Discussion of signature‐based models of preventive maintenance

We congratulate Asadi, Hashemi, and Balakrishnan for a comprehensive review of age-based maintenance strategies thorough study the statistic “signature” multi-component coherent systems. Among many contributions, Balakrishnan's paper also provides taxonomic framework with which to categorize classify plethora models. Section 2 is great importance, providing new researchers reliable nomenclature. What missing in this section, however, detailed explanation notation “coherent system,” will help clarify type systems signature applicable. Case article covers wide range scenarios, two exemplar are representative. However, case serves more as tutorial rather than proof superiority. The tables figures demonstrating (uniqueness of) optimization results, but not justifying advantages using signature. It would be much practical value compare relative performance models without As claimed article, one “important advantage other related concepts writing corresponding cost criteria that can take into account failure individual components costs”, we have couple comments about this. challenging extend application techniques maintenance, such condition-based predictive maintenance. Condition-based generally utilizes multi-view data (sensor monitoring data, manual inspection lifetime data), hence commonly accepted superior (that only data).1, Likewise, policies utilize streaming infer remaining useful lifetime.3 it difficult explicitly express an evaluation criterion (cost or availability) function policies, especially complex structures. curious applicability its data-rich settings. think there mainly technical hurdles surmounted before notion receives universal acceptance: assumption computation. While natural assume that, system, same distribution, untenable system has gone through multiple rounds (either perfect imperfect components). For “old” each component different (residual) vector s $$ \boldsymbol{s} no longer topological invariant (i.e., i ≠ A / n ! {s}_i\ne \left|{A}_i\right|/n! ). Consequently, greatly limit (to only). maintaining ubiquitous practice. example, need repeatedly update plan new) policy highly acquired information since operation. Other assumptions make brittle unlikely adopted general practice listed concluding remarks. Our viewpoint on should positioned formulation strategy competing existing objective formulations, auxiliary tool explaining component-level details engineers stakeholders. That is, what might call pragmatic signature-based formulation. tailored configurations decomposition further insights our decision making. interpret two-formulation paradigm context multi-fidelity optimization, where low-fidelity surrogate.4 In paradigm, restrictive signature, made mathematical convenience needs, now acceptable, could even extended strategies. group all components' distributions few clusters using, functional clustering.5 Structural utilized perform semi-supervised clustering.6 Each cluster corresponds type, then calculate survival Note do provide at decision-making point, interpretation, This partially overcome second hurdle—the Although briefly mentioned like comment computational complexity objectives. run-time includes calculation vector. apparent model computationally expensive, clear exactly time is. hope read future works quantify whole procedure big O notation, facilitate comparison selection among above way reduce burden: formulate when interpretation required. Another line approach replace exact (through permutation) Monte Carlo simulation. sampling presented Behrensdorf et al.7 constitutes early attempt numerically approximate focus evaluating efficiency algorithms convergence rate. On hand, efforts create signature-vector e-handbook ready reference standard well-regulated Finally, result from used warm initialization optimizing formulation, vice versa. To conclude discussion, argue built warrants investigation. believe powerful interpreting understanding polices generalized Moreover, exploratory analysis high-fidelity model, rough estimate importance optimal policy. suggested research provided guidelines tackling combinatorial problem. look forward reading materializing refuting ideas. Open access funding by IReL.