IoT & Sensor Data Analytics

With the rise of the Internet of Things (IoT), industries are awash in petabytes of Big Data from an increasing array of wired and wireless sensors. Gartner estimates that 20.8 billion connected items will be in use worldwide in 2020. These sensor networks are continuously monitoring and reporting on essential information like heart rates in a patient, production flow in an oil rig, or turbine wear in a hydropower plant. We help industries take advantage of this wealth of data by creating models to unlock the wisdom embedded within — to predict disease, reduce maintenance costs, anticipate equipment failures, etc. –driving a knowledge transformation of business operations.

IoT and Sensor Data Analytics Applications

We have extensive experience in the following industries helping our clients wade through high-volume data from sensor networks to discover actionable insights:

• Healthcare & Medical Devices
• Energy & Utilities
• Telecommunications
• Industrial Automation
• Automotive
• Aerospace
• Defense and Intelligence

Case Studies

With vast sensor data from connected devices, software usage logs, and equipment monitors, there are ample opportunities to deliver value. Whether you are new to sensor analytics or looking to augment existing capabilities, we can provide support where you need it most. Examples include:

Disease Event Detection Using Medical Sensor Data

A medical device manufacturer designed a sensor that is implanted within a major organ to detect disease events. The sensor is being tested on animal subjects with the intent to move to human trials and secure FDA approval for monitoring high-risk patients. We were engaged to provide machine learning models using the sensor data to identify abnormal activity predictive of the disease. The goal was to classify sensor traces as either normal or high-risk.

Results: Our ensemble model delivered a very high holdout sample accuracy of 96.9%, sensitivity of 93.8%, specificity of 98.9%, and specific animal accuracy between 93.9% and 99.4%. This proved that the device was better than the published state of the art at predicting the disease event in animals. As a result of this success, the device moved to human trials.

Using Sensor Analytics to Predict Natural Gas Well Freezing

We harnessed 20 years of detailed but noisy sensor readings from hundreds of gas wells to characterize transient well states for an international oil and gas exploration firm. The goal was to predict gas well shut-ins (blockages preventing production) four to six months in advance in a North American field, where winter months allowed only very limited access.

Results: Our model was far more accurate than the existing heuristic models and it allowed operations personnel to know which clusters of well pads should be prioritized for treatment. A second model was created which clearly showed when in the new well life cycle a plunger pump system should be installed. The estimated return on investment was less than a year.

Product Usage Analytics Improves Software User Experience

We successfully ingested over 1TB of software usage log data to build a user segmentation model for a major software client. We also created a custom visualization tool allowing the client to visualize and understand their log data in a way that was previously impossible.

Results: User segments defining unique software user personas were reproducible with 92% accuracy and have proven extremely valuable in helping SolidWorks understand user needs and behavior. Product Usage Analytics enabled SolidWorks to better understand, communicate with, and serve their users, facilitating more effective customer engagement and outreach.