Data Fusion for Environmental Process Control: Maximizing Useful Information Recovery under Data Limited Constraints

Snauffer, A. M.; Chauhan, Umang; Cogert, Kathryn I.; Winkler, Mari K.H.; Mueller, Amy

doi:10.1109/lsens.2018.2889274

Cited by 6 publications

(7 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…PSO provides a computationally efficient optimization method when the search space may be irregular (e.g., nonmonotonic, multiple local minima). Twenty individual particles were used per swarm as in Snauffer et al (2019). Momentum was set to 0.95 to simultaneously promote convergence and to mitigate the tendency of particles to run into established hyperparameter boundaries.…”

Section: Methodsmentioning

confidence: 99%

Predicting dissolved organic carbon concentration in a dynamic salt marsh creek via machine learning

Codden

Snauffer

Mueller

et al. 2020

Limnology & Ocean Methods

Self Cite

View full text Add to dashboard Cite

Dissolved organic carbon (DOC) is a master variable in aquatic systems. Resolving DOC dynamics requires high‐temporal resolution data. However, DOC concentration cannot be directly measured in situ, and discrete sample collection and analysis becomes expensive as temporal resolution increases. To surmount this problem, an option is to predict site‐specific DOC concentration with linear modeling and optical data predictors collected from high‐cost, high‐maintenance in situ spectrophotometers. This study sought to improve upon the accuracy and field costs of linear predictive DOC methods by using machine learning modeling coupled to low‐to‐zero cost predictors. To do this, we collected 16 months of in situ data (e.g., spectrophotometer attenuation, salinity, temperature), assembled freely available predictors (e.g., point in year, rainfall), and collected samples for DOC analysis, all in a salt marsh creek. At seasonal timescales, machine learning (coefficient of determination [R2] = 0.90) modestly improved upon the accuracy of linear methods (R2 = 0.80) but offered substantial instrumentation cost reductions (~ 90%) by requiring only cost‐free predictors (online data) or cost‐free predictors paired with low‐cost in situ predictors (temperature, salinity, depth). At intertidal timescales, linear methods proved ill‐equipped to predict DOC concentration compared to machine learning, and again, machine learning offered a substantial instrumentation cost reduction (~ 90%). Although our models were developed for and applicable to a single site, the use of machine learning with low‐to‐zero cost predictors provides a blueprint for others trying to model DOC dynamics and other analytes in any complex aquatic system.

show abstract

Section: Methodsmentioning

confidence: 99%

Predicting dissolved organic carbon concentration in a dynamic salt marsh creek via machine learning

Codden

Snauffer

Mueller

et al. 2020

Limnology & Ocean Methods

Self Cite

View full text Add to dashboard Cite

show abstract

“…). 19−21 On the latter point above, the known selectivity issues for ISEs, coupled with the well-documented challenges related to sensor interferences in chemically complex wastewaters, 15,22 point to the potential benefit of including commercial sensors characterizing known interferences for K + and Mg 2+ ISEs (e.g., Ca 2+ , Na + ), while the variability in bioprocesses generally suggests potential utility in data characterizing overall process chemistry, for example, pH, DO, NH 4 + , or Cl − . Machine learning (ML) approaches have proven useful as a data fusion technique when using sensor arrays in complex environments, improving the ability to capturing (usually nonlinear) underline patterns in spite of (often covarying) interferences.…”

Section: −mentioning

confidence: 99%

“…12,13 For instance, a recent study demonstrated real-time controls of dosing in a chemically driven phosphorous removal process using a soft sensor (comprised of pH, total suspended solids, electrical conductivity, oxidation reduction potential, level, and flow sensors) at a full-scale WWT plant. 14 Building on prior work using soft sensors for characterization of minority and/or difficult-to-measure constituents in natural waters and wastewaters, 15,16 this work assesses the feasibility of designing a soft sensor for phosphorus removal monitoring in relatively more sustainable biological processes such as EBPR with the goal of providing data at higher accuracy, higher resolution, lower lag time, and lower cost than existing solutions. Successfully applying online monitoring and control strategies like those above requires an understanding of both (1) the chemistry of the target process and (2) the expected interferences on candidate sensors, which can then be used to appropriately design the hardware backbone of a soft sensor solution.…”

Section: ■ Introductionmentioning

confidence: 99%

“…Yet, the use of arrays of sensors for direct and indirect (proxy) measurement of chemical constituents has been proven successful in related applications. − Soft sensors, that is, systems that utilize a tuned algorithm to estimate the concentration of a target analyte from readings of multiple individual sensing devices, − may provide a promising alternative to expensive and sometimes unreliable analyzers and probes for wastewater process monitoring and control. , For instance, a recent study demonstrated real-time controls of dosing in a chemically driven phosphorous removal process using a soft sensor (comprised of pH, total suspended solids, electrical conductivity, oxidation reduction potential, level, and flow sensors) at a full-scale WWT plant . Building on prior work using soft sensors for characterization of minority and/or difficult-to-measure constituents in natural waters and wastewaters, , this work assesses the feasibility of designing a soft sensor for phosphorus removal monitoring in relatively more sustainable biological processes such as EBPR with the goal of providing data at higher accuracy, higher resolution, lower lag time, and lower cost than existing solutions.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Design of a Soft Sensor for Monitoring Phosphorous Uptake in an EBPR Process

et al. 2022

Self Cite

View full text Add to dashboard Cite

Phosphorus (P) is a key nutrient targeted for removal by wastewater treatment, increasingly being achieved using biological processes such as enhanced biological phosphate removal (EBPR). However, commercial instrumentation for automated measurement of P is costly and provides only limited temporal resolution, constraining implementation of real-time controls in EBPR processes. This study designs a soft sensor for real-time controls using a suite of relatively low-cost sensors (ion-selective electrodes) to monitor P removal by measuring (1) secondary cations tightly coupled to bioP metabolism and (2) process bulk chemistry. Data collected from a highly instrumented lab scale reactor are used to evaluate which and how many sensors are required to achieve this goal. Machine learning (ML) approaches (support vector machines, nonlinear logistic regression, random forest, and Bayesian classification) are evaluated for sensor data fusion and coupled with a decision metric to operationalize the algorithm for reactor controls. Two key results emerge: (1) use of the slope of sensor data (mV/min) rather than raw data (mV) as the predictor signals significantly improves accuracy and resilience of the soft sensor-based system and (2) the K+ ion-selective electrode, in combination with any of the four ML algorithms, is sufficient to detect completion of P removal (within study sampling granularity of 4 min) with 100% accuracy in real time. High accuracy is maintained even as process chemistry is varied to increase the interference experienced by the sensors, indicating that this soft sensor is viable for use in real wastewater applications.

show abstract

“…PSO provides a computationally efficient optimization method when the search space may be irregular (e.g., non-monotonic, multiple local minima). Twenty individual particles were used per swarm as in Snauffer et al (2019).…”

Section: Model Setup and Evaluationmentioning

confidence: 99%

Dissolved organic carbon dynamics in a salt marsh creek

Codden¹

View full text Add to dashboard Cite

Data Fusion for Environmental Process Control: Maximizing Useful Information Recovery under Data Limited Constraints

Cited by 6 publications

References 13 publications

Predicting dissolved organic carbon concentration in a dynamic salt marsh creek via machine learning

Predicting dissolved organic carbon concentration in a dynamic salt marsh creek via machine learning

Design of a Soft Sensor for Monitoring Phosphorous Uptake in an EBPR Process

Dissolved organic carbon dynamics in a salt marsh creek

Contact Info

Product

Resources

About