Abstract Multivariate statistical process monitoring (MSPM) methods are significant for improving production efficiency and enhancing safety. However, to the authors’ best knowledge, there is no survey paper providing statistics of published papers over past decade. In this paper, several issues related MSPM reviewed studied. First, annual publication numbers journal articles concerning provided show active development important research field point out promising directions in future....

Abstract In this study, a two‐step principal component analysis (TS‐PCA) is proposed to handle the dynamic characteristics of chemical industrial processes in both steady state and unsteady state. Differently from traditional PCA (DPCA) dealing with static cross‐correlation structure auto‐correlation process data simultaneously, TS‐PCA handles them two steps: it first identifies by using least squares algorithm, then monitors innovation PCA. The time uncorrelated independent initial process....

Nonlinearity is extremely common in industrial processes. For handling the nonlinearity problem, this paper combines artificial neural networks (ANN) with principal component analysis (PCA) and proposes a new (NCA). NCA has similar network structure as ANN adopts gradient descent method for training, hence it same nonlinear fitting ability ANN. Furthermore, PCA's dimension reduction strategy to extract uncorrelated components from process data constructs statistical indices monitoring. The...

Several studies have applied the hidden Markov model (HMM) in multimode process monitoring. However, because inherent duration probability density of HMM is exponential, which inappropriate for modeling process, performance these HMM-based approaches not satisfactory. As a result, semi-Markov (HSMM), integrated mode into HMM, combined with principal component analysis (PCA) to handle feature, named as HSMM-PCA. PCA powerful monitoring algorithm unimodal and HSMM specializes division...

Outliers may cause model deviation and then affect the monitoring performance hence it is a challenging problem for process monitoring. The robust principal component analysis (RPCA) approach, which describes outlier components with sparse matrix identifies these using recovery most commonly used method to solve problems caused by outliers. However, because existing mathematical tools can only obtain nonsparse small element values, RPCA performs poorly during In this paper, we propose novel...

This paper introduces a novel monitoring method related to key-performance-indicators (KPIs), specifically tailored for the hybrid electric vehicle (HEV) powertrain system. The proposed establishes new KPI that better reflects performance of HEV Through application partial least squares and contribution plot method, it excels in minimizing data scale precisely faults. Diverging from current methodologies, this demands minimal prior knowledge solely relies on previously observed data....

Nonlinearity may cause a model deviation problem, and hence, it is challenging problem for process monitoring. To handle this issue, local kernel principal component analysis was proposed, achieved satisfactory performance in static For dynamic process, the expectation value of each variable changes over time, cannot be replaced with constant value. As such, data structure wrong, which causes problem. In paper, we propose new two-step analysis, extracts components then analyzes them by...

Abstract Recursive statistical process monitoring (RSPM) methods have superior performance for industrial processes, especially those with time‐varying characteristics, and recently been studied by many researchers. However, there is no survey paper yet that summarizes analyzes the existing RSPM methods. In this survey, approximately 60 papers related to are reviewed categorized from different aspects. Existing using classification criteria proposed in study based on ways of recursively...

To overcome the shortage of traditional temperature sensors, this paper adopts infrared thermal imaging technology for measurement. avoid spatial information loss issue during image data vectorization process, adopted relationship between pixels in principal component analysis (PCA) model training, which is called information-based PCA (SIPCA). Then, also used fault localization method to enhance location performance. Tested by an experimental tank system, proposed achieves better...

Abstract Traditional multivariate statistics‐based process monitoring (MSPM) methods are static algorithms, and the “time lag shift” method (TLSM) is most commonly used approach to handle dynamic issue. This paper proves in theory that two drawbacks exist TLSM‐based approaches: information unrelated real‐time data also analyzed, can be predicted by historical counted repeatedly both data. adopts orthonormal subspace analysis (OSA) these issues. OSA successfully separate into (the component)...

In order to address the non-Gaussian and dynamic features in process, this paper combined two- step principal component analysis (TS-PCA) preliminary-summation-based PCA (PS-PCA). By using summation two-step strategy, PS-TS-PCA can eliminate influence of relationship among process data feature at same time. And hence achieved good performance both monitoring. addition, also proposed new strategy correlation between two sample, proved that least squares algorithm estimate structure process.

Orthonormal subspace analysis (OSA) is proposed for handling the decomposition issue and principal component selection in traditional key performance indicator (KPI)-related process monitoring methods such as partial least squares (PLS) canonical correlation (CCA). However, it not appropriate to apply static OSA algorithm a dynamic since pays no attention auto-correlation relationships variables. Therefore, novel (DOSA) capture auto-correlative behavior of variables on basis KPIs accurately....

Preliminary-summation-based PCA (PS-PCA), was recently proposed to handle the non-Gaussian features of industrial processes. However, when PS-PCA is applied monitoring data with outliers, "summation infection" phenomenon occurs, which makes ineffective. To eliminate influence this paper proposes a novel robust (RPS-PCA) distinguishes outliers from faulty using consecutive detection results and then removes them. Because RPS-PCA only eliminates in normal process while retaining data, it can...

A enhanced principal component analysis (PCA), termed as Two-step PCA (TS-PCA), is proposed to handle the dynamic characteristic of industry processes. Differently from traditional (DPCA) using "time lag shift" structure, TS-PCA adopts a new structure present property in process data. By this can extract time-uncorrelated components data and use it for monitoring. In addition, update expectation standard variance at each step normalization.

Partial least squares (PLS) is a widely used multivariate statistical technique, which can be in neuroimaging, process monitoring, economics, etc.. Because the standard PLS trained by nonlinear iterative partial (NIPALS) algorithm, only obtain numerical solution rather than analytical solution. Therefore, it hard for subsequent theoretical analysis and optimization computationally intensive. This paper proposes an to take place of NIPALS. determination meaningfully contributes future PLS. In...

Abstract Traditional multivariate statistical-based process monitoring (MSPM) methods are effective data-driven approaches for large-scale industrial processes, but have a shortcoming in handling the redundant correlations between variables. To address this shortcoming, study proposes new MSPM method called minimalist module analysis (MMA). MMA divides data into several different modules and one more independent module. All variables strongly correlated, no exist; therefore, extracted...