An incredibly simple guide to starting a predictive maintenance program at your facility
[alert type=”info” icon-size=”hide-icon”]Key Takeaways:
- Predictive maintenance is a culture, fostered by forward-thinking executives and a dedicated, trained team of technicians.
- A mature preventive maintenance system with smooth processes and high schedule compliance provides an excellent foundation for predictive maintenance efforts.
- Start small with a critical asset and pilot a well-thought-out, planned program first.
- Establish a continuous implementation and improvement process to keep steps well documented for easy knowledge transfer.[/alert]
Predictive maintenance (PdM) can often be viewed as the panacea of maintenance issues these days. However, starting a predictive maintenance program can be complex and expensive, and not all facilities may be ready for such a program. In order to achieve success, a solid foundation must be followed by a methodical plan of action. Here’s a guide to help break down the process into easy-to-digest steps for your facility.
Are you ready?
Everyone thinks of predictive maintenance as a way to anticipate potential breakdowns and failures in equipment and, essentially, “catch them just in time.” Although that is a big part of predictive maintenance, it is not the core. Predictive maintenance is, instead, a philosophy. It focuses on using your actual asset operating conditions to optimize your facility as a whole.
Start with establishing a predictive maintenance mindset and culture. You need strong, visionary, and expert leadership as well as a trained, motivated maintenance team.
Once you have the people in place, you’ll need to look at the processes and technologies you are currently using. If you’re operating in a primarily reactive maintenance mode, you’ll need to move to a preventive maintenance program first. You’ll need the right tools to handle new processes on your complex assets. Implement or refine your computerized maintenance management system (CMMS) to serve as the foundation of your maintenance upgrade and help support predictive maintenance technology.
What are the steps to get started?
Let’s say your organization has laid a strong foundation.
You, as a group, have adopted the predictive maintenance attitude. Technicians work regularly to identify issues before they become problems. Emergency work orders are minimal. Your mature maintenance system helps processes run smoothly, and schedule compliance is high. You have some reliable historic data and a solid handle on your machines and assets. Essentially, preventive maintenance is the norm.
This well-oiled maintenance team is also led by visionary management. Leaders who are communicative, flexible, and ready to take on the future challenge of predictive maintenance. You are ready.
What’s next? Here are some simple steps to get started:
Step One: Start Small
No company has unlimited resources. Even with the right attitude, you’ll face some level of skepticism any time you implement a new system. That’s why it’s important to start small. Select one or two critical assets to pilot through an initial predictive maintenance implementation. That way you’ll be able to take steps slowly and address mistakes without affecting a large number of assets. There’s nothing like a successful pilot program to build the momentum you need in expanding a predictive maintenance program.
Step Two: Identify PdM Ready Assets
Determine which assets are ready for or deserving of a place in your new predictive maintenance program. For some assets, predictive maintenance simply is not the correct approach. These assets may be expendable or require low-cost solutions. Additionally, you don’t want to pilot a new program on your most expensive or most important asset. Just as a precautionary measure, start your predictive maintenance program with a less significant piece of equipment so that as you scale to more critical assets, your processes will be better than they were originally were.
To decide which assets are most critical, start with failure codes. Codes are asset-based or inspection-based. Common failure codes signify poor maintenance practices, user or operator errors, calibration issues, and asset defects. Some failure codes may not affect immediate performance, but others can cause significant downtime.
Another way to categorize your assets is by the type and number of tasks required to maintain an asset. For example, value-added tasks, comprehensive calibration tasks, “specialist” tasks, interval tasks and repeatable tasks should all be taken into consideration because they all demand high financial and human resources.
Evaluate both failure codes and maintenance tasks as well as the history and condition of particular assets. Frequent breakdowns indicate high repair costs and a great deal of downtime.
Think about how critical an asset is to your daily production line. If a machine is used infrequently, it can probably wait. Your ideal candidate for a pilot asset is one that is moderately used and often breaking down. It is important to address, and you won’t suffer huge losses if your predictive maintenance program takes a little while to refine.
For example, HVAC equipment is widespread and crucial in nearly all facilities. Routine maintenance is appropriate for these machines, but this approach doesn’t always discover the root cause of frequent failures. HVACs often require many little tasks scheduled based on time. Predictive maintenance, on the other hand, can analyze the current condition of HVAC equipment and schedule maintenance tasks only when they’re necessary to prevent it from failing.
Step Three: Identify Resources Required
The next step is to look at all the resources you will need to implement a predictive maintenance program. Here are the key categories:
- Labor. Evaluate the number of hours you’ll need to start and run your program. This should include supervisor and management hours, planning and updating hours, and total craft hours by job sector and skill level. You’ll also require support staff hours. These should include not only clerical assistance but also tangential departments like materials handling.
- Materials. Depending on which critical asset you decide to start with, you’ll need to identify the materials needed to move forward. This includes what items and levels you need to stock and when to replenish.
- Facilities. It’s important to take a hard look at your physical space and determine where predictive maintenance tasks will take place. This may include where to establish your shop, tool room, inventory, training, and repair space.
- Technology. A wide variety of tools and technology are available today. Determine which data collectors, sensors, infrared cameras, and analysis software you’ll need. There are several types of sensors, as well as many vendors you can purchase from. Monnit and TitanGPS are great resources with many types of sensors available! Below, you’ll find a list of a handful of the MANY sensor types:
- Temperature Sensors
- Pressure Sensors
- Chemical and Gas Sensors
- Level Sensors
- Vibration Sensors
- Motion Detectors
- Location Tracking,
- And more!
- Training. Investing in your team in terms of training, education, and certification will help you implement and run your program. It will also solidify that all-important culture.
Step Four: Implement Asset Monitoring and Begin Data Collection
At this point, you can begin monitoring your pilot asset. Monitor the pilot asset over as much time as you can before implementing your PdM program. This baseline data becomes crucial as it is the foundation for your entire pilot program. As any red flags or failures arise, perform and record needed maintenance. Look back through machine records, gather manufacturer information (on maintenance tasks, frequency, safety, etc.), and include conversations with maintenance technicians or operators who are familiar with your pilot asset. Consulting staff for working knowledge of the equipment will give you unique insights that you may not be able to pull from the hard data.
Purchase and install necessary sensors and begin collecting your PdM data. The sensors will monitor current conditions as well as the pilot asset’s failure codes. Data collection will be very specific to the critical asset you select. Three of the most common avenues for collecting data in many facilities:
- Electromechanical Systems: These are common in many facilities today. As a result, vibration sensors and monitoring usually play a huge role in predictive maintenance. In many cases, data collectors and software must work together to collect, monitor, and evaluate vibration energy.
- Thermography: This monitors infrared energy as a way to check machinery conditions. If certain surface areas are generating too much or not enough heat, a predictive maintenance system can notify a technician to explore the problem.
- Lubrication and Wear: Monitor the oil conditions of critical equipment and wear particles regularly. This data will help you make determinations about frequency of oil changes on equipment. Traditionally, this has been determined by time or usage. With predictive maintenance technology, you can use actual lubrication and wear data to ensure maintenance is completed at just the right time.
Note that it becomes crucial at this point to understand data sheets – or have someone or a system that does – so that you’re able to interpret and communicate the findings of this data.
Returning to our HVAC example from Step One, the most common PdM technology applied to HVACs are vibration sensors. With these sensors, you can better monitor structural vibration data from the compressor. Another common sensor used in HVACs are level sensors to monitor and analyze the machine’s oil.
Step Five: Create Machine-Learning Algorithms to Predict Failures
Everything begins to come together during this step. Using the data you’re collecting with your new sensors, you can develop algorithms that will predict equipment failure before it actually occurs. Many facilities will use Failure Mode and Effect Analysis (FMEA). FMEA is a systematic process analysis tool that helps to identify failure modes and its effects and causes. It also looks at frequency, detection rate, and severity.
Formulas, algorithms, and machine learning processes can then use this data to evaluate current conditions of critical assets as well as to predict the time left before a failure occurs. These are prognosis algorithms, which compare machine conditions at present with the baseline data you worked hard to collect in order to estimate when failure will occur. Depending on your industry and facility, a wide range of algorithms can be used. Once you’ve amassed enough data using your algorithms, you can begin to make intelligent maintenance decisions and improve your predictive maintenance program further.
Again, having a member of the team who is data-literate is crucial. Your algorithms are designed to collect and digest data, but you still need a person who can evaluate it and adjust the algorithms if needed.
Step Six: Apply to Pilot Asset
Now that you have the monitoring in place – continuously collecting data – and the algorithms to process it, your technology is ready to go. Apply it to your pilot asset.
“Applying” the predictive analytics to your pilot asset looks something like this:
- Sensors collect data and send it to the cloud,
- PdM algorithms are applied to your data center,
- Reports and insights are generated.
Cloud-based data and analytics automate the data collection, storage, and evaluation, making the whole process incredibly quick.
Once more, we can use our HVAC example. If the vibration sensors that you installed on your asset detect excess noise or vibration motion, the sensors will identify it as an abnormality by comparing this data to to baseline data or to the machine’s specifications. This predictive technology pinpoints the location and cause of the problem before it even occurs. Once the issue is addressed, sensors will continuously recheck the solution to ensure it is effective.
Similarly, the level sensors that you installed on your HVAC to monitor and analyze oil detect not only the amount of oil to indicate when it needs to be changed, but also it detects places where harmful wear is occurring. These sensors record wear data in various locations of the HVAC and point you to where damage control is the most urgent.
Overall, predictive technology and analytics on your HVAC will improve equipment life and minimize uneccessary maintenance costs.
Step Seven: Establish Continuous Implementation and Improvement Process
Once you’ve experienced some success, you’ll want to put into place a cycle of implementation and improvement as you bring your other critical assets on board. Be sure to run your data to measure against preset KPIs so you can illustrate the return on your investment and make a case for expansion.
According to industry experts, one of the key reasons for predictive maintenance failure is a lack of consistency and long-term methodology. In order for a program to function, grow, and evolve, facilities must monitor data on critical assets on a regular, ongoing basis. This means that the processes must be well understood and established. As you experience technician and talent turnover, you’ll need to be able to transfer the knowledge between the old and the new to keep the program moving.
Based on experiences of other successful implementations, you can expect a five percent to ten percent cost savings in operations, maintenance, and repair spending. Overall maintenance costs will be reduced by that amount as well. According to the Department of Energy, a predictive maintenance program has the potential to result in a tenfold return, a 40 percent decrease in downtime, and a 70 percent reduction in breakdowns.
Predictive maintenance promises to deliver incredible results to facilities that implement programs effectively. To recap, the steps to starting a predictive maintenance program at your facility are:
- Start small. Be sure you have strong leadership and a hard-working team in place, and don’t try and take it all on at once.
- Identify PdM ready assets. Choose your pilot assets wisely, not at random.
- Identify resources required. Prepare the labor, materials, facilities, technology (sensors), and training to start your PdM program.
- Implement asset monitoring and begin data collection. Install new technologies and collect baseline data.
- Create algorithms to predict failures. Design algorithms to process and analyze data from your new PdM technologies.
- Apply to pilot asset. Use PdM technologies and algorithms to monitor and provide insights on your pilot asset.
- Establish continuous implementation and improvement process. Synthesize learnings to prove your PdM program’s value and continue to improve before begin taking steps to scale.
In a business climate where markets grow more competitive each day, a predictive maintenance culture and program will give adopting companies a significant advantage. Start with fostering the right philosophy, invest in solid technology. Then, establish your program in small, methodical, and measurable steps. You’ll be surprised at how quickly you’ll build momentum within your organization once you start experiencing bottom-line results.