Choose your language:

Hong Kong
New Zealand
United Kingdom
United States

Implementing a Predictive Analytics Framework to Forecast Equipment Failure


Download PDF

A leading geophysical services company sought TEKsystems’ partnership in managing the high volume and velocity of their data, as well as developing predictive analytics to save time, money and resources.

Client Profile

The client, headquartered in the United Kingdom, provides seismic data acquisition services worldwide. TEKsystems has partnered with the client’s parent company since 2012.

Technologies Supported

Oracle Exadata x5-2 Data Machine, Oracle Exalytics x5-4, Oracle Business Intelligence 12c, Oracle R / Oracle Data Miner

Industry Landscape

The science of earthquakes, known as seismology, studies the interior of the earth. Seismologists also examine the impacts of other shocks, such as those caused by tsunami, drilling or volcanic eruption. Through the use of seismic technology, scientists create and then listen for vibrations in order to build images of the earth’s core. This analysis can uncover the location and magnitude of past disruptions in the earth.

Construction firms, oil and gas utilities, miners, manufacturers, governments and other businesses rely on this data to reduce risk and optimize the services they provide. These companies seek innovative technology to stay at the forefront of seismic data acquisition and analysis. The ability to receive and review the vast amounts of data, including insight into predicting equipment failure, is a competitive advantage, especially as oil prices decline and budgets for exploratory expeditions shrink.


Our client, a leader in marine and land seismic technology, enables the exploration and development of off-shore areas. For this engagement, the client deployed nine vessels in remote seas to conduct subsurface surveys over a 90-day period or more. These surveys create imaging of the ocean floor, and the reports used will support future mining and drilling decisions. Each vessel deployed 20 streamers, and sensors on each streamer collected approximately three trillion records at a rate of one gigabyte every three seconds. Sensors beamed information into a control room in near real time, and data scientists reviewed the output on large monitor screens.

The client had three main concerns with this arrangement:

  • First, the vessel’s control room contained multiple screens that all needed to be visually inspected. As streamers collected data, data scientists reviewed the screens for exceptions, which could indicate equipment failure. Any exceptions were recorded and investigated later, which took additional time to analyze. The high volume and velocity of the data, combined with the visual need for inspection, raised quality control concerns for the client.
  • Second, if a sensor failed, data would not be collected and the line would need to be recaptured. (The data was read in the control room in sequential horizontal lines, left to right and then right to left, repeatedly, as data was captured.) Sensor failure resulted in loss of time and data, which in turn increased the cost of operations.
  • Third, if there were any malfunctions in the equipment (not detected by the data analyst monitoring the process), the data captured would be corrupted by the noise generated by malfunctioning equipment. This would result in having to recapture the data for the entire line.

As a result, the client sought a new solution that could better accommodate the high volume of data being recorded by the vessels; perform real-time analysis and reporting of the data; and deliver algorithms to predict equipment failures and data quality issues well in advance to save time and resources.


The client was looking for a large partner and was open to solutions that could solve their problems with data volume and velocity. Based on past relationships with the parent company, TEKsystems and Oracle proposed a joint solution to meet the client’s needs. We explained that data monitoring in the control room, with a heavy reliance on visual inspection for exception reporting, could be improved through predictive modeling. The client was interested in our approach to this process improvement, as quality control was also a concern.

TEKsystems’ solution included the following components:

  • We proposed the implementation of Oracle Exadata Database Machine, due to its storage, speed of data ingestion and data retrieval along with the solid state machine required for deployment on the vessel. Exadata server uses a direct Oracle Call Interface (OCI) approach through C++ programs in order to achieve a one gigabytes per second ingestion rate, an improvement over the required rate of three gigabytes per second.
  • We would create a data model to support this faster ingestion rate as well as a faster data retrieval rate.
  • We would implement predictive analytics modeling using Oracle R and Oracle Data Miner technologies. Our modeling would be able to predict the possibility of equipment failure within 16 hours. We built two use cases:
    • Air gun leakage. In addition to using sensors to develop images of the ocean floor, the client used an air gun to look for cracks in the sea bed. At the surface of the ocean, the air gun pumps and forms an air bubble. The bubble travels to the ocean floor and bounces back, and the journey of the bubble is used to plot the depth and surface of the ocean. Air gun leakage may eventually result in the failure of the air gun and certainly the quality of the seismic data captured is perhaps not accurate.
    • Fiber optics failure. Fiber optic cables (the streamers) contain multiple sensors that transmit data back to the control room. Sensor failure represents a loss of approximately eight hours of work, and visual interpretation of data patterns is essential to interpret an exception and uncover equipment failure.

As part of delivering the solution to customer, we had a data scientist, solution architect and a technical lead on site along with technical analysts in our off-shore facility. Our Project Management Office (PMO) would track the project throughout its life cycle to ensure best practice methodologies were employed.


We implemented our pilot solution over a 10-week period. The TEKsystems team conducted a discovery sessions with key stakeholder and technical architects of the customer in week-1 to finalize the requirements and objectives for the POC and then introduced the data modeling component over the first three weeks, followed by the data ingestion phase and predictive modeling. At the conclusion of the 10-week period, the client conducted evaluations and a steering committee reviewed the work done to date. The steering committee determined the program was a success.

We achieved the following key wins:

  • Built predictive modeling to support requirements around detection and prediction of equipment failures and were able to detect equipment failures at least eight to 16 hours in advance, compared to the existing solution in place involving manual data monitoring. This could result in a huge potential cost savings by doing preventive maintenance through predictive analytics.
  • Designed a model to analyze data quality patterns and alert data quality issues in real time. This significantly drives down the number of people required to perform manual monitoring of data on big screens and also eliminates dependency on individual skills to interpret the data for quality issues.
  • Achieved data ingestion rate of one gigabyte per second, versus the expected one gigabyte per three seconds.
  • Developed an on-board reporting model for data analysis, which reduced the cycle to time for data analysis from three weeks to on demand near real time.
  • Achieved offline reporting; implemented Oracle Business Intelligence tool for the 10-week period.

Achievement of these wins enabled a significant competitive advantage for the client. The client decided to implement and operationalize this solution for vessels going forward, including TEKsystems in the process.

Key Success Factors

Oracle expertise

The client required a technical resolution to a complex problem and, to add further complication, the solution had to be deployed on a vessel operating 24x7 in a remote location. TEKsystems’ expertise and understanding of the Oracle suite of services allowed us to introduce a comprehensive offering that included modeling, analytics and reporting to meet the client’s needs for rapid data ingestion, as well as real-time detection and prediction of equipment failures.

Domain expertise

TEKsystems has deep domain expertise in manufacturing and specifically within the oil and gas industry. We have helped numerous oil and gas clients in implementing their analytics solutions by leveraging our functional knowledge expertise, combined with bleeding-edge technology stack, to solve their business problems.

Data science practice

The client needed to capture and analyze quality data for the entirety of the survey, but the appearance of a failed streamer or air gun would derail the entire survey and cost in the millions to recoup the effort. The data scientist was an essential member of the project team. More than just a data analyst, the data scientist was able to conduct predictive analytics. We conducted two use cases (fiber optics failure and air gun leakage) to help the client mitigate project risks and predict the failure of equipment in advance.

Client relationship

TEKsystems had partnered with the client and Oracle on past business intelligence initiatives for more than four years. When Oracle learned about the client’s unique situation, Oracle recommended TEKsystems as the implementation partner. We were able to produce an agile response that also fulfilled our customer’s requirement.

Send Us a Message
Choose one