Techniques for Predictive Maintenance

Predictive maintenance techniques help determine the condition of in-service equipment in order to predict when maintenance should be performed. This maintenance technique offers cost savings typical time based preventive maintenance.

The main value of Predicted Maintenance is to prevent unexpected equipment failures while maximizing resources. The key is “the right information at the right time”. By knowing which equipment needs maintenance, maintenance work can be better planned (spare parts, people etc.) and what would have been “emergency maintenance” are transformed to shorter “planned maintenance”. Other advantages include increased equipment lifetime, increased plant safety, fewer accidents with negative impact on environment, and optimised spare parts handling.

Predictive maintenance utilizes nondestructive testing technologies such as infrared, acoustic (partial discharge and airborne ultrasonic), corona detection, vibration analysis, sound level measurements, oil analysis, and other specific tests. New methods like Predictive Analytics use actual meter and indicator readings from the machinery to predict failures. Predictive Analytics is the type of statistical methods used by Predikto to help maintenance teams perform preventive maintenance and avoid a failure.

The chart below came from a Predictive Maintenance paper. You can find the entire write up at: http://www1.eere.energy.gov/femp/pdfs/OM_6.pdf

PredictiveTechnologiesChart

What is Predictive Analytics

Predictive analytics is an area of statistical analysis that deals with extracting information from data and using it to predict future trends and behavior patterns. The core of predictive analytics relies on capturing relationships between explanatory variables and the predictive variables from past occurrences, and exploiting it to predict future outcomes.

Predictive modeling draws from statistics and optimization techniques to extract accurate information from large volumes of data. Modeling techniques produce interpretable information allowing maintenance personnel to understand the implications of events, enabling them to take action based on these implications.

Predictive analytics builds on prior investments in enterprise asset management (EAM) systems, combines real-time data from sensors and other acquisition techniques with historical data to predict potential asset failures, and enables the move from reactive (scheduled, break-fix) to proactive (condition-based, preventive) maintenance.

Predictive analytics can be used to analyze the real-time data from the sensors in the context of historical data and asset information held in the EAM system to predict future conditions such as faults or failures and produce alarms or schedule maintenance or replacement. Predictive analytics also complement other existing systems such as data historians and SCADA systems.

Traditional Predictive Maintenance is missing a key component

Predictive Analytics complements and improves on traditional DCS, vibration, and other systems making up traditional predictive maintenance practices. These predictive systems are not intended to provide early warning. Rather, they are deployed to prevent significant equipment damage or failure.

Traditional monitoring systems are set to detect and prevent catastrophic damage. Alert levels must be set relatively broadly in order to prevent false alarms due to a wide range of operating conditions. Once a potential failure is detected, the time an operator has before equipment functional failure is often very brief, thereby limiting the options available to correct the situation.

Predictive Analytic software by using statistical models leverages existing infrastructure and systems to provide analysis and earlier warning of emerging issues.  Instead of reacting to the emergency alarm, statistical models for predictive analytics can give you warnings days in advance.

Predikto has developed these predictive analytics models.  Give us a call to see if your data and maintenance processes qualify.

$57 Trillion in infrastructure will be needed by 2030

McKinsey Global Institute released a study this year stating that the world will need $57 Trillion (with a T) infrastructure projects by 2030.  That spending is just meant to repair our existing infrastructure, which is in a fairly abysmal state in the U.S.

The most recent “report card” from the American Society of Civil Engineers gives the U.S.’s infrastructure a “D” overall and estimates that we need to spend $2.2 trillion over the next five years to get it up to snuff.

The article also discusses an interesting approach by Chile to prioritize infrastructure projects based on predicting the highest need.  A bulk of the infrastructure needs where in areas like roads, power grid, water, and telecommunication.  The US resembles countries like South Africa and China instead of other developed countries like Germany and Japan.

Advanced manufacturing and predictive techniques can help reduce the annual spend from $2.7 Trillion to $1.7 Trillion, but we like the overall theme of the report in that we need to get smarter at tackling these big infrastructure problems.  Similar to reactive vs. predictive asset maintenance, our infrastructure improvement plans should be more predictive and reactive.

Replace intuition with predictions

Many industrial manufacturing companies run a reactive maintenance organization. That means they wait for things to break before replacing parts. Studies have proven that running a preventive maintenance plan reduces expenses by up to 75% in the long run.  The challenge is that many of the Preventive Maintenance schedules were developed on intuition.  Changing the oil filter once a year is enough.  Most organizations are sitting on a wealth of data like prior downtimes (failures), work orders, parts, and the rich information captured in the indicators / meters.

Predikto has the skills and technology to analyze the maintenance data in order to provide a preventive maintenance plan and predictions to help reduce failures and reduce the overall cost of maintenance.

With Predictive Maintenance, organizations can spend less time and fewer resources repairing things too early or avoiding fixes when it is too late. Industry studies have documented cost reductions in the 3-5X range for catching problems early as opposed to catching a problem once there is significant damage to the equipment. Companies can spend more time focusing on what will happen next and be smarter about the preventive maintenance. It is much easier to fix a problem before it happens rather than reacting after the production floor has come to a screeching halt.

Bye ‘Big Data’. Hello ‘Smart Analytics’

John De Goes wrote an interesting blog posting on Venture Beat where he describes that the term “Big Data” is dead.  His arguments are strong and Predikto agrees with his statements.  We run into the term Big Data and Predictive Analytics a lot.  In many cases it does not apply and people have not agreed on a definition for Big Data.  Sometimes we see vendors using Predictive Analytics when their software or technology solution does not predict the future using machine learning or statistical methods.  The “Predictive Analytics” technology just sends an automated notification using advanced GUI to someone to react once a sensor has reached a threshold.  We also run into many cases where someone talks big data, but you can open the CSV files in Excel.

Predikto, like John De Goes, like the term smart data, but we like to focus on analytics that enables an action that impacts the bottom line.  We call this “Smart Analytics”.  When you combine the human skills from smart scientists, developers, process experts, and infrastructure gurus, with the technologies used to enable solutions, you have a winning combination.  It’s this careful balance of combining all components into a solution that enables an action that makes it really challenging and really FUN.

smartanalytics