Reliability is a new science: we are on the right way

TL;DR

This paper argues that reliability should be recognized as a scientific discipline by establishing its dual-structure of empirical and mathematical truths, supported by axiomatic principles and experimental verification.

Contribution

It introduces an axiomatic system for reliability science, develops a mathematical framework, and demonstrates its applicability across various systems and lifecycle phases.

Findings

Reliability is characterized as a scientific discipline with dual-structure.

Reliability principles include margin, degradation, and uncertainty.

Biandong Statistics measures dynamic uncertainty.

Abstract

Reliability has long been treated as an engineering practice supported by testing, statistics and standards, yet its status as a scientific discipline remains unsettled. From a philosophical perspective, scientific truth is characterized by a dual-structure that links empirical truth and mathematical truth, which requires an axiomatic system that is symbolically expressible and verifiable by universally repeatable controlled experiments. Building on this criterion, this paper examines whether reliability satisfies the dual-structure of scientific truth. Firstly, we analyze the philosophical foundations of the reliability problem, tracing its transition from experiential confidence and engineering practice toward scientific inquiry. Then, reliability science principles are introduced as an axiomatic system consisting of margin, degradation and uncertainty, which define reliability as the repeatability of system performance across time and space. Next, we present reliability science experiments as the empirical aspect of the dual-structure, where controlled and repeatable interventions are designed to verify the causal relations implied by the axioms. Furthermore, we develop the mathematical framework of reliability as the symbolic aspect of the dual-structure, articulating reliability laws through distance, relation and change, and developing a time-dependent measure, Biandong Statistics, to represent varying uncertainty beyond static descriptions. Accordingly, we argue that reliability is indeed a scientific discipline. The applicability of reliability science is demonstrated across engineering, living and social systems, and a unified logic for guiding engineering activities across the entire product lifecycle is provided, linking reliability to the conceptual, development, procurement, production and operation phases within a model-based structure.