signal and noise
Big Data — Verifiable Results
Airglide analyzes environmental and ship operating data to verify installed operational savings in reduced fuel consumption resulting from our Airglide™ Air Lubrication Systems.
A challenge for the ship owner or operator has been understanding the actual contribution of the ALS systems once installed.
To measure the performance contribution, data must be collected. In collecting this data, the two primary domains of data collection are found to be very noisy and complicated:
External Environmental Noise includes factors such as:
wind direction and velocity
ship heading changes
hull and propeller fouling
hull coating roughness
impact of diver cleaning maintenance
Internal Ship Noise includes on-board factors like:
engine and turbocharger condition and operating range
fuel transfer systems
auxiliary fuel consumption
bearing and shaft sealing status
fuel quality and type (ECA operation)
scrubber or exhaust waste heat energy recovery engine back pressure
This wide variety of exogenous data creates an extremely noisy data-collection environment. In collecting data to measure the actual reduction in ship drag and fuel consumption, statistical analysis and synthesizing of the huge amount of data is literally an area of intense research and development today.
Airglide, with its staff of world-class Ph.D. researchers, engineers, and programmers is leading the maritime industry in developing the tools that provide this answer to you in a timely manner with accuracy and precision.
"BIG DATA" = NEW ANALYSIS PARADIGM
The term “Big Data” has become an industry cliché and the relevant meaning is often lost. Unique and significant differences exist in analyzing “Big Data” compared to traditional analysis of “small data.” The
Traditional statistical analysis revolved around statistical significance and typically used the smaller sample data set to represent a larger data system. Today, in “Big Data” with advanced computing and storage power, we have access to either all, or the majority, of the data that represents the entire system. Statistical significance is not so relevant.
The data analysis is significantly different and becomes more a filtering or searching process rather than a modeling or scaling process.
Big data analysis is unique from historic data analysis in that new features become predominant, including Heterogeneity, Noise, Spurious Correlation, and Incidental Endogeneity versus Exogeneity Assumption.
Small data typically emanated from a single or few sources and the data was homogenous or relating to one population. In small data, an outlier is disregarded or ignored. In "Big Data", the number of different data sources provides the opportunity to consider sub-populations.
The massive amount of data in "Big Data" can lead to data noise that can mask important and relevant correlations.
c. Spurious Correlation
Alternatively, again due to the massive amount of data, incorrect correlations can be found based on data sets that occur not because they correlate but rather due only to the quantity of occurrences.
d. Incidental Endogeneity versus Exogeneity Assumption
Many of the statistical tools available in prior analysis with small data are not available in "Big Data" analysis due to the small data sets being able to correlate to a variable external to the data set and in "Big Data" the correlation is likely to a variable inside the data set.
The Ph.D. researchers, engineers, and programmers of Airglide are leading the application and implementation of these Big Data maxims to providing you, the ship owner or operator, with verifiable results.