Predictive maintenance leaves the back room as search for higher profits marches on

Oct. 18, 2004
The relentless search for higher profits is finally bringing predictive maintenance out of the back room.
Executives are gaining new knowledge of and respect for good maintenance practices. Years of cost- and staff-cutting and attrition have taken their toll on equipment reliability. ERP systems are red-flagging asset availability and costs of lost production as reasons why executive bonuses are flat. The best are coming to their plant engineering, maintenance and reliability departments looking for answers.But they’re not willing to spend more on labor. Benchmark studies show them their maintenance dollars are already excessive. In their opinion, costs are too high and they’re not getting enough return on investment.Those same studies tell them that predictive maintenance is high on the list of best practices. They know something about predictive maintenance, and now they want to hear about how industrial-strength applications can make their plants run both more cheaply and reliably.Management sees the light
Back in the boom years of the 1990s, while many other industrial facilities were allowed to dabble in complacency, deregulation and privatization, management was putting pressure on electric-power-generating facilities to significantly increase reliability and efficiency. “Ten years ago, we had a very strong preventive and corrective program,” says Roger Cole, reliability-based maintenance coordinator for AEP’s John Amos Power Plant in Big Scary, W.Va. (formerly known as the Big Scary Power Plant). “We had some predictive, but we didn’t leverage it. We had twice the workforce, so we were able to do a lot of testing.”Then the utility industry began to deregulate and cut costs. “We went through a period of shrinking workforce: When people left or retired, we didn’t replace them.” Cole says. “We got to the point where we didn’t have enough manpower to keep the plant running reliably, to avoid catastrophic failures and manage the workload. We were in a run-to-failure mode that became very expensive due to equipment damage and load curtailments.”In many industries today, efficiency has been sacrificed in the name of cost-cutting. “Maintenance saves $10, but it costs operations $20,” says Neil Cooper, general manager for Invensys Avantis. “It’s not that the technologies aren’t there. The problem is corporate hasn’t seen the proper role of the plant, and plant managers have not seen the big picture, so they drive maintenance costs down separately from trying to improve availability and output. Before we make maintenance decisions we have to weigh them in the context of corporate goals.”There hasn’t been an awareness at senior and corporate levels, but there is a change in progress -- a dawning realization that maintenance is not a necessary evil, it’s a significant contributor to the upside (Figure 1).From what our practitioner, expert and vendor sources tell us, we can see that there’s a tsunami of predictive technology unfolding as we write. Read on and you’ll be ready to talk with management about why now is the time to invest in more effective asset management. Simply put, it costs but it pays, and you need it to compete.Maintenance pays
Figure 1. Effective maintenance practices boost the bottom line by improving operating equipment effectiveness (OEE), reducing the cost of goods sold and avoiding capital expense. Streamline RCM
From high-resolution, easy-to-use infrared cameras you can actually afford to sophisticated vibration algorithms to oil analysis services that deliver timely reports you can put into action, there’s no shortage of practical predictive technologies. There is also a well-defined hierarchy of maintenance practices (Figure 2). The problems, as usual, are in implementation.Nowhere is this more obvious than with the engineers’ favorite steamroller, reliability-centered maintenance (RCM). Do a thorough root cause failure analysis, evaluate the risks, perform the prescribed preventive procedures and predictive inspections, and rock-solid reliability will surely follow.But it costs too much. “Proper RCM as defined by the original gurus is too expensive,” Cooper says. “It’s not practical. You have to do the 80/20 rule.”The essence of streamlined RCM (SRCM) is that 20% of the input gives 80% of the results. “SRCM is more cost-effective because it helps you focus more effort on critical failure modes,” says Scott Brady, general manager of decision support systems for SKF. “Whereas regular RCM makes you look at all failure modes, only to find that 70% are non-critical and not important,” he says.You don’t have to know everything about every asset. “We don’t analyze what happens if a washer fails on a pump foot, we look at what happens if the pump fails and determine the most common failure modes,” Brady says. “We identify critical systems, look at common failure modes, and put tasks in place to predict and prevent those failures. We gather what we need to know to keep the asset working, and use decision support to drive the data to decisions.”Use multiple technologies
Determining, monitoring and diagnosing the most common failure modes calls for multiple condition-monitoring techniques. Back at the John Amos Power Plant, technologies are now combined to detect and zero in on causes of incipient failures. “We use several technologies,” Cole says. “Infrared, vibration, oil analysis -- no one technology can do it alone. We might see a vibration, then take oil samples to help pinpoint the cause. Vibration doesn’t solve all the problems. It’s excellent for antifriction bearings and gears, for example, but not so good for sleeve bearings. And we’ve made our major finds with infrared.”Cole describes a 5,000 hp motor where the bearing failed on the inboard side. “It was so bad the rotor dropped into the stator, but the motor RTDs were all on the outboard side, so the motor temperature readings looked normal,” he says. “Infrared showed the inboard side was tremendously hot.” The plant does all its motor rebuilds, and now puts RTDs on the inboard side as well.Oil analysis, vibration, thermal analysis, etc., have traditionally been seen as individual systems. “That’s made them difficult to use,” says Rick Schlitz, director of the new Rockwell Center for Integrated Condition Monitoring in Houston. Vibration data might show that amplitudes are increasing, but what’s going on with the process? Maybe the load has increased and the vibration is not indicating a problem.“People couldn’t look at all the information, it was too hard,” Schlitz says. “Now data are being pulled together in integrated systems, in one location or at least communicating with plant historian and CMMS systems.”Bring systems together
Separate condition-monitoring systems are easier to implement, but have limited capabilities. Panhandle Energy, Houston, operates more than 10,000 miles of mainline natural gas pipeline extending from the Gulf of Mexico to the Midwest and Canada. With a heterogeneous and widely distributed mix of engine-driven natural gas compression equipment, some dating back to the 1920s, the company has been experiencing too many failures that put equipment off-line for six months at a time, at costs of $1 million to $2 million.“We have the ability to collect the data; we have the computers and the memory. The problem is it’s impossible to analyze it all,” says Patrick Pittman, Panhandle principal engineer of technical services. “So we’ve been collecting a lot of data that we’re not really using.” The company is implementing software that analyzes standard operating information -- vibration, pressures, speeds, temperatures -- to indicate the condition of equipment. Pittman says, “We get it from the existing instrumentation through the control systems and monitor it with predictive condition-monitoring software.”Emerging standards are facilitating systems integration (“Open season for open standards,” Plant Services, August, p. 64). “Adoption of open standards like the condition-based maintenance (CBM connector, OpenO&M, MIMOSA, OPC and ISA95) are making it easier to integrate condition-monitoring and analysis, diagnostics, and information systems,” says Ian Wray, vice president, product management for Indus. The vision goes beyond combining predictive technologies. “Separate departments must be integrated: condition monitoring with maintenance with information.”Condition information should be made available to production and business systems. Supply-chain excellence means mapping orders to the plant and efficiency levels, and asking plant managers what they need to do on the operating side to meet these requirements.Management is seeing the problem, but it’s up to engineers and maintenance professionals to show them the answers. “The conferences are buzzing,” Cooper says. “It began with the engineers; now it’s the maintenance guys. Operations people are the most cautious, but it’s the engineering and maintenance guys who are key. It starts in engineering or maintenance, or in the case of business systems, IT.”The benefits of plant-wide asset management go beyond improved production efficiencies. This truly collaborative effort involves input and action of all divisions of the enterprise. Maximum value is achieved when all resources are working together, interacting and sharing knowledge. This is especially critical at a time when plants are under the gun to do more with less. Collaboration helps address the erosion of experienced workers -- when senior operators and technicians leave the workforce, their experience goes with them and they are not likely to be replaced.“Collaborative enterprise asset-management systems help capture this valuable information and experience,” Cooper says, “And what price can you put on knowledge?”Find your place
Figure 2. A definite hierarchy of progressively more efficient maintenance practices has beenestablished over time. Where does your plant practice fall?Centralize monitoring
When information is integrated, who could resist making it available to Web browsers and handhelds so monitoring can be done at central locations and the results can reach out and touch everyone? “Costs are down and bandwidth is up,” Wray says. With intranet and Internet technology, centralized monitoring can alert local forces, and online monitoring can increase lead times for predicting failures. “A 10-minute or one-hour warning is not enough,” he says. “Plants need days or weeks to take efficient action.”Centralized analysis allows one team to do analysis for multiple plants or companies, which can leverage a large organization’s expertise over multiple plants and allow all sizes of companies use third-party service providers more efficiently.At both Panhandle Energy and Air Liquide, multiple sites feed data from CMMSs, plant WANs and portable data collectors to a centralized system where company engineers and vendors can analyze and troubleshoot.Results are sent back out to site CMMSs. “We’ve added real-time resource optimization capabilities,” Wray adds. “You can see the locations of technicians, assign them and route them efficiently with sophisticated workforce allocation and scheduling tools.”“Companies are using more services,” Schlitz says. For example, if you can’t justify or keep a vibration expert on staff, you can have vibration monitored by a service company. Such companies may have experience with a broad range of equipment and machinery. “We can establish a reliability database for a customer, which is valuable when looking at new equipment,” he says. “Our audit can establish drivers for a monitoring program, which can be carried out on or offsite. We can help companies establish a program and migrate from outsourcing to insourcing, typically during a period of one to three years.”Recognize patterns
Predictive condition-monitoring software bred for aerospace and advanced control applications is coming into industrial applications. Process plants have been focusing on advanced-control and process optimization to be competitive. “So plants are running at capacity -- pushed to the max -- in the process industries,” says Eddie Habibi, president and CEO for PAS. “The problem is reliability. A short, unexpected shutdown costs more than all the advanced controls buy them in a year.”In the past, with annual shutdowns, many incipient problems were caught with preventive maintenance and timely inspections. Now plants are running four or five years between turnarounds, and they can’t rely on good luck to catch problems.Software designed to tune processes within an operating envelope or statistically analyze variations in instrument readings and process parameters can detect gradual deterioration and call for attention before equipment breaks down. “On a compressor, for example, the operator will not see gradual deterioration because the alarms are all set at a fixed level,” Habibi says. “Problems come on gradually and are reflected in many variables, so you need pattern recognition. Normal versus abnormal operation should be seen not as an alarm setting, but as a multivariable parameter determination. If two or three variables are moving out toward the edge of the envelope, you know you have an abnormal situation.”Packages draw data from a combination of the control-system historian, online sensors, and manual data entry, and make decisions based on user-defined setpoints or proprietary algorithms. If significantly abnormal operating conditions are detected, the systems can alert the maintenance staff.“We’re using SmartSignal,” Panhandle’s Pittman says. “We didn’t have to add a lot of sensors to get it going, but we had to repair some of the existing sensors. We already have some vibration and temperature sensors, for example, exhaust temperature sensors on the 4,000 hp to 10,000 hp engines that drive natural gas compressors.”Panhandle is remotely detecting problems that used to require onsite physical inspection, and is able to act more quickly. “For example, we detected a fuel-injector gasket leak two to three days before it would have shut the engine down under the old system,” Pittman says. The leak increased fuel consumption and emissions, and could have escalated into a fire. The approach can work for any application where there is, “A multitude of variables and you’re having problems,” he says. “It recognizes trends that are not otherwise very visible.”Empower operations
Maintenance, operations, safety and planning are traditionally separate functions, but in more advanced companies, departments understand that they have the same goals and are cooperating to achieve them. The most critical cooperation is between maintenance and operations.With today’s computerized controls, operators are increasingly sitting in air-conditioned control rooms, not walking around the plant understanding what’s going on. But operators can be with machines 24/7, whereas maintenance personnel cannot. Where it makes sense, operators can be out in the field making inspection of machines and components. “The idea is to focus all activities on equipment effectiveness, activities that contribute to optimal production and asset return,” says Mark McGinn, director of product marketing for SKF.Operators can perform a subset of predictive and preventive tasks that range from handover procedures for shutdown and maintenance, post-maintenance inspection, and operator-involved failure analysis and prevention to operator-performed maintenance. “Not overhauling machines, but taking ownership of some of the maintenance tasks like checking lubrication levels, topping up, changing felts in paper industry, etc.,” McGinn says. “It’s really about change and taking ownership of the machines they’re operating, and it gives earlier indications of how the machines are performing.”Support decisions
Bringing operations, reliability and asset-management information together offers an important opportunity to capture knowledge and help make better decisions. “There’s a brain drain,” Brady says. “People are not going into maintenance and operations. The average person in the field is 50 years old, and 80% of our clients expect to lose 40% to 60% of their experienced people in next five years. But only 10% have a good apprenticeship program.”Knowledge that remains uncaptured or not passed on will be lost forever. Decision-support systems can capture the knowledge of personnel with years of experience in a plant with its own unique operating characteristics.Information from operations, reliability and asset-management systems can be used to support business objectives with strategies and decisions that optimize the efficiency of production equipment and maintenance personnel (Figure 3).Efficient assets
Figure 3. Information from operations, reliability and asset-management systems can be used to support business objectives with strategies and decisions that optimize the efficiency of production equipment and maintenance personnel.

“The missing links between data collected by sophisticated equipment and using that data to make decisions have been people,” McGinn says. “We automate 80% of it. We take the knowledge gained by experience over the past 20 to 30 years and put it into a system that uses that knowledge to detect problems, diagnose them and put the results where they can be used to make decisions.”

Find pre-instrumented equipment
Another trend is increased design and delivery of equipment with pre-integrated smart sensors and diagnostics. This eases applications such as pumps and compressors, where buying, installing and wiring up systems has been difficult and expensive.

Per-point costs are coming down, driven in part by wireless sensors. Software and diagnostic tools are in second and third generations, so they’re easier to use and don’t require specialized experts.

As this kind of equipment becomes increasingly commoditized, some vendors want to differentiate themselves (and generate service revenues). They’re providing pre-instrumented equipment and offering lifetime equipment services.

Other vendors are more reluctant to add predictive instrumentation. After all, they’re in the parts business, too.

Go over the wall
Any of the vendors mentioned in this article, and many others, stand ready to offer advice, equipment, software and services to get a plant started on effective predictive maintenance or they can take an existing program to the next level. Of course, all the typical obstacles to change are there, including the perennial Big Three: financial justification, management commitment and cultural change.

“Most of the companies that are trying to work on this are in brownfield plants, and the main barrier is the culture,” Cooper says. “You have to change the mind-set where maintenance guys are cutting overtime, cutting costs and spending all their time out there trying not to have anything break.” Meanwhile, no one has dropped the pressure on maintenance engineering to reduce costs. “For further cuts,” he says, “They have to look at predictive. What cost a lot of maintenance last year? How could we have avoided that?”

Your three main tools for overcoming the obstacles are leadership, training and metrics. One highly motivated individual is usually the difference between success and failure. “You need leadership by senior management that lives and breathes it,” Brady says.

Training is pivotal -- people need to learn what, why and how, as well as the benefits. Brady says, “If we can’t answer those questions to their satisfaction, the program will fail.”

Metrics are critical. Find measures that reflect maintenance effectiveness and follow through with them. “There are a lot of ways to attack the culture issue, but leadership, training and metrics are really how you drive it home,” Brady adds. “Not every plant can be successful.”

Success breeds success. Start by putting condition monitoring on the most critical equipment. Once the plant can see the value, issues will tend to drop away.

The John Amos Power Plant started digging out of its reactive rut eight years ago by teaming up with the Electric Power Research Institute (EPRI), identifying 30 motors and trying various techniques until they found a combination that works. Then they expanded the program to small equipment and pumps. Significant finds have more than paid for the program along the way.

“We haven’t solved all the problems yet,” Cole says. “People talk about the life cycle of predictive maintenance -- it starts off with high payback on low-hanging fruit, then as the plant gets more reliable and there are no big finds, management starts to see it as maybe not so valuable. So they cut it and end up back in a reactive rut. But we haven’t seen that. After eight years, we are still finding potential problems and saving significant costs.

 “One of the biggest obstacles is management doesn’t want to give personnel up to perform predictive programs. Bite the bullet and do what you have to: Farm it out, train somebody -- it’s going to pay for itself.”

Figures 1-3 courtesy of SKF Reliability Systems.

Paul Studebaker is editor in chief of Plant Services magazine. E-mail him at [email protected].