Keywords
sequencing batch reactors, business process, event calculus
Start Date
1-7-2012 12:00 AM
Abstract
Sequencing Batch Reactors (SBRs) provide several advantages in terms offlexibility and robustness to variations in the inflow concentrations and sludge alterations.Adjusting the duration of the load, reaction and discharge phases, it allows to optimizethe treatment, saving time and energy. Optimal policies can be defined by observingand analyzing some chemical and physical parameters such as pH, redox potential anddissolved oxygen concentration. Various Artificial Intelligence (AI) techniques have beenproposed and used to recognize the state of the biological processes inside the plant,using the signal trends and changes as indirect indicators. In particular, the terminationof a process (typically, denitrification in the anoxic phase or nitrification in the aerobicone) can be estimated by a management system and used in control policies. In thispaper, we point out that this recognition task is only part of the responsibilities of a potentialEnvironmental Decision Support System (EDSS) managing an SBR, and by nomeans the only one which can take advantage of AI techniques. In fact, the entire controland management system could be defined and implemented using declarative AItechniques, deployed within a uniform execution environment. In particular, we proposetwo alternative but similar models of the SBR operation cycles: one is based on workflows,exploiting the BPMN2 (Business Process Management Notation v.2) standard forthe definition and execution of business processes; the other instead is founded on theprinciples of (Reactive) Event Calculus. Both representation capture the operational behaviorof the SBR, supporting the state transitions and the actions associated to eachstate. The transitions themselves are driven by events, i.e. relevant state changes. Theevents are identified either directly, through the sensor system installed on the plant, oranalysing a combination of other more elementary events, and conditioned by the actualstate of the plant. The correlations between events, states and control actions, as wellas their consequences, will eventually be defined using rules, which will also encode thenecessary knowledge to deal with exceptional conditions, accounting for a more flexiblesystem.
Modelling SBR cycle management and optimization using Events and Workflows
Sequencing Batch Reactors (SBRs) provide several advantages in terms offlexibility and robustness to variations in the inflow concentrations and sludge alterations.Adjusting the duration of the load, reaction and discharge phases, it allows to optimizethe treatment, saving time and energy. Optimal policies can be defined by observingand analyzing some chemical and physical parameters such as pH, redox potential anddissolved oxygen concentration. Various Artificial Intelligence (AI) techniques have beenproposed and used to recognize the state of the biological processes inside the plant,using the signal trends and changes as indirect indicators. In particular, the terminationof a process (typically, denitrification in the anoxic phase or nitrification in the aerobicone) can be estimated by a management system and used in control policies. In thispaper, we point out that this recognition task is only part of the responsibilities of a potentialEnvironmental Decision Support System (EDSS) managing an SBR, and by nomeans the only one which can take advantage of AI techniques. In fact, the entire controland management system could be defined and implemented using declarative AItechniques, deployed within a uniform execution environment. In particular, we proposetwo alternative but similar models of the SBR operation cycles: one is based on workflows,exploiting the BPMN2 (Business Process Management Notation v.2) standard forthe definition and execution of business processes; the other instead is founded on theprinciples of (Reactive) Event Calculus. Both representation capture the operational behaviorof the SBR, supporting the state transitions and the actions associated to eachstate. The transitions themselves are driven by events, i.e. relevant state changes. Theevents are identified either directly, through the sensor system installed on the plant, oranalysing a combination of other more elementary events, and conditioned by the actualstate of the plant. The correlations between events, states and control actions, as wellas their consequences, will eventually be defined using rules, which will also encode thenecessary knowledge to deal with exceptional conditions, accounting for a more flexiblesystem.
