Mustapha Ouladsine

Generalization and analysis of sufficient conditions for PCA-based fault detectability and isolability

Abstract Fault detectability and fault isolability concepts are necessary to be developed in order to determine whether the process faults are detectable and isolable. In PCA framework, the development of these concepts has been limited to the use of some particular detection indices. This paper provides an extension and a generalization of the fault detectability and the reconstruction-based fault isolabilty concepts in order to be valid for the use of any detection index having a quadratic-form. Fundamental fault detectability as well as fault isolability analyses based on the combined index compared to those using both SPE and Hotelling’s T2 indices are investigated. In addition, we have illustrated the proposed analyses from a simulation example. This one highlights the advantage of the combined index into the isolation of some process faults that have not large enough magnitudes to be isolable neither by SPE index nor by Hotelling’s T2 statistic.

Fault prognosis for Discrete Manufacturing Processes

Abstract This paper deals with a fault prognosis method, based on the extraction of a health indicator (HI) from a large amount of raw sensors data, applied to Discrete Manufacturing Processes (DMP). The HI is extracted by locating the significant points of machine which are related to the degradation. The dynamics of HI is then analysed and modelled using an appropriate stochastic process. The adaptive aspect of the prediction model allows the updating of the Remaining Usesul Life (RUL) estimation. The developed approach is applied on a real case provided by ST-Microelectronics, where experimental result shows its efficiency.

Health Index Extraction Methods for Batch Processes in Semiconductor Manufacturing

This paper deals with a study of three methods for health index (HI) extraction in semiconductor manufacturing equipments. The first method uses degradation reconstruction-based identification with basic principal component analysis (PCA), the second one uses multiway PCA and the last one extracts HI from the significant points related to degradation. A comparison of these methods are made discussing about their efficiency and shortcoming for the implementation. The studied methods are applied on two data sets: 1) a simulation case and 2) a real case provided by ST-Microelectronics, where experimental results highlight the advantages and limits of each one.

Application of PCA for efficient multivariate FDC of semiconductor manufacturing equipment

With the evolutions in sensing technologies and the increasing use of advanced process control techniques, terabytes of data are recorded today from manufacturing equipment during the process of semiconductor devices. These large amounts of data are then operated by FDC systems to assess the overall condition of the equipment. In this paper, we consider the Exponential Hybrid-wise Multiway Principal Components Analysis (E-HMPCA), a PCA-derived model that include an Exponentially Weighted Moving Average component, for the condition monitoring of a Chemical Vapor Deposition tool in STMicroelectronics Rousset 8” fab. In order to work directly on temporal signal from equipment sensors, the application of Dynamic Time Warping for data synchronization is also presented. A real-occurred failure case is used to highlight the benefits of this approach on detection efficiency improvement and monitoring complexity reduction.

A note on unknown input interval observer design with application to systems prognosis

This paper is a contribution to the problem of systems prognosis. More precisely, in this work, the goal is to introduce an unknown input interval observer for a class of uncertain system with an unknown input for systems prognosis. The observer synthesis is applied to unknown variables estimation of an electromechanical oscillator with a non-stationary two-well potential to illustrate the pertinence of the proposed approach.

Active Fault Diagnosis and Major Actuator Failure Accommodation: Application to a UAV

Interest in Unmanned Aerial Vehicles (UAVs) is growing worldwide. Nevertheless there are numerous issues that must be overcome as a precondition to their routine and safe integration in military and civilian airspaces. Chief among these are absence of certification standards and regulations addressing UAV systems, poor reliability record of UAV systems and operations. Standards and regulations for airworthiness certification and flight operations in the military and civilian airspaces are being studied (Brigaud, 2006). In this respect, the USAR standard suggests a mishap rate of one catastrophic mishap per one million hours (Brigaud, 2006). To reach such performances, upcoming technologies have the promise of significantly improving the reliability of UAVs. In this connection, a detailed study (OSD, 2003) shows that most of the breakdowns are due to system failures such as propulsion, data link and Flight Control Systems (FCS). These latter include all systems contributing to the aircraft stability and control such as avionics, air data system, servo-actuators, control surfaces/servos, on-board software, navigation, and other related subsystems. As regards FCS, it is recommended in (OSD, 2003) to incorporate emerging technologies such as Self-Repairing Flight Control Systems (SRFCS) which have the capability to diagnose and to repair malfunctions. In this respect, Fault-tolerant control (FTC) are control systems that have the ability to accommodate failures automatically in order to maintain system stability and a sufficient level of performance. FTC are classified into passive and active methods. The analytical fault-tolerant control operation can be achieved passively by the use of a control law designed to guarantee an acceptable degree of performance in fault-free case and to be insensitive to some faults. However, the passive methods are unsuitable to deal with a significant number of faults. In particular, for an aircraft, it may be tricky to design an a priori controller able to accommodate the whole of the faults affecting the control surfaces. By contrast, an active FTC consists of adjusting the controllers on-line according to the fault magnitude and type, in order to maintain the closed-loop performance of the system. To do so, a fault detection and isolation (FDI) module which provides information about the fault is required (Noura et al., 2009). Active FTC mechanisms may be implemented either via pre-computed control laws or via on-line automatic redesign. Active Fault Diagnosis and Major Actuator Failure Accommodation: Application to a UAV 7

Virtual Metrology applied in Run-to-Run Control for a Chemical Mechanical Planarization process

This paper deals with missing data in semiconductor manufacturing derived from a measurement sampling strategies. The idea is to construct a virtual metrology module to estimate non measured variables using a new modified Just-In-Time Learning approach (JITL). The aim of this paper is to integrate estimated data into product control loop. In collaboration with our industrial partner STMicroelectronics Rousset, the accuracy of the proposed method is illustrated by using industrial data-sets derived from Chemical Mechanical Planarization (CMP) process that enables us to compare results obtained with the classical and the modified version of JITL approach. Then, the contribution of the estimated data is shown in product quality improvement.

Heart rhythm characterization through induced physiological variables

Atrial fibrillation remains a major cause of morbi-mortality, making mass screening desirable and leading industry to actively develop devices devoted to automatic AF detection. Because there is a tendency toward mobile devices, there is a need for an accurate, rapid method for studying short inter-beat interval time series for real-time automatic medical monitoring. We report a new methodology to efficiently select highly discriminative variables between physiological states, here a normal sinus rhythm or atrial fibrillation. We generate induced variables using the first ten time derivatives of an RR interval time series and formally express a new multivariate metric quantifying their discriminative power to drive state variable selection. When combined with a simple classifier, this new methodology results in 99.9% classification accuracy for 1-min RR interval time series (nu2009=u20097,400), with heart rate accelerations and jerks being the most discriminant variables. We show that the RR interval time series can be drastically reduced from 60u2009s to 3u2009s, with a classification accuracy of 95.0%. We show that heart rhythm characterization is facilitated by induced variables using time derivatives, which is a generic methodology that is particularly suitable to real-time medical monitoring.

Splitting method for spatio-temporal sensors deployment in underwater systems

In this paper, we present a novel stochastic optimization algorithm based on the rare events simulation framework for sensors deployment in underwater systems. More precisely, we focus on finding the best spatio-temporal deployment of a set of sensors in order to maximize the detection probability of an intelligent and randomly moving target in an area under surveillance. Based on generalized splitting technique with a dedicated Gibbs sampler, our approach does not require any state-space discretization and rely on the evolutionary framework.

Product quality prediction using alarm data : Application to the semiconductor manufacturing process

In the complex manufacturing processes, high quantity of products might be rejected. This can be due to the no detected failures. To evaluate the processing of manufacturing steps, alarms are setting off to indicate failures. However, industrial plant operators often receive many more alarms than they can manage, which include correlation. A poor alarm system may cause nuisance alarms and thus alarm floods, which reduces the ability of operators to take actions. This paper aims to identify unnecessary alarms within a large amount of event data. We prove the equivalence between similarity approaches in case of sparse binary data. The second purpose of this paper is the product quality prediction based on historical alarm events by using a regularized regression method. To demonstrate the effectiveness of these tools and their utility in the product quality prediction, we present an industrial case study based on alarm and scrap data collected from a semiconductor manufacturing process. Application results show the practicality and utility of the proposed methodology for both alarm management and product quality prediction.