Sewer flow meters can be very laborious and expensive to maintain. Regular meter services and spot check measurements can quickly eat up your metering budget. This article outlines how the use of data analytics can optimize your metering accuracy and servicing costs. Here is a quick overview of the key elements of this concept:
- Manual spot measurements of depth and velocity are too inaccurate to rely upon as the sole means of verify meter accuracy and detecting meter drift.
- Meter servicing requires expensive confined space entry. Using data analytics to optimize when these service visits are performed can save a lot on your metering program.
- The best approach to understand meter accuracy includes an array of tools such as spot measurements, continuous data review, scatter plots, hydraulic null modeling, mass flow balance and dye testing. Each of these tools are described in this article.
- The use of data analytics, especially continuous data review and scatter plots, can separate real flow trends from metering issues, and help identify when a meter needs to be serviced and what type of servicing is needed.
- The use of data analytics like those demonstrated in this article can optimize your flow metering program by improving the accuracy of the flow meter data, providing confidence that the flows recorded are correct, and significantly reducing flow metering costs by performing expensive field visits only when needed.
A sewer system is a challenging environment to make measurements. Area-velocity flow meter sensors are often placed on the bottom of the pipe beneath the sewage, making them prone to fouling, ragging and damage from moving sediment and debris. Even meters with sensors above the sewage like a radar or laser-based meter are placed in a tough environment with gases, water vapor and splashing that can interfere with the meters.
These challenges result in the need for frequent meter servicing and maintenance. This servicing often entails visiting the metering site to perform maintenance that includes downloading the data, checking the meter data, performing spot checks of the depth and velocity, and cleaning the probe. Because these activities often require entering the manhole, a multiple-person crew is needed for confined space entry procedures. This is very laborious and can be very expensive. Meter servicing costs are often the single largest expense in performing flow metering. Optimizing the frequency and scope of these visits with data analytics can greatly reduced flow meter servicing costs.
Limitations of Manual Depth and Velocity Readings
A common QA/QC measure for sewer flow meters is to make manual spot measurements of the depth and velocity using an independent measurement technique. Depth is often measured with a steel ruler or rod, and velocity is often measured with a portable velocity sensor like the one shown below. The manual measurements are then compared to the meter measurements to verify that the meter is working properly. This methodology is useful for checking if the flow meter is “in the ball park” with the flow measurement. Caution should be applied before adjusting metering data to match these measurements, for the reasons outlined below.
Spot measurement techniques can vary in their accuracy. Depth can usually be accurately measured to within a 1/8 or 1/4 inch, but turbulence or waves in the flow can lower the accuracy. Velocity is harder to accurately measure with a portable probe due to the following challenges:
- Turbulence in the flow
- Timing-varying velocity (due to turbulence)
- Varying velocity across the cross section of the flow
- Interference of the probe, cord or rod with the flow
- Low depths or high velocities
These challenges necessitate making several measurements over time and at different locations in the flow cross-section and averaging them to generate an aggregate cross-sectional and time-averaged velocity. If this sounds challenging, it should! This is especially challenging when deep underground, in a dark confined space with sewage rapidly flowing by! Considering these variables, a field technician would be doing well if the spot velocity measurement was within 20-25% of the actual velocity. For certain conditions, like low flow, high velocity or a turbulent location, it is possible for the accuracy to be much lower.
Given the inaccuracies of the spot checks of depth and velocity, the resulting flow may be only accurate to within 25-50%. This is certainly not accurate enough to use for calibration or adjustment of the flow meter data. Such measurements should only be used to spot check whether the meter is operating correctly or whether there are gross measurement errors in the metering data. Other techniques must be used in conjunction with spot measurements to understand the accuracy of the meter performance. It can be almost impossible to detect a meter drift on the order of 10% with spot measurements that are only accurate to 25-50%. That is where deeper analytical tools are needed.
Toolbox for Understanding Meter Accuracy
There are several analytical techniques that can be used in addition to spot measurements to understand the accuracy of meter data. These are briefly described below with links to more detailed articles and resources on each item.
- Continuous meter data review – Just examining the continuous depth, velocity and flow data recorded on a regular basis is a simple and effective way to detect metering issues, data drops, drifts, and ragging on the sensor. This is covered in more detail in the next section.
- Scatter plots – A scatter plot is a plot of the depth versus the velocity (or flow) recorded by a meter and are very useful tools to detect whether a meter is operating in normal ranges or may be having some issues. The scatter plot article covers the basic concepts and more detail is provided in the next section for using scatter plots to optimize meter servicing.
- Hydraulic null modeling – When a sewer is flowing under uniform flow conditions, the expectation is that the scatter plot will trend along the Manning’s curve. But for more complicated hydraulics, a fully dynamic hydraulic model may be needed to evaluate the reasonableness of the meter data. The technique involves developing a detailed hydraulic model of the collection system, and injecting the observed meter flow into the model. Comparisons can then be made between the observed depth and the model depth to understand the meter performance. (Null modeling refers to the absence of a hydrologic model, thus focusing the model results entirely on the accuracy of the hydraulic model and meter data.)
- Mass flow balance – Mass Flow Balance (MFB) is a process that compares the flows from upstream meters to downstream meters to check the continuity of the flows in the metering network. A good description of the process is contains in a previous blog post on Best Practices for Flow Meter Data QA/QC. Here is a video that demonstrates the process.
- Dye testing – Dye testing is an accurate methodology of measuring instantaneous flows in a sewer using a fluorescent dye. A good description of the process is contains in a previous blog post on dye testing.
Analytics to Optimize Meter Servicing
The remainder of this article focuses on how to use data analytics to optimize meter servicing. Combining the use of regular data downloads and data analytics can optimize meter servicing costs. These techniques can alert the field crews to when a meter is having an issue that necessitates field servicing, reducing expensive meter servicing field visits to only when there is a known meter issue.
Telemetry on the meters can facilitate this process, but even without telemetry, these techniques can optimize meter servicing. This can be done by manually downloading the data periodically from the surface (no entry crew required) and then reviewing the data and only servicing the meters that need it. In either case, the data analytics described below can optimize meter servicing and greatly reduce metering costs.
The base data used for these examples is from a real flow meter. To illustrate the impacts of changing flows or sensor drift more clearly, the data has been modified to exaggerate these impacts of these effects for the examples below. These examples show how these concepts work extremely well for reviewing real metering data to detect issues and mobilize field servicing crews.
Base Meter Performance
The base meter performance shown in this example is a period when the meter is working well. The continuous review of the flow, depth and velocity show a consistently repeating pattern with no visible sensor drift. The depth and velocity both trend up and down together very tightly, resulting in a tight scatter plot that follows a Manning’s relationship very closely. For a meter with this level of performance, there is no need for an expensive field service. Simply reviewing the data regularly with tools like this can save a tremendous amount on field servicing.
Real Decreasing Flows
This example shows how a meter would perform when the flows in the system are decreasing. This is very common as long-term system flows in a sewer system can trend up or down over long periods of time from ground water levels. This example could be from a system that has not had much rain, so that ground water levels are dropping, resulting in a long-term reduction in ground water infiltration. Let’s examine how to differentiate between flow trend shifts and sensor drifts using data analytics.
Notice that the continuous flow plot below shows a downward trend in the second half of the month. At first glance, it might appear that there is sensor drift. However, notice upon examining the depth and velocity continuous plots that both the depth and the velocity are trending down together as well. It would be unusual to have a sensor drift in two independent measurements at the same time that are exactly correlated. This is a tell-tale sign that the decrease in flow is real and that the lower flow caused both the depth and velocity to trend lower, following the Manning’s equation.
This can be seen clearly on the scatter plot below. The plot has been split into two, with the blue points from the first half of the month, and the red points from the second half of the month when the flows were trending lower. Notice that the red points continue to follow a Manning’s trend line, albeit at lower flows. Again, this is a tell-tale sign that the meter sensors are working fine, and that the downward drift happened at both sensors together because the real flow was trending down.
To have confidence that the meter is performing well, it is necessary to carefully examine all three continuous plots and the scatter plot. When this detailed examination is done, it can validate that the meter is doing fine and that the trending patterns are real. There is no need for an expensive field visit to a site like this.
Meter Drift Example
This example shows how a meter would perform when the depth sensor is drifting, resulting in poor meter accuracy. This can occur if the sensor fails, or if there is ragging or debris interfering with the sensor. In this case, it is necessary to perform a field service right away. Let’s examine how to detect that using data analytics.
Notice that the continuous flow plot below shows a upward trend in the second half of the month. Furthermore, notice that only the depth measurements are trending up and the velocity continues to exhibit a consistent repeating pattern. This is a tell-tale sign that the depth sensor is drifting and is in need of servicing. This is also very clear on the scatter plot below, which exhibits several parallel trend plots as the depth sensor drifts higher and higher. Baring some unusual, intermittent system operation, there is no normal hydraulic condition that would cause the scatter plot to deviate from the Manning’s line like that. This too is a tell-tale sign that the depth sensor is having an issue and drifting.
From these observations, we can deduce several things:
- The upward flow trend is not real.
- The depth sensor is drifting and in need of servicing.
- The cause of the depth drift is likely not from ragging on the sensor, because ragging would have interfered with the velocity measurement too.
- The likely cause is a malfunction or fouling of the depth sensor itself.
The depth sensor at this meter needs to be serviced and cleaned, and then tested to verify that the cleaning fixed the issue. If cleaning does not fix the issue, the sensor may have failed and may require replacement. Diagnostics like this can help optimize when to send a field crew out to do a confined space entry to service the meter, and even identify the type of sensor servicing that will be needed.
- Improving the accuracy of the flow meter data.
- Providing confidence that the flows recorded are correct.
- Significantly reducing flow metering costs by performing expensive field visits only when needed.
H2Ometrics was developed specifically to aid owners and engineers in rapidly performing these types of data analytics. Here is a link to short overview video about the platform. Here is a link to several case studies that demonstrate how our customers are using these tools. Below are links to several tutorial videos that demonstrate many of these tools with the H2Ometrics platform.