Polytechnique > Research > Directory > A professor or researcher

Directory of Expertises

Photo of Soumaya Yacout

Soumaya Yacout

Department of Mathematical and Industrial Engineering


Research interests

 Diagnosis and Prognosis of Industrial Process Performance and Condition Based Maintenance

Engineering systems, such as aircraft, transportation systems, manufacturing systems, and industrial processes such as mining, are becoming more complex and are subject to failure modes that are difficult to detect and to explain. This situation has a negative impact on the systems' reliability, availability, safety , serviceability, maintainability and productivity. On-line, real-time fault detection, analysis, diagnosis and prognosis tools can assist the operator in achieving his/her mission of making the systems functional and safe to use quickly and efficiently. Recent advancements in Condition - Based Maintenance (CBM) and Equipment Health Management (EHM) produced new and innovative methods for diagnosis and prognosis of systems condition. Yet, these methods still have limited aplicability due to their structural limitations, to their complexity, and to their lack of readiness to use. In most industries, the most used diagnostic systems are still based on human expertise. This expertise, although very valuable, still has limitations, in particular in situations where multiple failures modes or multiple inputs to the system are interacting in an unknown new correlated manner that has not been seen or documented before. Moreover, this expertise can be lost due to age or retirement or resignation, thus leading to the loss of knowledge. Even if provisions are taken for knowledge transmission, this process may be time consuming and exhausting. Equipment Health Management systems need improvements in order to overcome these limitations, and to increase their efficiency, accessibily, applicabitity, and explicability power.

This research aims at developing an integrated system for diagnosis and prognosis that is generic and applicable to a variety of engineering systems. The system is capable of processing data with different data analytics techniques, and via data mining , it extract hidden information in the form of patterns, find correlations between inputs and root causes that can explain different known and unknown performances, in an exploitation and exploration phases. The tool is based on artificial intelligence notions of machine learning and testing, and use a reinforcement learning approach to improve its learning capabilities with time and  via accumulated knowledge, and as the computational capabilities of computers are enhanced.

A new software called cbmLAD is developed at École Polytechnique and it is now available for use.


Decision making based on Knowledge Discovery and Data Exploitation for Industrial Processes

    Condition monitoring is the process of monitoring the operating characteristics of a process so that changes and trends of the monitored characteristics can be used to predict the present and the future process ' performance, and in order to control process's output.

Recently, there have been considerable research efforts to develop condition monitoring technologies for industrial processes and systems. The use of these technologies has resulted in the acquisition of large amounts of data and has given life to new fields of research, namely data mining and knowledge discovery from databases, and data analytics. In the past, one of the main problems in industrial setting was  the lack of data needed to support the decision making process.  Nowadays, most companies use one or more condition monitoring technologies and possess considerable databases containing performance indicators for their processes.  Consequently, researchers are now interested in finding techniques to extract information and interpret it accurately.

 The objective of the proposed research is to use a new approach called Logical Analysis of Data (LAD) in order to exploit the databases of industrial processes ,to assist in decision making, and  to improve process performance. LAD is  a data mining artificial intelligence approach that is based on pattern recognition. It is a combinatorics and optimization-based data analysis technique. Historical data containing performance indicators and process output measurements are exploited in order to generate patterns that characterize the process performance. Data fusion techniques based on the generated patterns are used in order to combine information which is coming from  different sensors. Pattern-based clustering techniques are developped. Multi-clasification problems are  solved, and reinforcement pattern- based learning techniques with intelligent agent are developed.  




NSERC subjects

  • 1606 Operations management


© École Polytechnique de Montréal
Find Peoble or units | Site plan | Key word search | Conditions(*) this link leads to a section in french.