ADMI, LLC 3620 Pelham Rd. PMB 351, Greenville, SC 29615    (843) 513-2130
Surface and Groundwater

Data Mining is being applied to an array of problems related to the interactions between natural and man-made systems. These interactions are becoming increasingly important as growing populations and development place heavier burdens on our environment. ADMi has been successfull in applying data mining to an array of problems related to the interactions between natural and man-made systems with both surface water (streams, lakes, rivers, and esturaries) and groundwater systems.

Common issues encountered are:
  • Many stakeholders - federal and state agencies, utilities, communities, environmental groups
  • Competing interests with potential disagreements
  • Varying scientific/computer know-how amongst the users
Sample Project 1

Beaufort River TMDL

Needs: Determine Total Maximum Daily Load (TMDL) to support permitting process for three wastewater reclamation facilities (WRF) operated by Beaufort-Jasper Water and Sewer Authority and the U.S. Marine Corps.

Solution: ADMi and USGS developed a model to determine TMDL which simulated the impacts of the WRFs' BOD and NH3 discharges, and the non-point source rainfall runoff on river’s dissolved-oxygen concentrations under widely ranging tidal and weather conditions

Results: The model, developed using artificial neural networks (ANN), was packaged in an easy-to-use decision support system (DSS). The DSS and was found to be unusually accurate and defensible when compared to traditional finite-difference estuary models, allowing the new permits to be issued approximately 2½ years after the project’s start. Projects of similar complexity in the Myrtle Beach and Charleston areas required approximately 10 and 15 years for permits to be issued.

Sample Project 2

FERC Re-Licensing of Hydroelectric Facilities

Needs: Re-licensing of six hydroelectric facilities on the Pee Dee River and protection of fresh water intake in Myrtle Beach. Stakeholders included Alcoa Power Generating, Progress Energy, the Pee Dee River Coalition of industrial and municipal water users, and the South Carolina Departments of Natural Resources and Health and Environmental Control.

Solution: ADMi and USGS developed a model of the estuarial system composed of the Pee Dee and Waccamaw Rivers, and the Atlantic Intracoastal Waterway for determining how hydroelectric operations contribute to salinity intrusion events.

Results: The model, developed using artificial neural networks (ANN), was packaged in an easy-to-use "decision support system" (DSS), providing broad regulatory and stakeholder involvement in the permitting process. The DSS was quickly accepted by all stakeholders. Important findings were that hydroelectric operations contribute to salinity intrusion events during droughts; however, tidal conditions brought by offshore storms can cause uncontrollable intrusion events.

Water and Wastewater

Water and wastewater utilities are facing growing challenges to improve regulatory performance while having to meet the growing cost of chemicals and power. ADMi has been successful in optimizing treatment facilities in order to reduce disinfections by-products and TOC and optimize chemical addition which has resulted in substantial chemical savings.

Event detection systems (EDS) that monitor water quality in distribution systems for deliberate or inadvertent contamination events have become commercially available in the last few years. ADMi has done extensive research for the Water Research Foundation examining the causes of false alarms in EDS as well as researching alternative approaches that would reduce the occurrence of false alarms and false negatives. The same distribution data used for EDS can also be leveraged to detect and account for causes of conventional problems such as low chlorine residual, nitrification, and disinfection by-products as a means to improve as-delivered water quality and lower operating costs.

Faced with limited finances, increasing rates, limited water resources and increasing performance demands, accurate water demand forecasting is becoming an increasingly important element of developing sustainable utility infrastructure and operations. Improved water demand forecasting can be used by utilities to: 1) improve the timing and design of major capital improvements; 2) enable an optimal use of cash resources and long-term financing; 3) greatly assist with the design and implementation of any rate increases; and 4) assist with the balancing of water resources over the short-term

Sample Project 3

Optimizing THM Levels at a Large Southeastern Water Treatment Plant

Needs: Trihalomethane (THM) is a carcinogen that forms when organics in source water are oxidized by chlorination. THMs are measured by laboratory instruments that are expensive to buy and operate, therefore, most facilities send water samples to outside labs for analysis on a regular basis. A large municipal plant in the southeast needed to predict THM is real-time.

Solution: ADMi developed a model from historical process and laboratory data using a form of AI called an artificial neural network (ANN). This allows the system to automatically adapt to process changes by "retraining" models as new data is collected.

Results: ADMi delivered a THM control system that was integrated with a preexisting SCADA system, to predict THM concentration in real-time. THM predictions during a seven-month trial matched lab-results to ±11 ppb.

Sample Project 4

Causes of False Alarms in current EDS and an Alternate Approach 

Needs: Reducing false alarms in Event Detection Systems – Water Research Foundation Project #4182

Findings: ADMi's research found that the variability of normal WQ parameters can be both high relative to their ranges, and apparently random, making it impossible to reliably discriminate between normal behavior and events making a single-site approach to event detection. We say “apparently random” because the distribution system and upstream processes obey physical laws, however, the causes of measurement variability are unaccounted for, either because explanatory data is in hand, but not used, or explanatory data does not exist.

Results: Use a multivariate, multisite analysis by using upstream monitoring data to account for variability in downstream signals. Empirical models would be used to predict WQ parameters at the target site. A "big" prediction error would trigger an event.

Single-site EDS is blind to what is going on upstream of the site. Multi-site EDS accounts for parameter variability between upstream and target site. Large prediction error = event
Sample Project 5

Forecasting Short-Term Water Demand 

Needs: A medium sized coastal utility in the southeast wanted to better understand the factors influencing their short-term water demand.

Solution: ADMi studied ten years of historical data using data mining methods including statistics, artificial neural network models, and multi-dimensional data visualization.

Results: ADMi successfully quantified the sensitivity of demand to changing growth, rates and meteorological forming. More than 90% of the demand variability was attributable to the Baseline and weather.

Weather Impacts

Today, few disagree that climate change and sea-level rise are occurring, and they are affecting our water resources. Traditional mechanistic models have had difficulty simulating the complex interactions between the weather and surface and groundwater systems. ADMi has used artificial neural network models and other data mining techniques to complement physics-based models to evaluate threats to municipal water intakes in estuaries, and climate and groundwater use impacts on aquifer and lake levels.

Sample Project 6

Simulation of Salinity Intrusion using Climate-Change Scenarios

Needs: To develop a method for assessing risks to freshwater intakes located in estuaries.

Solution: Developed empirical estuary model using long-term weather and hydrologic data that accurately represents the intrusion process near an intake. Because of past droughts and storms, these data already encompass the expected ranges of future weather and hydrologic conditions.

Results: Model delivered in spreadsheet-based decision support system that is easily used by utility personnel. Model can be run with incremental changes to historical streamflows and sea-level-rise to determine how the frequency, magnitude, and duration of intrusion events might change. Optionally, the model can be run using output from global circulation/climate models (GCM) to estimate the impacts of their predictions of future weather patterns.

Project Participants: ADMi, US Geological Survey, University of South Carolina and the South Carolina Sea Grant Consortium. The project was sponsored by the Water Research Foundation, Beaufort-Jasper Water and Sewer Authority (BJWSA), and the National Oceanic and Atmospheric Administration.
Sample Project 7

Impacts of Climate and Groundwater Use on Water Availability in Central Florida

Needs: To develop empirical model to quantify how variations in both rainfall and groundwater use affect water levels and flows in aquifers, lakes, and springs in central Florida. Model will complement deterministic models which have been developed, but have difficulty simulating the complex interactions between the weather and the surface and subsurface environments in a karst terrain.

Solution: Developed models that predict weather and groundwater use impacts on water levels in 23 wells and 22 lakes, and flows from 6 springs.

Results: For nearly all sites, groundwater use was found to explain much less of the observed variability in water levels and flows than climatic forcing, although relative groundwater use impacts are greater during droughts. However, both future climate and groundwater use should be considered for sustainable management of the resources. Delivered decision support systems allows user manipulation of rainfall and groundwater use.

Project participants: ADMi and US Geological Survey. The sponsors were the St. Johns River Water Management District, the South Florida Water Management District, and the Southwest Florida Water Management District.
Industrial Process Optimization

Today's manufacturers must constantly respond a marketplace that anticipates new products that bring better quality and value. Processes such as testing, quality control , and environmental compliance can be improved through finding and fixing causes of poor performance. In molding and metal forming there are many variables such as die geometry, lubricants, release agents, quenches , and heat treatments that affect final part dimensions. Complex variable interactions can create tolerance stack up in machining, joining, and assembly . Data mining allows technologists to visualize how multiple variables interact, and then predict how to best minimize problems and optimize processes.

Oil & Gas

Producers know that how they acquire, produce, and deliver reserves says everything that’s important to customers and investors. Optimizing these processes means getting the most out of human and capital assets to maximize recovery and return on capital while minimizing expenses. ADMi can use its extensive experience in data mining, visualization, and optimization to help producers improve productivity while lowering operating costs.

Sample Project 8

Polymer Film Manufacturer, Southeastern US

Needs: To determine causes of unexpected film breaks in the final stage of polymer film manufacturing process.

Solution: Applied several data mining methods to determine the specific patterns of process conditions that were most highly correlated to film break events. The project started by working with the client’s engineers to understand the process. Some 160 variables, sampled at 1-minute intervals for six months (over 40 million measurements), were examined. The problem was especially challenging because combinations of variables and time delays had to be exhaustively searched to identify key patterns

Results: Recommended changes in control system logic and operating practices greatly decreased the variability in polymer flow characteristics. Subsequently, the line went from an uptime rating of 82% to 94%, to become the client’s best operating line. Annual savings are about $5 million.

Sample Project 9

Optimizing Injector Performance

Needs: To apply data mining to understand factors that affect injector performance. ADMi partnered with Advantek International.

Solution: A producer provided 11 years of North Sea injector and water quality data for 14 wells in two different blocks. The data were compiled in ADMi's iQuest™ data mining environment where they were digitally filtered to remove erratic influences from surface operations. Dynamic prediction models were then synthesized for each well using "artificial neural networks" (ANN), a machine learning technique, and "multivariate phase space reconstruction" (MPSR) from Chaos Theory. Finally, the models were used to perform sensitivity analyses to reveal the dynamics and relative impacts of each variable in ADMi's iVision™ visualization software.

Results: Findings included - a good response to acidizing in one block but not the other; time delays of one to three months for the strongest impacts to appear (depending on the variable); the impact of DO was more than twice that of SS and BI; and chlorides (CL) had a marginal effect (despite speculation that they helpfully suppress biological activity). While the value of these results would be magnified by an expanded study, their uniformity and clarity reveal a pathway forward for helping producers optimize their injection operations.

To learn more contact John Cook, 843.513.2130,

i  delivers powerful solutions to problems of real-time monitoring, prediction, forecasting, outlier detection, and decision support.
Copyright 2017   |  ADMI, LLC 3620 Pelham Rd. PMB 351, Greenville, SC 29615     (843) 513-2130
replica watches sale
Web Design by
Hosting by