PenSim Project:
PenSim stands for Penicillin Simulator. We have developed a number of Graphical User Interfaces (GUI) for this simulator that can be run under different platforms. PenSim relies on unstructured mathematical model we have developed by extending existing models in the literature to account for engineering variables. The original simulator was developed in Matlab environment and compiled into C modules. These C modules were also integrated into a web-based GUI application. Recently, this simulator has been converted into C code for better performance and it is run on our web server. A stand-alone version that can be run under any Windows 32-bit environment was also developed and can be downloaded using the second link below. Stand-alone version (PenSim v2.0) allows users to introduce faults to the input variables and produces nicer graphs if desired.
PenSim v.2.0 Stand-alone simulator (for Windows 95/98/NT/2000/XP)
Development of Web-based Simulator for Education and Research
Unstructured mathematical model and its simulator were adapted into a web server application. This way simulator can be reached remotely and without consuming CPU power on the local computers. Graphical User Interfaces were developed. This application is used as class projects (fermentation process control and biochemical engineering) in various chemical engineering departments throughout the world. A stand-alone version was also developed and posted on the web to be used as data generator for studying empirical modeling, controller performance assessment and process monitoring techniques. This version has also received attention from scientific and industrial community.
PenSim v.1.0 Web-based simulator
Development of Complex Mathematical Models for Fed-batch Penicillin Fermentation
Unstructured and structured (detailed) mathematical models and simulators are developed for penicillin fermentation. Models are representative for a range of secondary metabolites productions with non-growth associated reaction kinetics. Models produced consistent results with published data and also impressed professionals involved in industrial microbial fermentations. Unstructured model is extended for use as a data generator in studying empirical modeling, process performance and quality prediction, and controller performance monitoring techniques.
Data-based Modeling of Batch/Fed-batch Processes
Linear techniques based on reduced space modeling such as principal components analysis (PCA) and projection to latent structures (PLS) have successfully used and extended for modeling batch process data. Batch data challenges including unequal, unsynchronized trajectories, outliers, and missing data have been resolved. A detailed set of techniques and software were developed for data length equalization and time alignment of trajectories. Research has resulted in new insight into integrated development and deployment of data alignment methods and statistical process monitoring techniques.
Process Fault Detection and Diagnosis
Online batch process performance monitoring and quality prediction techniques were developed based on the empirical models. Detailed comparative studies were undertaken in the development of efficient fault detection and diagnosis (FDD) framework. Techniques developed and implemented include adaptive PCA, multiway PCA and PLS, multiblock PCA and PLS, and multiscale PCA. Performance of the techniques was tested with both simulated and industrial data. Research has also been conducted for developing FDD techniques for multistage batch processes.
Knowledge-Based Systems for Intelligent Process Operations
Online process performance monitoring and quality prediction techniques were integrated into knowledge-based systems for automation and process supervision in real-time. G2 is used as KBS development tool. Our research has resulted generic KBS software for automated detection and diagnosis of abnormalities in batch processes. Integrated use of statistical inference and process specific knowledge has provided improvements in diagnostic capabilities of the monitoring framework. Fed-batch penicillin fermentation simulator is also integrated into supervisory KBS to generate data in real-time for case studies.
|
