Use Cases

Optimize Efficiency of Power Plants

NeuCo, a leading provider in electric power producer optimization,wanted to develop a way to model and optimize power plant settings to perform at maximum efficiency, increasing profits for the plants.

Optimize Efficiency of Power Plants

Solution: nD machine learning technology was used to model multiple types of power plant operations. Plant combustion processes were modeled in order to optimize plant control systems, leading to 0.75%-1.5% improvements in boiler efficiency while also reducing emissions by 10-15%. Furnace cleaning operations were also modeled and optimized, leading to efficiency improvements of 0.75%-1.5% along with improvements to emissions performance. These efficiency improvements of around 1% could often be translated into annual savings of approximately a half million dollars for the power plants using this system.

Discover Your Solution

Light Rail Power Circuit Analysis

A large urban transportation organization wanted to find the source of power spikes causing damage to power circuit hardware and gain better visibility into their DC power delivery system.

Light Rail Power Circuit Analysis

Solution: nD’s machine learning and visualization technologies were used for a proof-of-concept to evaluate billions of datapoints from prototypes of high frequency current and voltage monitoring instruments. The final solution will handle visualization and monitoring of millisecond-based data from across the entire train system efficiently and reliably to identify anomalies.

Discover Your Solution

Electricity Generation System Transformation Planning

Hawaiian Electric (HECO) wanted to understand how and when to value a wide variety of generating assets while they transformed Hawaii’s electricity generation to 100% renewable over a 20 year period. They turned to Black and Veatch, a leading engineering, consulting and construction company for consulting.

Electricity Generation System Transformation Planning

Solution: Black and Veatch used nD technology to analyze, visualize, and compare results from thousands of detailed electricity generation and dispatch simulations which combined to include trillions of data points. Information gained from these simulations was then used to estimate the value of distributed energy resources like rooftop solar, electric vehicles and demand response programs under different grid transformation scenarios. The result was a better understanding of how different transitional approaches would yield better cost savings for the company and its customers.

Discover Your Solution

Detecting and Classifying Oil Spills Using Remote Sensors

Researchers from Universita' di Perugina in Italy, Institute of Physics in Lithuania, and Universität Oldenburg in Germany wanted to improve on an established method to detect and classify oil spills on remote water surfaces using laser-based sensors.

Detecting and Classifying Oil Spills Using Remote Sensors

Solution: Previous research determined that oil spills could be classified using laser-based fluorescence spectroscopy by taking these sensors and comparing them to a database of known signatures using neural networks. nD classification technology was used to implement an improved learning system that was more resilient to adding new classifications to the database of oil fluorescence spectra than previous methods without sacrificing accuracy. This allowed the system to be continuously retrained without invalidating previous detections, resulting in a more robust system.

Discover Your Solution

Plant Equipment Monitoring and Diagnostics

Black and Veatch is a leading engineering, consulting and construction company working in such industries as power, telecommunications, water, and government/defense. They needed to automate and increase the scale of the asset management services, particularly the monitoring of real-time utility data.

Plant Equipment Monitoring and Diagnostics

Solution: nD machine learning technology was used to create and manage thousands of models of plant equipment operating parameters. These models were then used with real-time streaming sensor data to provide 24/7 monitoring to a variety of customers primarily in the energy sector. To date, a wide variety of emerging hardware and operational issues have been detected early, saving millions of dollars.

Discover Your Solution

Improving Mining Safety

As a researcher for the Council for Scientific and Industrial Research (CSIR) in South Africa, Teboho Nyareli wanted to increase safety in the mining industry by improving the subjective process of determining if walls in mines are intact.

Improving Mining Safety

Solution: Traditional mining wall testing involves tapping on walls and subjectively listening to the sound it makes. Experienced miners can usually tell the difference between safe walls and unsafe walls. nD classification technology was used to develop models that could take these sounds and determine whether a wall was safe based on limited samples. The classifier was then embedded into a prototype device which records the sound and processes it to provide immediate objective feedback. This initial success should lead to further refinement of both the models and the device.

Discover Your Solution

Detecting Impending Faults in High Voltage Power Systems

As the Principal Engineer for Artificial Intelligence (AI) at Qualitrol-DMS in the United Kingdom, Fraser Cook wanted to classify ultra-high frequency (UHF) electromagnetic discharges from high voltage power equipment to identify whether the unit was about to fail.

Detecting Impending Faults in High Voltage Power Systems

Solution: The most common causes of electrical failure in Gas Insulated Switchgear (GIS) generate Partial Discharge (PD) prior to dielectric failure of the plant. These PD pulses create resonances that can be detected in the Ultra-High Frequency (UHF) range. nD classification technology was used to develop the main classifier in a proprietary system to classify these UHF discharge patterns to effectively detect and determine the types of faults present. Based on a database of over 3 million samples, Fraser reported being able to classify the results with “extremely high accuracy” using his system.

Discover Your Solution

Forecasting Extreme Weather

As a Professorial Research Fellow at Central Queensland University focusing on environmental issues, Dr. John Abbot wanted to develop better medium- and long-term weather forecasts with a focus on extreme events for better emergency and insurance planning after severe flooding events hit Australia.

Forecasting Extreme Weather

Solution: Dr. Abbot was able to use nD machine learning technology to generate improved extreme weather forecasting models. Using the “skill level” weather forecasting metric, scores for the new method were in the range 25% to 80% while previous prediction methods were between -20% and 20% which was generally equivalent to regular climatology. Based on his research, the approach is being applied to broadening geographic areas in hopes of predicting extreme events before they happen to mitigate losses both for insurance companies and those insured.

Discover Your Solution

Determining Seabed Characteristics

Dr. Vladan Babovic of the Danish Hydraulic Institute wanted to develop an automated way to interpret data from Side Scan Sonar images of the sea floor to reduce reliance on skilled technicians and improve data collection.

Determining Seabed Characteristics

Solution: Side Scan Sonar produces gray-scale images corresponding to the acoustic reflectance of objects. It is typically difficult to recognize and classify objects based on a single feature, but spatial order of the gray level transitions gives “texture” characteristics to the image. Dr. Babovic was able to apply nD classification and feature extraction technologies to create a system to efficiently identify seabed materials like clay/mud, sand, eel grass and gravel without the need for skilled technicians. Furthermore, this process could be repeated automatically to track changes in the seabed over time as new scans were available.

Discover Your Solution

Reducing Soft Tissue Injuries in Athletes

MP Murphy & Associates Pty Ltd is a consulting firm that focuses on Information Technology and Risk Management. They were tasked by sporting clubs with optimizing player conditioning and training programs to reduce in-season injuries while improving performance.

Reducing Soft Tissue Injuries in Athletes

Solution: nD machine learning technology was used to model, analyze and optimize the training loads and conditioning reports of athletes on sports teams over the course of multiple seasons. This analysis has allowed medical and conditioning staff to better assess new player fitness, better predict in-season performance, and determine optimal activity levels to balance player performance and injury risk. And in one case, enabled a major sporting club to reduce player injuries by 57%, causing it to become the #1 rated team in the league in many player availability categories.

Discover Your Solution

Detecting Hidden Explosives and Bioagents

A team of researchers from the New Jersey Institute of Technology Chemical Engineering and Physics departments wanted to develop a method for detecting hidden explosives and bioagents in Terahertz (THz) spectral images in an effort to improve security.

Detecting Hidden Explosives and Bioagents

Solution: Terahertz (THz) radiation spans the far infrared range with wavelengths ~0.3mm and typically transmits through plastics and many other materials. Different types of materials reflect or transmit THz spectra in distinct ways. Using this information, the researchers processed images of scans at multiple THz frequencies that included hidden explosives and bioagents to determine if these objects could be detected. The team was able to use nD machine learning to effectively analyze these images and determine which frequencies and methodologies were most effective at detecting the lethal agents.

Discover Your Solution

Enhancing Educational Outcomes

As an instructor at Fort Lewis College in Southwestern Colorado, Cameron Cooper wanted to improve the outcomes for students enrolled in Developmental Mathematics courses.

Enhancing Educational Outcomes

Solution: nD machine learning was used to take input factors such as learning styles, responses to questionnaires, high school GPA and standardized test scores to classify students as potentially being at risk based on problems previous students encountered. Students identified as “at risk” were placed in a 1-hour per week support class which addressed issues of frustration tolerance, test anxiety, and much more. The addition of this class for these students increased the overall student success rate by 8% in the first semester it was implemented. Additional work was then done to expand the program to other areas of the college and nationally.

Discover Your Solution

Detecting Characteristics in Satellite Images

As a researcher at Metria Miljöanalys (an environmental consulting group in Sweden), Mats Rosengren wanted to use multi-spectral sensor data from satellites to extract statistical information and detect changes in forests and other land uses.

Detecting Characteristics in Satellite Images

Solution: Mr. Rosengren used nD classification technology to effectively differentiate between multiple classes which were often overlapping and sometimes had low separability. Based on these classifications, Metria Miljöanalys was able to create applications to map forest types, track changes in forestation and land use, and perform other environmental analyses. Along with government clients, forestry companies found these maps and analyses to be essential in their planning and forest management processes.

Discover Your Solution

Authorizing Radiology Services

As the Director of Medical Informatics at MedSolutions (a Radiology Benefits Management (RBM) company), Scott Perry wanted to help insurance companies cut costs and improve services by providing faster authorization for radiology services without sacrificing accuracy.

Authorizing Radiology Services

Solution: nD classification technology was used to generate models to determine whether radiological procedures (such as MRIs) were warranted given their location on the body and other data. Over 100 different models were developed using over 100,000 patient records. The resulting models had an error rate as low as 1%. Using these models, the response time for most insurance authorizations improved to minutes, which reduced internal cost for the insurance companies.

Discover Your Solution

Predicting Snowfall Depth

As a professor of Atmospheric Sciences at the University of Wisconsin - Milwaukee, Dr. Paul Roebber wanted to develop a snow density system to improve in the prediction of snowfall depths.

Predicting Snowfall Depth

Solution: nD prediction technology was used as an integral part of a snow density system which substantially improved over the sample climatology and is now used in official forecasts. In testing, the system correctly diagnosed over 60% of the snow event cases examined, which was a substantial improvement over the less than 42% correct using sample climatology and the less than 52% correct using the National Weather Service “new snowfall to estimated meltwater conversion” table. When expressed as forecasting skill scores, it offered an increase of 75% - 183% over the next most skillful approach.

Discover Your Solution

Predicting the Emplacement of Improvised Explosive Devices (IEDs)

Researcher Warren Lerner wanted to determine the feasibility of predicting potential Improvised Explosive Devices (IEDs) placement based on terrain features, lines of sight and target-vehicle traffic.

Predicting the Emplacement of Improvised Explosive Devices (IEDs)

Solution: For his doctoral dissertation, Warren Lerner wanted to determine whether it was feasible to predict the location of emplaced IEDs based on information related to historical IED detonation events, such as the presence of certain terrain features, visual and radio-frequency lines of sight and the volume of target-vehicle traffic during a 24-hour period. Using nD classification technology, initial tests on theoretical data correctly classified 85% of cases in flat areas and 87% of cases in mountainous areas. Based on these findings, it was concluded that there was a strong case for further research in this area.

Discover Your Solution

Classification of Text from Internet Searches

Researchers at Europe University Viadrina in Germany wanted to develop an automated solution to scan website listings for innovative businesses to include in a business database.

Classification of Text from Internet Searches

Solution: Europe University Viadrina in Germany has developed a database of business Internet applications. However, a method was needed to scan the rapidly growing number of websites to determine which businesses to include in the database. nD classification technology was applied to classify business applications as “interesting” or “not interesting” based on characteristics parsed from the company information. In initial testing, their best approach classified over 84% of new data correctly.

Discover Your Solution

Planning Autonomous Robot Movement

As a researcher at the Slovak Academy of Sciences, Danica Janglová wanted to develop an autonomous way to navigate a robot through a series of obstacles.

Planning Autonomous Robot Movement

Solution: nD machine learning technology was used to develop two key subsystems in testing an approach to effectively guide a robot through obstacles. The first subsystem used nD technology to process the data from ultrasonic sensors to create a virtual map of the environment as the robot traversed through the obstacles. The second subsystem used nD technology to develop a path planning algorithm to safely allow the robot to reach its destination based on its environmental awareness.

Discover Your Solution