Manufacturers and operators employ array of maintenance strategies, all of which can be broadly categorized as below:
- Corrective Maintenance
- Preventive Maintenance
- Predictive Maintenance
Corrective maintenance is the classic Run-to-Failure reactive maintenance that has no special maintenance plan in place. The machine is assumed to be fit unless proven otherwise.
- High risk of collateral damage and secondary failure
- High production downtime
- Overtime labor and high cost of spare parts
- Machines are not over-maintained
- No overhead of condition monitoring or planning costs
Preventive maintenance (PM) is the popular periodic maintenance strategy that is actively employed by all manufacturers and operators in the industry today. An optimal breakdown window is pre-calculated (at the time of component design or installation, based on a wide range of models describing the degradation process of equipment, cost structure and admissible maintenance etc.), and a preventive maintenance schedule is laid out. Maintenance is carried-out on those periodic intervals, assuming that the machine is going to break otherwise.
- Calendar-based maintenance: Machines are repaired when there are no faults
- There will still be unscheduled breakdowns
- Fewer catastrophic failures and lesser collateral damage
- Greater control over spare-parts and inventory
- Maintenance is performed in controlled manner, with a rough estimate of costs well-known ahead of time
Predictive Maintenance (PdM) is an alternative to the above two that employs predictive analytics over real-time data collected (streamed) from parts of the machine to a centralized processor that detects variations in the functional parameters and detects anomalies that can potentially lead to breakdowns. The real-time nature of the analytics helps identify the functional breakdowns long before they happen but soon after their potential cause arises.
- Unexpected breakdown is reduced or even completely eliminated
- Parts are ordered when needed and maintenance performed when convenient
- Equipment life and there by its utilization is maximized
- Investment costs for implementing the condition-based monitoring (CBM) system
- Additional skills might be required to effectively use the CBM system effectively
Predictive maintenance, also known as Condition Based Maintenance (CBM) differs from preventive maintenance by basing maintenance need on the actual condition of the machine rather than on some preset schedule or assumptions.
For example, a typical preventive maintenance strategy demands automobile operators to change the engine oil after every 3,000 to 5,000 Miles traveled. No concern is given to the actual condition of vehicle or performance capability of the oil. If on the other hand, the operator has some way of knowing or somehow measuring the actual condition of the vehicle and the oil lubrication properties, he/she gains the potential to extend the vehicle usage and postpone oil change until the vehicle has traveled 10,000 Miles, or perhaps pre-pone the oil change in case of any abnormality.
Underlying preventive maintenance is the popular belief that machine failures are directly related to machine operating age, which studies indicate not to be true always. Failures are not always linear in nature. Studies indicate that 89% of the problems are random with no direct relation to the age. Table 1 showcases some of these well-known failure patterns and their conditional probability (Y-axis) with respect to Time (X-axis).
Fig: Failure Probability Curves
Complex items frequently demonstrate some infant mortality, after which their failure probability either increases gradually or remains constant, and a marked wear-out age is not common. Considering this fact, the chance of a preventive maintenance avoiding a potential failure is low, as there is every possibility that the system can fail right after a scheduled maintenance. Thus, preventive maintenance imposes additional costs of repair. Predictive Maintenance reduces such additional costs by scheduling maintenance if and only when a potential breakdown symptom is identified.
However, the costs of monitoring equipment and monitoring operations should not exceed the original asset replacement costs; lest the whole point of Predictive Maintenance becomes moot. Studies have estimated that a properly functioning CBM program can provide savings of 8% to 12% over the traditional maintenance schemes. Our customer reports indicate the following industrial average savings resulted from initiation of a functional predictive maintenance program:
- Reduction in maintenance costs: 25% to 30%
- Elimination of breakdowns: 70% to 75%
- Reduction in equipment or process downtime: 35% to 45%
- Increase in production: 20% to 25%
Predictive analytics with M2M telematics provides deep insights into the machine operations and full functionality status – giving rise to optimal maintenance schedules with improved machine availability. The enormous data streamed in realtime from sensors attached to the machine/vehicle is processed with big data architectures to enable anomaly detection and failure prediction in real time.
The demo video below showcases the Predictive Maintenance in operation with an electric locomotive airbrake as M2M predictive analytics case study. It demonstrates computing the Failure rate and MTTF from realtime operating conditions, and how alternate scenarios can be evaluated for a what-if analysis that can benefit from big data architectures.
More details can be found in my M2M Telematics & Predictive Analytics paper.
Gopalakrishna Palem is a Technology Management & Strategy constultant specialized in Big data architectures and M2M predictive analytics. During his 12+ year tenure at Microsoft and Oracle, he helped many customers build their high volume transactional systems, distributed render pipelines, advanced visualization & modeling tools, real-time dataflow dependency-graph architectures, and Single-sign-on implementations for M2M telematics.
When he is not busy working, he is actively engaged in driving open-source efforts and guiding researchers on Algorithmic Information Theory, Systems Control and Automata, Poincare recurrences for finite-state machines, Knowledge modeling in data-dependent systems and Natural Language Processing.
He can be reached at [email protected]