Switching Conditional Random Field Approach to Process Operating Mode Diagnosis for Multi-Modal Processes

Accurate diagnosis of process modes for industrial processes is critical to safe and reliable operation of processes. The hidden Markov models (HMMs) have been widely employed to solve the real-time process mode diagnosis problems for multi-modal processes. However, restricted by the inherent conditional independence assumptions, the process mode diagnosis performance of HMMs tends to be degraded as these assumptions can be easily broken in reality. Alternatively, the conditional random field (CRF) model has been proposed in the context of process monitoring and proven to outperform the HMMs. In this work, we extend the CRF framework to the mode diagnosis of the processes that have multiple operating conditions, by designing a new framework, termed as, a switching CRF (SCRF). In the proposed framework, multiple linear-chain CRF models are proposed which have the capability to switch between each other in accordance with a scheduling variable that is indicative of the operating conditions. The expectation-maximization algorithm is employed for parameter estimation. To validate the performance, a numerical example is employed. The results demonstrate that the proposed SCRF approach shows superior diagnosis performance to the linear-chain CRF based method.

[1]  Andrew McCallum,et al.  An Introduction to Conditional Random Fields , 2010, Found. Trends Mach. Learn..

[2]  Si-Zhao Joe Qin,et al.  Monitoring Non-normal Data with Principal Component Analysis and Adaptive Density Estimation , 2007, 2007 46th IEEE Conference on Decision and Control.

[3]  Jeff A. Bilmes,et al.  A gentle tutorial of the em algorithm and its application to parameter estimation for Gaussian mixture and hidden Markov models , 1998 .

[4]  Lawrence R. Rabiner,et al.  A tutorial on hidden Markov models and selected applications in speech recognition , 1989, Proc. IEEE.

[5]  B. Kowalski,et al.  Partial least-squares regression: a tutorial , 1986 .

[6]  Andrew McCallum,et al.  Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data , 2001, ICML.

[7]  Dimitar Filev,et al.  Online Nonlinear System Identification with Evolving Spatial-Temporal Filters , 2018, 2018 Annual American Control Conference (ACC).

[8]  Paul Geladi,et al.  Principal Component Analysis , 1987, Comprehensive Chemometrics.

[9]  Jorge Nocedal,et al.  On the limited memory BFGS method for large scale optimization , 1989, Math. Program..

[10]  Henry A. Kautz,et al.  Extracting Places and Activities from GPS Traces Using Hierarchical Conditional Random Fields , 2007, Int. J. Robotics Res..

[11]  Biao Huang,et al.  A novel approach to process operating mode diagnosis using conditional random fields in the presence of missing data , 2018, Comput. Chem. Eng..

[12]  D. Rubin,et al.  Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[13]  Xunyuan Yin,et al.  Distributed State Estimation of Sensor-Network Systems Subject to Markovian Channel Switching With Application to a Chemical Process , 2018, IEEE Transactions on Systems, Man, and Cybernetics: Systems.

[14]  S. Zhao,et al.  Monitoring of Processes with Multiple Operating Modes through Multiple Principle Component Analysis Models , 2004 .

[15]  Richard D. Braatz,et al.  Fault Detection and Diagnosis in Industrial Systems , 2001 .

[16]  Georg Heigold,et al.  Equivalence of Generative and Log-Linear Models , 2011, IEEE Transactions on Audio, Speech, and Language Processing.

[17]  Xunyuan Yin,et al.  Distributed output‐feedback fault detection and isolation of cascade process networks , 2017 .

[18]  Si-Zhao Joe Qin,et al.  Survey on data-driven industrial process monitoring and diagnosis , 2012, Annu. Rev. Control..

[19]  Alireza Fatehi,et al.  Operating condition diagnosis based on HMM with adaptive transition probabilities in presence of missing observations , 2015 .