Conditional predictive Bayesian Cramér-Rao Lower Bounds for prognostic algorithms design

Acuña, David E.; Orchard, Marcos E.; Saona, Raimundo

doi:10.1016/j.asoc.2018.01.033

Cited by 8 publications

(4 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [74,75], there was an attempt to provide a more general insight, but it was restricted to Markov processes. Also, there was an underlying hypothesis of statistical independence in the proposed probability measures; although these measures have still proven to be useful to define a functional cost criterion for prognostic algorithm design [76]. Moreover, despite the fact that hazard zones are well known and accepted in the PHM community [13], the current state-of-the-art formalization of failure prognosis problem [77,78] still defines failure events with the classical deterministic threshold approach, is restricted to events over Markov processes, lacks of mathematical demonstrations, and has led to inconsistencies when computing FPT probability distributions with methods different from those simulating complete state trajectories of systems [74].…”

Section: Failure Prognosis In the Discipline Of Prognostics And Health Managementmentioning

confidence: 99%

Computation of time probability distributions for the occurrence of uncertain future events

Acuña-Ureta

Orchard

Wheeler

2021

Mechanical Systems and Signal Processing

Self Cite

View full text Add to dashboard Cite

Section: Failure Prognosis In the Discipline Of Prognostics And Health Managementmentioning

confidence: 99%

Computation of time probability distributions for the occurrence of uncertain future events

Acuña-Ureta

Orchard

Wheeler

2021

Mechanical Systems and Signal Processing

Self Cite

View full text Add to dashboard Cite

“…Compared to predictive and forecast modeling methods [7] [13] and approaches with the usage of anomaly/fault prognostics and RUL (remaining useful life) estimation [14] [15] [16], which are able to predict quality downtrends through health indicators or other types of problems (failures, degradations, ...) in the future with a certain horizon, the essential point is that anomaly detection operates in a fully unsupervised manner (typically based on process data/models), which means that no quality information about the current process/system/product states needs to be available to establish appropriate predictors or forecast models. Therefore, 1.…”

Section: Motivation and State-of-the-artmentioning

confidence: 99%

On-line anomaly detection with advanced independent component analysis of multi-variate residual signals from causal relation networks

Lughofer

Zăvoianu

Pollak

et al. 2020

Information Sciences

View full text Add to dashboard Cite

Anomaly detection in today's industrial environments is an ambitious challenge in order to detect possible arising faults/problems, which may turn into severe waste during production, into defects or even into damages of systems components, at an early stage. Data-driven anomaly detection from multi-sensor networks faces challenges in proper data (information) fusion methodologies to establish adequate modeling cycles, whose outcome are models characterizing the anomaly-free reference situation based on which new on-line data are compared with, i.e. how much these deviate from them (pointing to potential faults). In this paper, we propose a new approach which is based on i) causal relation networks (CRNs) representing the inner causes and effects between sensor channels (or sensor nodes) in form of partial sub-relations, which are modeled by non-linear (fuzzy) regression models for characterizing the (local) degree of influences of the single causes on the effects, ii) an advanced analysis of the multi-variate residual signals obtained from the partial relations in the CRNs, which employs independent component analysis (ICA) to characterize hidden structures in the fused residuals (a significant change in these indicates an anomaly) and iii) automatized control limits on the energy content of latent variables obtained through the demixing matrix from ICA. Suppression of possible noise content in residuals ---to decrease the likelihood of false alarms ---, is achieved by performing ICA-based residual analysis solely on the dominant parts. Our approach was successfully evaluated for a real-world manufacturing process in the context of micro-fluidic chip production, where customer complaints arose about the quality of the chips during a specific production cycle $\rightarrow$ our approach could detect the anomaly in the process (leading to the bad quality chips) with negligible delay based on the process data recorded by multiple sensors in two production phases (injection molding and bonding), while it produced lower false alarm rates than several related and well-known state-of-the-art methods for (unsupervised) anomaly detection and while it also caused much lower parametrization efforts (in fact, none at all) to produce reliable results. Keywordson-line anomaly detection; causal relation networks; advanced multi-variate residual analysis; dominant parts of independent component analysis; automatized control limits; on-line production systems

show abstract

“…Early recognition of arising problems in form of downtrends in product quality or in form of cases where measures for product quality are expected to fall out of allowed tolerance limits in a latter/final stage of production. For the first task, usually techniques from the fields of forecasting [5] and prognostics [6], [7] are applied, which can either rely on process parameter settings (static case) or process values recorded over time (dynamic case). Supervision of process parameters were proposed in [8] (there for a phone camera lens injection molding machine) and in [9] (there for plastic injection moulding).…”

Section: Motivation and State-of-the-artmentioning

confidence: 99%

“…Our approach is leaned on the concepts demonstrated in [32] for dynamically adjusting the forgetting factor Λ in the adaptive learning and evolution process of the fuzzy models. This factor is integrated 1. in the recursive fuzzily weighted least squares formulas for the rule consequents, see Equations (7) to (9).…”

Section: Dynamic Forgettingmentioning

confidence: 99%

Autonomous supervision and optimization of product quality in a multi-stage manufacturing process based on self-adaptive prediction models

Lughofer

Zăvoianu

Pollak

et al. 2019

Journal of Process Control

View full text Add to dashboard Cite

T. 2019. Autonomous supervision and optimization of product quality in a multi-stage manufacturing process based on self-adaptive prediction models. Journal of process control [online], 76, pages 27-45. Available from: https://doi. AbstractIn modern manufacturing facilities, there are basically two essential phases for assuring high production quality with low (or even zero) defects and waste in order to save costs for companies. The first phase concerns the early recognition of potentially arising problems in product quality, the second phase concerns proper reactions upon the recognition of such problems. In this paper, we address a holistic approach for handling both issues consecutively within a predictive maintenance framework at an on-line production system. Thereby, we address multi-stage functionality based on i)data-driven forecast models for (measure-able) product quality criteria (QCs) at a latter stage, which are established and executed through process values (and their time series trends) recorded at an early stage of production (describing its progress), and ii) process optimization cycles whose outputs are suggestions for proper reactions at an earlier stage in the case of forecasted downtrends or exceeds of allowed boundaries in product quality. The data-driven forecast models are established through a high-dimensional batch time-series modeling problem. In this, we employ a non-linear version of PLSR (partial least squares regression) by coupling PLS with generalized Takagi-Sugeno fuzzy systems (termed as PLS-Fuzzy). The models are able to self-adapt over time based on recursive parameters adaptation and rule evolution functionalities. Two concepts for increased flexibility during model updates are proposed, i.) a dynamic outweighing strategy of older samples with an adaptive update of the forgetting factor (steering forgetting intensity) and ii.) an incremental update of the latent variable space spanned by the directions (loading vectors) achieved through PLS; the whole model update approach is termed as SAFM-IF (Self-Adaptive Forecast Models with Increased Flexibility). Process optimization is achieved through multi-objective optimization using evolutionary techniques, where the (trained and updated) forecast models serve as surrogate models to guide the optimization process to Pareto fronts (containing solution candidates) with high quality. A new influence analysis between process values and QCsis suggested based on the PLS-fuzzy forecast models in order to reduce the dimensionality of the optimization space and thus to guarantee high(er) quality of solutions within a reasonable amount of time (→ better usage in on-line mode). The methodologies have been comprehensively evaluated on real on-line process data from a (micro-fluidic)chip production system, where the early stage comprises the injection molding process and the latter stage the bonding process. The results show remarkable performance in terms of low prediction errors of the PLS-Fuzzy forecast models (showing mostly lower errors t...

show abstract

Conditional predictive Bayesian Cramér-Rao Lower Bounds for prognostic algorithms design

Cited by 8 publications

References 9 publications

Computation of time probability distributions for the occurrence of uncertain future events

Computation of time probability distributions for the occurrence of uncertain future events

On-line anomaly detection with advanced independent component analysis of multi-variate residual signals from causal relation networks

Autonomous supervision and optimization of product quality in a multi-stage manufacturing process based on self-adaptive prediction models

Contact Info

Product

Resources

About