Chemicals are an operating cost for a water treatment plant. Operators can reduce costs by dosing just enough chemical to achieve desired water quality but not overdosing.

In this application we consider a water treatment plant that takes raw water from a surface water source and treats it before delivering to a water supply network. The raw water contains suspended solids which are clumped into larger aggregates through the addition of a coagulant then separated from the water using a filtration process.

The water treatment plant uses a jar test to determine the coagulant dose. In this manually performed lab test a sample of the raw water inflow is obtained and divided among a series of jars. A difference dose of coagulant is added to each jar then all jars are agitated at a consistent rate. Agitation is stopped and the jars are allowed to settle for a period of time. The turbidity of the raw water sample in each jar is then measured and the dose rate resulting in lowest turbidity is defined as optimum.

The disadvantage of the above test is that it is formed at discrete times, and as such can not react to a continuously changing raw water quality. 


A predicted value model is trained with inputs including surface water source depth, raw water flow, raw water turbidity, streaming current detector, turbidity of filtered water, disinfection chlorine dose rate, and residual chlorine . The model is trained to predict the optimal coagulant dose rate determined by jar testing.

Once the model is trained, the VROC system provides dashboards to validate the effectiveness of the model as well as model statistics.

VROC ingested the historical data and provided insightful results as to the major factors responsible for failures. In addition, VROC was able to identify transient factors such as one month where the chemical supplier to the plant supplied incorrect chemicals. VROC correctly identified that the chemical was the significant factor causing failures for that particular month. 

The VROC system allows this customer to save significant time, rather than spending thousands of engineering hours on each problem they can quickly identify the root cause of individual failures or events and concentrate on fixing the problem rather than their previous method of comparing trends or using calculations in Excel.
Accurate recommendation of optimum setpoints provided continuously in real time.


  • Continuous monitoring of water source, weather, and plant operating characteristics to provide real time optimum setpoints.
  • Reduced reliance on manual jar testing.
  • Reduced chemical use and more efficient and effective chemical dosing based on plant and water source conditions.