Schedule a Software Walkthrough
Software Reliability Growth Models
Models that are overkill

Ohba Inflection S-Shaped Model

Intended to capture the case in which some faults mask others. An overly complex model is not recommended by IEEE 1633.

S-Shaped NHPP
S-curve driven by mutual dependency between faults, not testing ramp-up
1984
Introduced by Mitsuru Ohba
Masked-Fault Data
Best when some faults are known to be hidden behind others
Overview

An S-curve caused by faults hiding behind other faults

It assumes some faults are effectively invisible to testing until other faults are found and fixed first — a mutual dependency, or masking, effect that's common in complex systems with layered logic.

This model is overkill because you can simply use different models when inflections occur. The academic community drives for one model that fits all assumptions, when in fact you can simply use a selection of models.

This model is not one of the software reliability growth models recommended in the IEEE 1633. Requs Trend employs models that are recommended.

Model Family

S-Shaped, Inflection NHPP

Adds an inflection parameter to the Goel-Okumoto model to represent mutual dependency between faults.

Introduced

1984 — Mitsuru Ohba

Developed to address cases where purely exponential SRGMs proved insufficient for data showing fault masking effects.

IEEE 1633 Context

Standard S-Shaped Option

A recognized, standards-consistent alternative when the data shows a fault-masking-driven S-curve rather than a testing ramp-up.

01 — The Math

How the Inflection S-Shaped Model works

The model introduces an inflection parameter c that represents the degree of mutual dependency among faults — how much some faults are hidden behind others until those others are fixed.

§1Mean Value

Mean Value Function

Formula
m(t) = a (1 − e^(−bt)) / (1 + c e^(−bt))
Cumulative expected failures, shaped by the inflection parameter c

As c approaches zero, this reduces to the standard Goel-Okumoto exponential curve; larger values of c produce a more pronounced S-shape.

What the Terms Mean
  • m(t) — expected cumulative number of failures by time t
  • a — total expected number of faults (the asymptote)
  • b — the fault detection rate
  • c — the inflection (mutual dependency) parameter — higher values mean more fault masking
§2Intensity

Failure Intensity Function

Formula
λ(t) = ab(1+c) e^(−bt) / (1 + c e^(−bt)
Instantaneous failure intensity, derived from the mean value function

Like the Delayed S-Shaped model, intensity rises before it falls — but here the shape is driven by the inflection parameter c rather than a separate ramp-up term.

What the Terms Mean
  • λ(t) — instantaneous failure intensity at time t
  • a, b, c — same parameters as the mean value function
  • Larger c values push the peak of λ(t) later in the testing period
§3Assumptions

Key Assumptions

Model Assumptions

The core NHPP assumptions still apply, with an added assumption specifically about fault interdependency.

Assumption List
  • A finite total fault count a exists in the software
  • Some faults are mutually dependent — undetectable until other faults are found and removed first
  • The degree of this masking effect is constant and captured by a single inflection parameter c
  • Fault repair is instantaneous and perfect — no new faults introduced
§4Estimation

Parameter Estimation

Method

a, b, and c are estimated using maximum likelihood estimation or nonlinear least-squares regression. With three parameters instead of two, the model needs a data set large enough to distinguish the inflection effect from simple exponential decay.

What's Being Estimated
  • a — the eventual total fault count
  • b — the underlying fault detection rate
  • c — the strength of the fault-masking/inflection effect
02 — Strengths & Limitations

Strengths, limitations, and best fit for the Inflection S-Shaped Model

STRENGTHS

Models a real, distinct failure pattern

Captures fault masking directly, a mechanism the Delayed S-Shaped model doesn't represent, since its S-curve comes from a different cause (testing ramp-up, not fault dependency).

LIMITATIONS

Overly complex

Multiple models that handle different inflections is a better idea.

BEST USED FOR

Not recommended by IEEE 1633

Faults hide other faults when there is an increasing fault rate. Using different models is a better idea.

DATA REQUIREMENTS

Needs enough data for three parameters

As a three-parameter model, it requires a larger, cleaner data set than the two-parameter exponential models to fit reliably.

See this model fit against real project data.

Start with the online demo of Requs Trend or a discussion of your current reliability growth modeling approach.