets. The linear model suffers from low Adjusted R-squared which has been enhanced by introducing new features like the number connection of fibre and copper in the data set and using GBM model.
The company needed to increase the ability to predict the number of trouble tickets in extreme weather/climate conditions as well as normal weather days. The dataset provided by the client a was a log of trouble tickets due to weather conditions as well as weather forecasts collected from different weather stations over the past three years.
Pre-processing and Exploratory Data Analysis (EDA):
Following analysis of the data sets, missing values were fixed and unrelated information was removed to focus on relevant features. Weather data was merged with daily ticket data and variables are closely examined to find the relationship with the prediction. Extreme weather days are identified and data value are explored to find any possible reason for high ticket volumes.
The clean data was split into train and test with 70:30 ratio. We built a multi-linear regression model with moving average of weather variables. The adjusted R^2 has shown result for normal weather date vs. extreme weather date as 57% vs. 51% for fibre connection and 73% vs 68% for copper connection respectively.
This result indicates that we needed to identify other important variables relating to ticket numbers so we re-engineered the analysis for a more accurate prediction.
We noted that the number of tickets within the same global region, under same weather forecast condition are varying. The reason is directly related to the number of relevant fibre and copper connections.
We added this feature in the data set and have used GBM with fine tuning parameter for regression prediction and found the more accurate prediction for both weather and extreme weather date.
As a result of our analysis and recommendations, accuracy of weather predictions reached to 82% and 79% for extreme weather condition dates. This can be compared with the result of existing in-house model near to 50%.
Implementation of new HR system by Major Bank
Sector : Banking
Our client was one of the largest private sector banks in India with over 23,000 employees. The HR department was using spreadsheets to calculate and process salaries at the end of every month. Completing this onerous manual task demanded a huge dedicated team of 20+ FTEs. As well as the high costs, the existing salary processing process was highly prone to human error.
We implemented an automated Corporate Salary solution with the help of RPA solution. As part of this, we implemented eight bots to automate end to end salary processing function.
- Reduced the average handling time from up to 30 minutes to less than 12 minutes.
- FTE count reduced by 50%
- High BOT accuracy of 99.99% was observed in production environment.
- ROI after implementing RPA automation was achieved in just four months.