Appropriate monitoring levels

Determining how much predictive maintenance you need to do

Over the past few years, the question of how much maintenance should be done has come up over and over again. I have been asked this question by college students doing their term papers, asset care managers putting together new or revised maintenance plans, as well as management consultants seeking information for their clients.

My answer is always the same--It depends. Now, you deserve and explanation.

Let's start by going to the top.You can specify the level of predictive maintenance in your company first by understanding what your company produces. Then you must factor in your company's strategic basis for competition, that is, price, quality, and delivery. Then, one must consider regulatory concerns such as EPA, OSHA, and perhaps insurance or other similar requirements.

Finally, it behooves one to consider the particular corporate value structure in a company. One should realize that no matter how well thought out and sophisticated a predictive maintenance program seems, if it is inconsistent with corporate goals or philosophy, it probably won't work.

Understanding your facility's product
Here, several cases should be considered. If usable space is the end product, then building support equipment--heating, cooling, ventilating, and electrical equipment--are likely to be important. The question that must be answered before one can ascribe a level of predictive maintenance to the equipment is: What happens if the equipment experiences an outage? Does your company or asset care department lose a customer? Can that customer continue to function, perhaps less efficiently or conveniently, without fully operational support equipment. At this point, you and your customer's answer to that question is the only one that counts, not mine.

If the space is unusable or your company will experience severe economic penalties because of an equipment outage, then you must ensure that the outage does not occur. Predictive maintenance becomes the viable method for avoiding equipment outages.

The second case is that of a production system--the facility produces something. Here is where the question of corporate competition arises. Is your company competing on the basis of price, quality, delivery, which one or all three? Each of these competitive factors dictates different levels of maintenance, especially predictive maintenance.

Price--If you intend to be the low-cost producer in your business, then the price structure is going to be based on the lowest unit cost per item produced. That translates into greatest number of units produced per unit of time. The tacit assumption: the production assets must be running for all scheduled production time units. If production assets are not running, unit costs and the associated prices must rise. Here too, the asset care function must also be the lowest cost producer.

In this scenario, asset care or maintenance might easily be viewed as an expense. Predictive maintenance is much less expensive than breakdown maintenance. It's time to calculate the differences in cost. Again, you have the answers for your facility.

Quality--If your company chose to compete on the basis of quality, then predictive maintenance becomes a foregone conclusion. Quality is necessarily a function of predictable and repeatable processes. Further, equipment outages translate into shutdowns and startups. Products produced during those occurances tend to be nonconforming and out-of-spec. In other words, they are not able to meet customer needs. Statistical process control techniques that ensure quality depend heavily on predictive maintenance. Frankly, predictive maintenance is critical to the success of any facility that supposes to compete on the basis of quality. Workable budgets based on the value of lost sales due to non-conforming beans will be of special interest to the bean counters.

Delivery--Should your company hang its haton its ability to deliver product on time, every time, meeting production schedules is the critical element in its competitive strategy.

Although the asset care function may have difficulties gaining access to production equipment, hard data--the kind generated through predictive technologies--about equipment condition becomes an absolute. Here reliable production processes are key to corporate success. Without reliable equipment, later than expected deliveries destroy the strategic advantage. Although some production folks find it difficult to release an asset to maintenance, they generally find it infinitely more difficult to miss their appointed output due to an unpredicted equipment outage.

The nit of all this: corporations fund strategy, not making someone's job easier. Further, seldom does a corporation care whether an engineer or technician feels adequately entertained by a chunk of technology.

Regulatory considerations
Complying with regulations such as those of OSHA or the EPA is a fact of life in our time. Similarly, companies that are ISO certified have an obligation to ensure that their process equipment does exactly what it is supposed to do. Further, the costs of non-compliance are astronomical. That being said, one can view the information generated and collected by predictive technologies
as being a critical facet of corporate compliance programs.

Those production assets that can be considered as having the potential to injure someone or to violate an environmental regulation definitely should be included in any predictive maintenance program. The information generated through predictive technologies is a primary defensive mechanism to avoid failures that have the potential to cause injury to people or the environment. If you have such systems, you need to know that those systems will not fail. Your equipment records, generated through predictive technologies, that show those assets in sound condition are of critical importance to you and your company from the legal perspective. Simply stated, by using predictive maintenance, the problem of equipment outages never occurs.

Identifying appropriate cycles
Most equipment responds to the environment in which it is used and to how it is used or misused. The operating environment includes such things as levels of moisture, presence of corrosive chemical in the atmosphere, ambient dust or particulate matter, and temperature. How an asset is used or abused focuses on factors such as machine loading, stop and start cycles, and operator and maintenance technician skill levels.

Environmental conditions--The conditions under which you operate your production assets often dictate the frequency and amount of predictive data. Plants located where corrosives, abrasive dust, or high levels of moisture are present necessarily will demand high levels of predictive maintenance. For example, machine elements like bearings in rotating equipment, process cooling equipment, and process ovens are susceptible to corrosion, dirt, and moisture. The very nature of the materials--bearings, shafts, fans, heat exchangers--in poor environments is such that equipment degradation is occurring. The question is not if, but how fast.

Your experience in your plant should provide an index to the life expectancy of the equipment components. You can use an initial frequency based on something less than the shortest known life span of that particular type of component in your situation.

Under adverse environmental conditions strategically critical equipment should receive frequent if not continuous monitoring. Specifically that means on-line monitoring where possible. Also, to prevent responding to false signals, it may be appropriate to use two separate technologies to monitor a single point.

For example, suppose that you have a critical fan application in your process. Further suppose that if the fan drops out because a bearing fails you lose production time and violate an EPA regulation. Here, I suggest that bearing vibration and temperature monitoring are necessary. Also, if it's my system, I want it hard wired with permanent pickups for automatic data collection and trending.

Keep in mind that both technologies indicate pending bearing failure and both technologies operate on different principles. Ergo, they confirm and compliment each other. A false reading
because a sensor or pickup failed on one system will not be confirmed by the
other system.

Now, consider the same fan in, say a warehouse application. What happens if it drops out. If there is no impending danger to people or property you might not even care. There fore, the question is: Is it still there?

Operating cycle--Many equipment items seem to do very well under steady-state conditions. That often translates into slow or almost imperceptible wear on machine elements. In many cases you can monitor these processes and systems on a periodic basis. Again, monitoring frequencies of something on the order of half of the shortest known life span of important machine elements may be sufficient. Your knowledge of the plant and environment is your primary source
of information.

However, if you have an application in which the equipment item experiences frequent stops and starts or variable loading, the monitoring frequency should be increaseddramatically. Also, depending on how critical nature of the asset is, it easily becomes a candidate for continuous monitoring with hardwired sensors. Equipment that experiences variable and surge loads may never see a steady state. Typical examples of this type of equipment loading occur with frequent machine setups, on and off cycling that is too rapid for proper bearing lubrication, and thermal shocking
in process heating equipment. It becomes highly desirable to use hardwired predictive technologies in these types of applications.

Proper alarm limits
Where does one get accurate information on specific alarm limits? At this point, I'll assume that you understand how your process works and how different elements work in conjunction with each other.

Ideally, you should be able to get specifics on appropriate alarm limits from the equipment manufacturer. Yet, most of the equipment specifications I've seen include information on equipment capacity and power requirements for the motors. There isn't a lot of specifics on vibration levels and so on.

However, even the basic information is not always available. For example, consider situations in which an equipment item is of foreign manufacturer. In that case the manufacturer's representative may not have the information on hand.

Also consider equipment that has been rebuilt or upgraded by a local machinery builder--this includes the maintenance engineer who became fed up with poor equipment design. Again, there probably isn't any hard information on appropriate alarm limits. Here, as above, you get to ferret out the information yourself.

Which technology is most suitable for a particular plant? It is entirely dependent upon the plant.

As a matter of fact, when you are purchasing critical spares, you can include in your purchase order that all replacement components come with appropriate upper and lower limits. That means that if you are buying bearings, you need to know what the maximum displacement, velocity, and spike energy limits are so you can set your alarm points for your vibration monitoring equipment.

You also need to know what the maximum operating temperature is so you can set your alarm limits if you are monitoring temperature. You may also wish to know what the maximum noise level is for serviceable bearings if you are using ultrasonic equipment in your predictive maintenance program.

Although the component manufacturer can be your primary source of information regarding alarm limits, the predictive maintenance equipment supplier is also a source for limits. Those folks understand their technology in depth and should not be overlooked as a source of information. However, you must consider that although they have broad experience, the equipment you have may be specialized and therefore may have different alarm limits.

Further if your system is capable of trending, you can request that component manufacturers provide you with information about useful component life. That means that you may be able to tap into the manufacturers research. For example, you may know that a particular bearing has a certain tolerable range of vibration. You may also know the failure point--that maximum tolerable vibration level. You may be able to find out the point at which that bearing is definitely going to fail--the point on the trend line that clearly indicates pending failure according to the manufacturer. As well, critical temperatures and noise levels may also be available for your predictive program if you ask. Keep in mind that what you are seeking is definitive and actionable information. You should be in a position to tell the folks in production that a particular equipment item is functional and will continue to run or that it must be taken out of service to replace a specific component.

The goal is not to see how much data you can collect but to get actionable information on your production system.

For example, rotating equipment nominally calls for vibration monitoring. Yet in the case of motors, a complete picture of equipment condition may also demand power monitoring. Further, both thermography and ultrasonics sense pending failures in rotating equipment.

If your critical applications hinge mostly on your process equipment's ability to transfer heat, the likely technology for you is thermography. This is suitable for ovens, boilers, and heat exchangers. If steam is an important utility in your plant, I'd also be very interested in using ultrasonics. Depending on how well your system is insulated and accessible, ultrasound may be your only flag to a hidden leak.

Thermography is also a primary source of information in analyzing electrical switching and power transmission equipment. Nothing shows weak electrical elements--connectors, fuses, breakers, transformers, and so forth--as clearly and readily as thermal imaging. The question is, what happens to the production levels in your plant if that electrical component drops out of service. While you're looking at that, it's always worthwhile to consider just how many production assets drop out with it.

Another approach to predictive maintenance that may be most appropriate for your plant deals with fluids. This is a group of technologies. One measures the amount of contaminants in fluids to determine if they are clean enough to do their intended job. The focus here is on preventing the lubricating fluid from ever getting dirty--the proactive approach. One can also perform oil analysis to determine if a lubricant is still capable of lubricating through looking at the chemistry of the lubricant. Then comes the ferrography. Here you can determine through inspecting the metallic material in an oil which components are wearing.

Frankly, predictive maintenance is critical to the success of any facility that supposes to compete on the basis of quality.

Frankly, how much information one gathers and the type hinges on the law of diminishing returns. Those who use predictive technologies must always be in a position to cost-justify their expenditures. How much information you glean and how often you gather depends upon the critical nature of your process and equipment.

In a perfect world every piece of equipment would arrive at the plant hard wired and ready for connection to a continuous monitoring system that interfaced with the computerized maintenance management system. You would have a complete report on the actual equipment condition down to the last detail--all the base line data and alarm limits for your predictive maintenance program. But that is not current reality.

Therefore, the goal is not to see how much data you can collect but to get actionable information on your production system. That goal must align itself with your corporate strategy.

So what is the appropriate level of predictive maintenance? You are the most likely person to make
that judgement.