High-dimensional Monitoring Research Articles

Time Series Feature Selection Method Based on Mutual Information

Time series data have characteristics such as high dimensionality, excessive noise, data imbalance, etc. In the data preprocessing process, feature selection plays an important role in the quantitative analysis of multidimensional time series data. Aiming at the problem of feature selection of multidimensional time series data, a feature selection method for time series based on mutual information (MI) is proposed. One of the difficulties of traditional MI methods is in searching for a suitable target variable. To address this issue, the main innovation of this paper is the hybridization of principal component analysis (PCA) and kernel regression (KR) methods based on MI. Firstly, based on historical operational data, quantifiable system operability is constructed using PCA and KR. The next step is to use the constructed system operability as the target variable for MI analysis to extract the most useful features for the system data analysis. In order to verify the effectiveness of the method, an experiment is conducted on the CMAPSS engine dataset, and the effectiveness of condition recognition is tested based on the extracted features. The results indicate that the proposed method can effectively achieve feature extraction of high-dimensional monitoring data.

Dynamic comprehensive quality assessment of post-harvest grape in different transportation chains using SAHP–CatBoost machine learning

Abstract Objectives Food quality assessment is critical for indicating the shelf-life and ensuring food safety or value. Due to high environmental sensitivity, the post-harvest quality of fresh fruit will undergo complex changes in the supply chain, with various dynamic quality-related features. It is difficult to efficiently and accurately extract comprehensive quality feature of post-harvest fruits from high-dimensional monitoring data with heterogeneous characteristics (numerical and categorical). Therefore, we proposed a dynamic comprehensive quality assessment method based on self-adaptive analytic hierarchy process (SAHP) integrated with the CatBoost model. Materials and Methods By adaptive weight optimization, the SAHP was utilized to analyze the multi-source quality information and obtain the quantized fusion value, as an output sample of CatBoost machine learning. Then, using heterogeneous monitoring data as input, the CatBoost model was directly trained through unbiased boosting with categorical features for dynamic assessment of overall quality status. Results Three quality index monitoring data sets for ‘Jufeng’ grape in different transportation chains (normal temperature, cold insulation, and cold chain) were individually constructed as the research samples. Furthermore, compared to other machine learning methods, the SAHP–CatBoost had more accurate results in comprehensive quality feature extraction. In actual transportation chains, the mean absolute error, mean absolute percentage error, and root mean squared error of dynamic comprehensive assessment were limited to 0.0044, 1.012%, and 0.0078, respectively. Conclusions The proposed method is efficient in handling heterogeneous monitoring data and extracting comprehensive quality information of post-harvest grape as a robust shelf-life indicator. It can reasonably guide post-harvest quality management to reduce food loss and improve economic benefits.

Health assessment of water pumps using high-dimensional monitoring data

Abstract With the development of IoT monitoring equipment, an increasing number of monitoring indicators are employed to monitor the operational status of water pumps, thereby resulting in the challenge of data redundancy. This paper proposes an algorithm for predicting the health status of pumps that integrates multiple monitoring variables. Initially, the original dataset is classified using the maximum relevance minimum redundancy method. Next, principal component dimensionality reduction is used to reduce the dimensionality of the classified dataset. Finally, a long and short term memory neural network is employed to construct the association model between monitoring data and equipment health. The proposed algorithm takes into account the correlation between variables and the negative impacts of long-term dependence on the prediction results. It is capable of predicting abnormal working conditions, which has been experimentally verified in the Xiasha Pumping Station located in Hangzhou. The algorithm was compared with LR, SVM, and RNN algorithms, and it was found that the proposed algorithm achieved the highest prediction accuracy.

Distribution-free Phase-II monitoring of high-dimensional industrial processes via origin and modified interpoint distance based algorithms

Improvements in measuring devices, the development of sensing technologies, automated record-keeping and cloud storage facilities have offered us a large volume of data streams involving many variables for data-driven decision-making. Consequently, monitoring high-dimensional (HD) processes has become an essential and exciting topic in statistical process monitoring. HD monitoring poses a significant challenge because considerable uncertainty is involved in estimating process parameters when an adequately large reference sample is unavailable before monitoring. Thus, designing an effective, conveniently implementable monitoring scheme for the HD processes remains a challenging task that has yet to be effectively addressed. This paper develops two distribution-free exponentially weighted moving average (EWMA) schemes to monitor multivariate and HD processes based on origin and interpoint distances. Previous works on interpoint distance-based charts are generally not distribution-free unless only one reference point is used as a benchmark to compute the interpoint distances. We offer an improved and general way to achieve conditionally distribution-free properties closely. The proposal also combines the advantage of the nonparametric approaches for jointly monitoring the process location and scale parameters with the better shift detection performance of an EWMA scheme. The new EWMA schemes can be easily implemented even when the data dimension is large with a small number of Phase-I samples. Control limits using the median run length as a metric are evaluated and tabulated for some cases to implement the proposed chart quickly. Extensive comparative studies have been conducted to verify the efficacy of the proposed monitoring schemes. Finally, the proposed schemes are illustrated with a real problem of monitoring high-dimensional quality characteristics from semiconductor manufacturing processes.

Optimized Apriori algorithm for deformation response analysis of landslide hazards

With the explosive development of data mining technology, higher requirements have been put forward to analyze the response of landslide deformation. However, related algorithms, such as the traditional Apriori and FP-Growth algorithms, are still in the starting period of being applied to landslide hazards. Due to monitoring data characteristics, some problems were encountered while applying these algorithms, such as poor applicability and low computational efficiency. Therefore, we propose an optimized Apriori algorithm to solve the above problems. The optimized algorithm strictly controls the front and rear itemsets' construction process and stores the factors and deformation events according to their dimensions and levels. In addition, some key calculation processes of the algorithm are well-parallelized. Based on the monitoring data of the Baishuihe landslide, three experiments were designed to verify the performance of the proposed algorithm. The results show that when the strong association rules with high factor dimension and level characteristics were obtained, the proposed algorithm's computation time was 1/432 and 1/80 of the Apriori and FP-Growth algorithms, respectively. The proposed algorithm has significant advantages in analyzing massive high-dimensional monitoring data of landslide hazards.

Self-supervised anomaly pattern detection for large scale industrial data

Detecting the anomalies in a large amounts of high-dimensional data has been a challenging task. In the Industry 4.0 environment, large-scale high-dimensional monitoring data features the complex pattern of high level semantics. In order to provide enterprise-wide monitoring solutions, it is necessary to identify the high-level semantic patterns of the anomalies in these data without splitting them. Existing end-to-end deep neural networks for time series are capable of recognizing the high-level semantics in natural language or speech signals, but they are barely applied in real-time anomaly detection of industrial data because of the large time costs. In this paper, we leverage the self-supervised contrastive learning methodology and propose a Composite Semantic Augmentation Encoder (CSAE) to provide an appropriate representation of industrial data and implement quick detection of anomalies in industrial application environments. CSAE is a non-sequential deep neural network with two augmentation layers and a mandatory layer. The two layers of data-augmentation are built to expand the size of samples of both low-level semantic anomalies and high-level semantic anomalies, which enables CSAE to discover diverse anomalies and improves its accuracy of high-level semantic pattern recognition. The mandatory layer is built to compress and reserve the temporal information in the industrial data to accelerate the anomaly detection. Therefore, as a non-sequential contrastive learning model, CSAE has faster training convergence than the usual sequence models. The experiment results have verified that CSAE can achieve higher prediction accuracy with less time consumption than existing machine learning models in the tasks of high dimensional anomaly pattern detection.

Aero-engine gas path system health assessment based on depth digital twin

Aero-engine health assessment is of great significance for accurately understanding the health status of aircraft, supporting maintenance decision-making and ensuring flight safety. However, aero-engine has the characteristics of complex structure, fault coupling and state nonlinearity, coupled with the constraints of many factors such as acquisition means, analysis methods and the limitation of abnormal data. It is difficult to obtain a mapping relationship that fully characterizes its operating status through monitoring information. Therefore, this paper proposes a health assessment method based on depth digital twin, which can be used for real-time monitoring of aero-engine operation state. Firstly, the mechanism model is constructed for the multi-scale simulation of aero-engine gas path system. Combined with the advantages of dynamic learning and self-optimization of deep learning method, the data-driven model for data prediction is constructed, and the two are fused to realize the depth digital twin of aero-engine. Then, the digital twin model is used to simulate the high-dimensional monitoring data generated during the operation of aero-engine. Finally, a multi-scale one-dimensional convolution neural network model (MultiScale1DCNN) is proposed to analyze the simulated data, so as to assess the real-time health status of aero-engine. Through the simulation test of aero-engine sensor data, it is verified that the digital twin model has high reliability. Compared with the traditional simulation model, it has higher accuracy. In the aero-engine health assessment tests, the MultiScale1DCNN model can accurately identify the failure mode and assess the failure level, and has high assessment accuracy. In several assessment tests, the assessment accuracy rate is above 96%. The test results show that the health assessment method can accurately reflect the health status of aero-engine, and has certain real-time performance, which shows that it has high engineering application value.

An early warning method for mechanical fault detection based on adversarial auto-encoders

The vibration signal of mechanical equipment is non-linear and non-stationary, and it is difficult to fully reflect the operation state of equipment through traditional fixed threshold alarm method. While the early warning method based on multi-feature parameter fusion relies on manual experience to extract features, which is difficult to ensure the accuracy of the extracted features and cannot achieved good early warning effect. To solve this problem, a feature self-learning method based on adversarial auto-encoders is proposed in this paper, which encodes high-dimensional monitoring data in normal state into low-dimensional vectors with certain statistical laws and uses it as a benchmark to detect abnormalities in the operating state of the equipment in time by measuring the difference between the encoded features of real-time monitoring data and the benchmark. The actual application cases of reciprocating compressors show that the proposed method can detect the weak signs of equipment fault at the early stage, and realize early warning. At the same time, by comparing with the Auto-Encoders network-based warning method and the Dirichlet process mixture model-based warning method, It is verified that the method in this paper has more advantages in terms of warning accuracy and warning time.

Multivariate control charts for monitoring process mean vector of individual observations under regularized covariance estimation

ABSTRACT Multivariate control charts are generally used in industries for monitoring and diagnosing processes characterized by several process variables. The applications of charts assume that the in-control process parameters are known and the charts’ limits are obtained from the known parameters. The parameters are typically unknown in practice, and the charts’ limits are usually based on estimated parameters from some historical in-control datasets in the Phase I study. The performance of the charts for monitoring future observation depends on efficient estimates of the process parameters from the historical in-control process. When only a few historical observations are available, the performance of the charts based on the empirical estimates of the process mean vector and covariance matrix have been shown to deviate from the desired performance of the charts based on the true parameters. We investigate the performance of the multivariate Shewhart control charts based on several shrinkage estimates of the covariance matrix when only a few in-control observations are available to estimate the parameters. Simulation results show that the control charts based on the shrinkage estimators outperform the charts based on existing classical estimators. An example involving high-dimensional monitoring is provided to illustrate the performance of the proposed Shrinkage-based Shewhart chart.

Multivariate Monitoring Workflow for Formulation, Fill and Finish Processes.

Process monitoring is a critical task in ensuring the consistent quality of the final drug product in biopharmaceutical formulation, fill, and finish (FFF) processes. Data generated during FFF monitoring includes multiple time series and high-dimensional data, which is typically investigated in a limited way and rarely examined with multivariate data analysis (MVDA) tools to optimally distinguish between normal and abnormal observations. Data alignment, data cleaning and correct feature extraction of time series of various FFF sources are resource-intensive tasks, but nonetheless they are crucial for further data analysis. Furthermore, most commercial statistical software programs offer only nonrobust MVDA, rendering the identification of multivariate outliers error-prone. To solve this issue, we aimed to develop a novel, automated, multivariate process monitoring workflow for FFF processes, which is able to robustly identify root causes in process-relevant FFF features. We demonstrate the successful implementation of algorithms capable of data alignment and cleaning of time-series data from various FFF data sources, followed by the interconnection of the time-series data with process-relevant phase settings, thus enabling the seamless extraction of process-relevant features. This workflow allows the introduction of efficient, high-dimensional monitoring in FFF for a daily work-routine as well as for continued process verification (CPV).

Spatial-Temporal Features Based Sensor Network Partition in Dam Safety Monitoring System.

Many various types of sensors have been installed to monitor the deformation and stress in the dam structure. It is difficult to directly evaluate the operation status of the dam structure based on the massive monitoring data. The sensor network is divided into multiple regions according to the design specifications, simulation data, and engineering experiences. The local results from sub-regions are integrated to achieve overall evaluation. However, it ignores the spatial distribution of sensors and the variation of time series, which cannot meet the real-time evaluation for the dam safety monitoring. If the network partitions can provide the preliminary foundation for analyzing the dynamic change laws of the dam’s working conditions in a real-way, we should consider the similarity of structure and stresses in the local region of the dam and the correlation among the monitoring data. A time-series denoising autoencoder (TSDA) is proposed to represent the spatial and temporal features of the nodes by compressing high-dimensional monitoring data. Then, a network partitioning algorithm (NPA) based on spatial-temporal features based on the TSDA is presented. The NPA ensures that the partition results can support the analysis of the physical change laws by introducing the auxiliary objective variable to optimize the network partition objective function. Experimental results on the public datasets and a real dataset from an arch dam demonstrate that the proposed network partition algorithm NPA can achieve better partition performance than TSDA+K-Means and TSDA+GMM. The NPA can improve the silhouette coefficient by 45.1% and 58.4% higher than the TSDA+K-Means and TSDA+GMM, respectively. The NPA can increase the Calinski-Harabaz Index by 30.8% and 61.6%, respectively.

Spatial rank-based high-dimensional monitoring through random projection

High-dimensional process monitoring becomes increasingly important in many applications where the joint distribution of process variables is usually unknown and not normal, and requires nonparametric methods for analysis and monitoring. However, when the process dimension is much larger than the reference sample size, most traditional nonparametric multivariate control charts fail due to the curse of dimensionality. Furthermore, when the process goes out of control, only a few (sparse) dimensions will be influenced, which increases the difficulty for both detection and diagnosis. To address these problems, this article proposes a new nonparametric monitoring scheme for high-dimensional processes. This scheme first projects the high-dimensional process into several subprocesses using ensemble random projections for dimension reduction. Then, for each subprocess a local nonparametric control chart is constructed based on the spatial rank test. Finally, all the local charts are fused together for decision making. Furthermore, after an out-of-control alarm is triggered, a diagnostic framework is proposed based on the square-root LASSO algorithm. Numerical studies together with real-data examples demonstrate the efficacy and applicability of the proposed methodology.

A data-driven approach of performance evaluation for cache server groups in content delivery network

In industry, Content Delivery Network (CDN) service providers are increasingly using data-driven mechanisms to build the performance models of the service-providing systems. Building a model to accurately describe the performance of the existing infrastructure is very crucial to make resource management decisions. Conventional approaches that use hand-tuned parameters or linear models have their drawbacks. Recently, data-driven paradigm has been shown to greatly outperform traditional methods in modeling complex systems. We design a data-driven approach to building a reasonable and feasible performance model for CDN cache server groups. We use deep LSTM auto-encoder to capture the temporal structures from the high-dimensional monitoring data, and use a deep neural network to predict the reach rate which is a client QoS measurement from the CDN service providers’ perspective. The experimental results have shown that our model is able to outperform state-of-the-art models.

Anomaly Detection Approaches for Semiconductor Manufacturing

Smart production monitoring is a crucial activity in advanced manufacturing for quality, control and maintenance purposes. Advanced Monitoring Systems aim to detect anomalies and trends; anomalies are data patterns that have different data characteristics from normal instances, while trends are tendencies of production to move in a particular direction over time. In this work, we compare state-of-the-art ML approaches (ABOD, LOF, onlinePCA and osPCA) to detect outliers and events in high-dimensional monitoring problems. The compared anomaly detection strategies have been tested on a real industrial dataset related to a Semiconductor Manufacturing Etching process.

Statistical Learning Methods Applied to Process Monitoring: An Overview and Perspective

The increasing availability of high-volume, high-velocity data sets, often containing variables of different data types, brings an increasing need for monitoring tools that are designed to handle these big data sets. While the research on multivariate statistical process monitoring tools is vast, the application of these tools for big data sets has received less attention. In this expository paper, we give an overview of the current state of data-driven multivariate statistical process monitoring methodology. We highlight some of the main directions involving statistical learning and dimension reduction techniques applied to control charts in research from supply chain, engineering, computer science, and statistics. The goal of this paper is to bring into better focus some of the monitoring and surveillance methodology informed by data mining techniques that show promise for monitoring large and diverse data sets. We introduce an example using Wikipedia search information and illustrate a few of the complexities of applying the available methods to a high-dimensional monitoring scenario. Throughout, we offer advice to practitioners and some suggestions for future research in this emerging area of research.

Some Current Directions in the Theory and Application of Statistical Process Monitoring

The purpose of this paper is to provide an overview and our perspective of recent research and applications of statistical process monitoring. The focus is on work done over the past decade or so. We review briefly a number of important areas, including health-related monitoring, spatiotemporal surveillance, profile monitoring, use of autocorrelated data, the effect of estimation error, and high-dimensional monitoring, among others. We briefly discuss the choice of performance metrics. We provide references and offer some directions for further research.