KCF Technologies Blog

Maintenance Practices Have Major Effects on HVAC Energy Consumption

SmartDiagnostics® vibration sensors on a chiller in an university HVAC facility.
A report copyrighted 2012, posted online under "Exhisting Building Retrofits" by the Institute for Building Efficiency ("an initiative of Johnson Controls"), bears the unremarkable title Studies Show: HVAC System Maintenance Saves Energy.  This unattributed brief is interesting in that it charts nine different "Causes of Degraded Energy Performance in HVAC Equipment," and tersely summarizes both the "HVAC Maintenance Solution" and the "Estimated Impact on Chiller Energy Consumption."

The three different studies summarized in the report concluded that the best HVAC maintenance practices could produce energy results of five to 20 percent.  "In contrast," one report concluded, "poor maintenance practices can increase energy use by 30 to 60 percent"--a clear sign not only that are all maintenance practices are not equal, but that some may actively defeat their own purposes of reducing energy use and expense.

The report also painted a picture of the state of current HVAC maintenance practices in the United State today:

"There are three basic approaches to maintaining HVAC systems in buildings:
  1. Reactive maintenance.  Under this management practice, used by 55 percent of companies, HVAC systems run until a problem or failure occurs.  (This strategy is also called run-to-fail maintenance.)
  2. Preventive (or scheduled) maintenance.  This practice used by 31 percent of companies, includes the periodic maintenance of HVAC equipment, generally as prescribed by the manufacturers.
  3. Predictive maintenance.  Practiced by 12 percent of companies, this strategy differs from preventive maintenance by basing maintenance on the actual condition of the machine, rather than on a preset schedule.  Predictive maintenance can be the most cost-effective over the long term, but does require technology infrastructure investments up front."
Though it cautions conventionally that "more studies are needed" to obtain greater precision, it concludes, "The HVAC industry can develop better tools to help building owners and facility managers evaluate the relationship between maintenance costs and energy costs and support investment in the appropriate maintenance approach."

Photo by Christopher Shannon/KCF Technologies.  All rights reserved.

SmartDiagnostics® Feature Highlight: Powering Up

The SmartDiagnostics® vibration sensor node (VSN) is optimized for low-power and high-fidelity vibration and temperature sensing.  The VSN uses KCF Technologies' proprietary wireless protocol to transmit the full dynamic vibration spectrum over the air on a nearly continuous basis.  The VSN's very small size and robust industrial design enable its reliable use on a wide variety of machines and in harsh industrial settings.

VSNs can be powered by a 3.6 volt, AA, lithium battery or by an energy harvester.  The decision to use a battery or an energy harvester is one of personal preference.  The advantages of using battery power include: simplicity, low cost, and relatively long life.  This inexpensive battery has an impressive eight year maximum life span with a 1-hour collection interval and a 4-second ping rate.  Theoretically, a plant manager shouldn't have to think about a battery replacement for more than six years.

On the downside, while the lithium battery will last many years, it must still be monitored and eventually replaced.  Some environments do not allow the use of lithium batteries due to potential hazards.  The advantage of an energy harvester is that, once installed, it will power a sensor virtually forever without maintenance.  On the downside, energy harvesters are more expensive and they relay on sufficient environmental energy to generate power.  Learn more about SmartDiagnostics® energy harvesters here.

Can You Afford NOT to Use Predictive Maintenance?

The preceding is a paraphrase of the title of Steve Carson's thoughtful 2009 article on the Water and Wastewater Blog and WaterAndWastewater.com.  (Carson wrote for Multirode, which "supplies pump station controllers and supervisors; level-sensing devices, web-based monitoring services; SCADA software; panels; and engineering and integration services" to the water and wastewater industries and municipal systems nationwide.)

America in the 21st century is well aware that it has a considerable investment in utility infrastructure, and knowledge of the condition of these critical assets has become crucial to planning maintenance and improvements.  "The problem for wastewater utilities," Carson observes, "is that often the pipe network is 100 years old and so corroded that the whole pipe network needs replacement--and it's the largest asset class in terms of replacement value."

Carson details three approaches to caring for pipes, pumps, and motors as follows:
  • " 'Run to fail,' i.e., wait until the pump or motor fails and--usually--race out and fix it or replace it;"
  • "Preventive maintenance, i.e., periodic maintenance of a pump or motor to avoid waiting for it to fail;"
  • "Predictive maintenance - the utility determines the state of each asset and can plan for servicing, or replacement, of a pump or motor."
Conceding that most wastewater treatment operations use the first approach ("usually due to lack of resources"), he also notes "a significant proportion" have adopted the second, or preventative maintenance, approach, though most of these utilities wonder "whether they are doing too much, or too little."

Carson prefers the third way, "a much more proactive approach, generally known...as Condition Based Monitoring, which in practice is the same as Predictive Maintenance."  To work, however, this latter approach requires data that many plants lack, and are unsure of how to collect.  He suggests five key types: insulation resistance of motor windings, pump flow rates, pump volume per cost of energy, detailed pump fault data, and vibration analysis (noting that "to get vibration data, you need...sensors.").

Carson concedes "that adopting predictive maintenance strategies doesn't mean that every failure will be known in advance."  Emergencies will always be with us.  However, "the benefit you have with...predictive maintenance is that you can now have a lot more confidence of the state of you assets and you run maintenance program more proactively and most cost-effectively."  Best of all, because you have accurate, data-based knowledge of your system, "you can plan the most effective capital works, replacement, or service program."

How to Easily Save Trillions of Dollars

SmartDiagnostics® vibration sensor on a cooling tower fan.
Can it really be easy to save money with continuous condition monitoring systems?  What's the secret?

Here's a more direct question:
"What would I do if I had access to continuous vibration data on all of my rotating equipment?"  If you are a normal person like my mother, my wife, and most of my friends, that's probably a question that you never have asked yourself, and never will.  (Note: My dad was doing vibration monitoring projects on Sikorsky helicopters back in the early 1970s, but that is a special case.)

Fortunately for KCF Technologies, the question of how to make use of loads of vibration data is one that is often asked by our customers and industrial partners.  Why?  Because there is great promise in installing small, low-cost, wireless sensors to monitor equipment and save lots of money in industrial processes.The promise for savings is huge, estimated by the Department of Energy (DoE) at $2.5 Trillion per year in U.S. industry (which incidentally would solve most of the country's financial problems and guarantee American competitive advantage).

So what's the problem?  The problem most commonly described is that it's hard to make useful, actionable information out of reams of vibration data.  That has been the case until recently, because vibration monitoring has required sensors, sophisticated acquisition systems, and a well-trained vibration technician or engineer.  Each pump, compressor, turbine, air handling unit, chiller, fan, etc. has its own mechanical behavior and vibration characteristics and requires detailed expertise to fully understand the complexities that would impact up-time and reliability.  This has generally relegated vibration monitoring to only the critical assets of a plant where the expertise and monitoring equipment can be justified.  However, many pieces of important equipment (most, actually, in the balance of plant) are skipped.

So what's the secret to make sense of the data on all the other rotating equipment?

The secret is to start simple, break down the problem with some assumptions, and incorporate a wealth of readily available knowledge into easy-to-use, or automatic, software.  This is the core assumption used in SmartDiagnostics®:
  • Rotating equipment vibrates at an overall level, and at certain frequencies, and the vibration increases before it fails or needs to be shut down for maintenance/repair.
From that, the problem can easily be solved:
  1. The frequencies can be easily calculated for some of the most basic types of equipment and failure modes, and built into downloadable templates.
  2. The vibration levels can be set based on standard ISO Vibration Severity recommendations
  3. Both the frequency bands and vibration levels can be adjusted to a specific machine based on a baseline (if the software is easy to use).
Starting in 2004 with a DoE project to tackle low-cost, ubiquitous, wireless sensors, KCF has been working to to solve this problem, and we've got at least a big part of it solved.  It's inexpensive and easy enough to achieve a positive return on investment in under a year.  Check back for examples and case studies of how our customers are doing this today.

For more information on how our customers are doing this, check out the templates for monitoring templates on KCF's SmartDiagnostics® costumer resource page.

Post by Dr. Jeremy Frank, President of KCF Technologies, Inc.

Photo by Christopher Shannon/KCF Technologies.  All rights reserved. 

SmartDiagnostics® Feature Higlight: Measuring Vibration Severity

SmartDiagnostics® enables you to view the amplitude of vibration in your machinery.  Amplitude is an indicator of the severity of a vibration.  Amplitude can be expressed as one of the following engineering units:
  • Velocity: the speed of movement
  • Acceleration: the force associated with movement
In most situations,it is the speed, or velocity, amplitude that gives you the most useful information about the condition of the machine.  Velocity measurement is effective over a wide range of frequencies from low to high.  Velocity is typically expressed in units of inches per second (ips) or millimeters per second (mm/sec).

Velocity is the most useful measure of vibration because it gives us the same diagnosis across all frequencies.  0.5 ips at 1,000 rpm indicates the same vibration severity as 0.5 ips at 10,000 rpm.

Acceleration is the rate of change of velocity and is the measurement of the force being produced.  Acceleration is typically expressed in units of mm/sec2 (millimeters per second squared) or gravitational forces (Gs: 1G = 32.17 ft/sec/sec).  Acceleration is frequency reated, in that 1 g at 1,000 rpm is not the same vibration severity as 1 g at 10,000 rpm.  Acceleration generally emphasizes high frequency peaks in a spectrum.

Amplitude can be expressed in terms of its peak value, or what is known as its Root-Mean-Square (RMS) value.  The peak velocity amplitude of a vibrating machine is simply the maximum (peak) vibration obtained by the machine in a given time period.  By contrast to the peak velocity amplitude, the RMS velocity amplitude of a vibrating machine tells us the vibration energy of the machine--the higher the vibration energy, the higher the root-mean-square velocity amplitude.

Predictive Maintenance Tip: The Power of Sharing PdM Information and Techniques

The Calpine story Kevin Nordenstrom tells above is convincing proof that predictive maintenance work, but the road was not without its bumps, especially at the start.  As he notes, "One of the challenges we faced with the program was getting the individual Calpine plants on board with the PdM program.  In particular, it was challenging to justify and demonstrate the value of PdM because of the difficulty in measuring and documenting the cost savings of potential events that were ultimately prevented and never occurred."

Just three years before the $1.5 million savings noted in Nordenstrom's story, "only a handful of Calpine's plants routinely practiced PdM techniques, with each conducting its own PdM activities in a non-standardized manner.  As a result, no uniform technology or strategy was used, sharing of best practices and successes was limited, and each plant handled equipment failures in its own unique manner."

"After reviewing the options and potential strategies for meeting up-time and, reliability, and profitability goals, senior management decided to develop and begin implementing a comprehensive PdM program," he recalls.  "Simply stated, we can't allow our customers to be without power. To make sure that doesn't happen, we needed a uniform PdM program that could readily implemented at the majority of our plants."

And, that's where the game-changing power of predictive maintenance really shines, sharing information to make the best maintenance decisions across scores of plants, all sharing the same objectives.

"Rolled out to the individual plants..., the new PdM program consisted of a variety of proactive tools and techniques, including condition-monitoring hardware and software from Rockwell Automation...By implementing common procedures and practices among the various plants, Calpine is [able] to share experiences and documented successes, which will help drive further efficiencies and cost savings across plants.  Cost savings also resulted from implementing an enterprise-network system."

Going from isolated maintenance specialists making their individual decisions alone, virtually "in the dark," to a dedicated PdM team sharing their experiences, notes, and observations across dozens of power plants magnifies the power and value of predictive maintenance enormously.  And, if Calpine can do it, so can other large networks of facilities sharing similar tasks and technologies--almost certainly including yours.

When Power Plant Downtime can Mean Massive Losses, Predictive Maintenance Makes a Big Difference

Published by Putnam Media, Plant Services reaches over 35,000 manufacturing plants across all U.S. processing and OEM industries, including utilities and power generation facilities.  Its mission is to chronicle how American industry leaders are "continuously retrofitting, updating, and redesigning their existing facilities with up-to-date technology," in pursuit of reduced downtime, increased up-time, and therefore greater profits.

Nowhere does performance matter more than in the electrical industry, where downtime at a large natural gas power plant can cost $11,000 an hour.  Yes, that's $264,000 a day.

At the height of its success in 2004, San Jose-based Calpine Corp. had 92 energy centers in 21 states , Canada, and the United Kingdom, with a combined capacity of 22,000 MW.  It was two years previously that Calpine fully embraced predictive maintenance (PdM) as the way to maximize profitability.  This is well documented by Kevin Nordenstrom, Calpine's predictive maintenance engineering manager, in his 2005 Plant Services feature, "Forced outages and MRO costs reduced with PdM."

"with this much potential profit at stake," Nordenstrom wrote, "maintaining energy production, reducing forced outages (caused by unplanned shutdowns, or equipment failure), and avoiding generator deratings (when a unit fails to deliver power at its rated capacity) are top priorities for management and employees alike."

After considerable in-plant research and considerations of all its options, Calpine installed its condition-based maintenance Enshare system in 66 of its U.S. plants, achieving "nearly $1 million in cost savings" annually.  Seasoned maintenance personnel at each plant conducted vibration analysis "on equipment deemed critical to maintaining production, including gas turbines, steam turbines, cooling fans, and critical oil and water pumps."

Determining which equipment should be monitored was relatively simple: if it could cause expensive downtime, either through failure, or due to the high cost of replacement, it was deemed worth monitoring.

"My team of eight PdM engineers and technicians perform different PdM activities at each of the plants," Nordenstrom recorded, "including vibration analysis, online motor analysis, off-line motor analysis, as well as infrared thermography, and lubrication and transfer oil analysis....This computes to approximately 600 points of data collected at each of the...plants that have enrolled in the PdM program."

Vibration readings were collected monthly with hand-held data collectors.  "With this information, the team can analyze potential problems and recommend corrective actions before performance is affected."

The overall result?  For the last two quarters for which PdM savings were recorded, the "documented cost avoidance...exceeded $1.5 million."

"Overall, the maintenance program...enabled Calpine to maximize system reliability by identifying, and correcting, potential problems before a forced outage occurs, or production is interrupted;" Nordenstrom concluded, enabling its natural gas, and geothermal steam power plants to "...carefully plan, and time, equipment repair, and replacement, to avoid any sudden process interruptions."

Some Pump Related Calulations You May Want to Know

Motor bearing frequencies for pumps are calculated based on the running speed and the bearing geometry.  For rolling element bearings, the following formulas are used:
  • Ball Pass Frequency Outer Race (BPFO) = Nb/2 x S x (1 - (Bd/Pd x cos(th))
  • Ball Pass Frequency Inner Race (BPFI) = Nb/2 x S x (1 - (Bd/Pd x cos(th))
  • Fundamental Train Frequency (FTF) = S/2 x (1 - (Bd/Pd x cos(th))
  • Ball Spin Frequency (BSF) = Pd/2Bd x S x (1 - (Bd/Pd x cos(th))
  • Nb = number of rolling elements
  • S = speed (revolutions per second, in Hz)
  • Bd = ball diameter
  • Pd = pitch diameter
  • th = contact angle (degrees)
  The following guidelines can be used as a quick reference:
  • Ball Pass Frequency Outer Race (BPFO) = Nb x S x 0.4
  • Ball Pass Frequency Inner Race (BPFI) = Nb x S x 0.6
  • Fundamental Train Frequency (FTF) = S x 0.4
  • Ball Spin Frequency (BSF) = S x 1.6
For example, a centrifugal pump with eight blades operating at 1,800 RPM with 19 rolling elements in each bearing will have the following frequencies:
  • Fundamental Train Frequency: 12Hz
  • 1x Motor Speed: 30Hz
  • Ball Spin Frequency: 48Hz
  • 2x Motor Speed: 60Hz
  • Ball Pass Frequency Outer Race (BPFO): 228Hz
  • Blade Pass Frequency (FPF): 240Hz
  • Ball Pass Frequency Inner Race (BPFI): 342Hz
The vibration spectrum will display a peak at each frequency noted above.  The actual frequency will be slightly lower as the speed slows under load with motor slip of a few percent.  The analysis should be performed over a frequency band accommodating this range of speed.

For variable speed pumps, the motor speed will vary over a range of speeds based on the required pressure, flow rate, or peak efficiency.

Each peak is analyzed for a trend in amplitude against pre-set warning and alarm levels.  Specific information may be available from individual manufacturers, or through operational specifications.  The following guidelines provide a useful start for analyzing the FFT max amplitudes in each band for a pump:

General (rotation and bearing locations including motor speed, FTF, BSF, BPFO, and BPFI)
Level Amplitude (max)
Warning 0.10 in/s
Alarm 0.25 in/s
Shutdown 0.62 in/s

KCF Making the Things You Work with Smarter