OptimalTREND™ Wiki / The DIPF curve
Technical
The DIPF curve
The classic P-F curve, from Nowlan and Heap's 1978 Reliability-centered Maintenance, describes only the detection window before failure. The reliability community has since expanded it to the DIPF curve – Design, Installation, Potential failure, Functional failure – and, with safety added, the SDIPF curve. It plots resistance to failure across the whole asset life and, more usefully, shows where the failures are really introduced.
From P-F to DIPF
Nowlan and Heap's P-F curve made one point: inspect an item to detect a potential failure, an identifiable physical condition that shows functional failure is imminent, so you can intervene before it fails. The DIPF curve extends that backward to Design and Installation, the two stages where most failures are actually introduced, and the SDIPF curve extends it again to put Safety first. The classic P-F curve turns out to be only the last segment of a much longer story.
The four points
| D – Design | Inherent reliability is specified and designed in: materials, duty and margins, redundancy, maintainability and design-for-inspection. Around 80% of an asset's lifecycle cost is fixed here. |
|---|---|
| I – Installation | Installation and commissioning preserve or destroy that design reliability. Misalignment, imbalance, soft foot, contamination, wrong lubrication and poor commissioning cut resistance to failure from day one. |
| P – Potential failure | The point at which the developing defect first becomes detectable by a condition-monitoring technology. Where P sits depends on the technique's sensitivity. |
| F – Functional failure | The point at which the asset can no longer meet its required standard. Beyond it lies the fault domain: heat, noise, secondary damage and safety exposure. |
The four domains
The curve divides the life into four domains, and each calls for different work:
| Design | Reliability is specified and designed in. Around 80% of lifecycle cost is fixed in plan, design and build, so this is the highest-leverage domain of all: right sizing, right specification, design for inspection and maintainability. |
|---|---|
| Proactive | Installation, commissioning and precision practice: precision alignment, balancing, torqueing and lubrication, contamination control and defect elimination. Resistance to failure is preserved here, and the largest, cheapest recurring gains are made. |
| Predictive | P to F. Condition monitoring detects the developing defect and you plan the intervention before failure. This is OptimalTREND™'s home – the classic P-F interval. |
| Fault | Past F. The asset is failing, cost multiplies and safety is exposed. The whole point of the other three domains is to never operate here. |
Diagnosis, not prediction
An important distinction the reliability leaders draw: condition monitoring diagnoses, it does not predict. Detecting an identifiable physical condition that shows failure is imminent is a diagnosis of a defect that already exists, like a doctor's test for a virus – testing, not forecasting. Statistical prediction, "this bearing will fail within two years", is fine for a population and useless for one asset; to predict an individual failure you must already have the defect. OptimalTREND™ detects and diagnoses the real, developing condition and estimates the time left from its own trend, not a population statistic.
Where the leverage really is
The DIPF curve reframes where to spend effort. Figures widely cited in the reliability literature: about 80% of lifecycle cost is fixed in plan, design and build; as much as 60% of failures and safety issues are preventable by design; and 30 to 40% of breakdowns trace to poor design or condition. Most corrective work found by predictive maintenance could have been eliminated upstream – by precision installation, alignment, balancing, torqueing and lubrication, better specification and better sourcing. Defect elimination in the design and proactive domains removes the defect before it ever reaches P; condition monitoring in the predictive domain is the safety net, not the strategy.
Detection point by technology (the predictive domain)
Within the predictive domain, different technologies detect the same developing defect at different points, so the technology sets the warning you get. Typical order and lead time for a rolling-element bearing:
| Acoustic emission / ultrasound | Earliest. Incipient sub-surface and lubrication distress. Months. |
|---|---|
| Vibration, high-frequency envelope | Early. Repetitive impacts before the velocity spectrum. Weeks to months. |
| Vibration, velocity spectrum | Mid. Defect frequencies and harmonics established. Weeks. |
| Oil and wear debris | Mid. Metal from the wearing surfaces. Days to weeks. |
| Temperature / thermography | Late. Friction heat once advanced. Days. |
| Audible noise, hot to touch | Very late. Hours to days; effectively at F. |
Net P-F and the monitoring interval
The usable window is the net P-F interval: the P-F interval minus the time to detect, plan and act. Two rules follow, both examinable CMRP material:
- The monitoring interval must be no more than half the P-F interval, so at least one inspection reliably falls inside the window. A bearing with a two-month P-F interval needs monitoring at least monthly.
- The net P-F interval must be long enough to act: to raise the work, get the parts and take the outage. If it is not, the task detects a failure you still cannot prevent.
Why age-based maintenance still fails
The Nowlan and Heap study behind RCM found six failure patterns, and only a minority show a wear-out age:
| Pattern F, infant mortality – about 68% | High early, then constant. The majority of items. |
|---|---|
| Pattern E, random – about 14% | Constant failure rate, independent of age. |
| Pattern D, rapidly rising then constant – about 7% | No useful wear-out life. |
| Pattern C, gradually rising – about 5% | Slow rise, no clear knee. |
| Pattern B, wear-out – about 2% | Classic end-of-life. |
| Pattern A, bathtub – about 4% | Infant mortality plus wear-out. |
About 89% of items (patterns D, E and F) have no useful wear-out age, so scheduled overhaul cannot improve them and often introduces infant mortality – itself an Installation-domain defect. What works is precision in the design and proactive domains, and detecting the rest on condition within the predictive domain.
Where OptimalTREND™ fits
OptimalTREND™ works the predictive domain: it diagnoses the developing defect as early as possible, moving P to the left and lengthening the interval, monitors continuously so nothing is missed, and estimates remaining useful life from the real trend. The design and proactive domains – precision design, installation and RCM so the defect never forms – are the work of OptimalAvailability Studio™ and precision practice across the Optimal ARaaS® platform. Together they lift the whole DIPF curve, not just its predictive tail.