EIS Solution Fameworks

Tool for 2 stage tolerance optimization

Tools for Capital Maintenance Planning

Supply Chain Planning under Uncertainty

Tool for fixed-structure parametric identification

Methodology for identification of linear and  nonlinear dynamic systems

A library of tools / methods for the development of Virtual Sensors using operational data

An object oriented module for online multivariable control and optimization of linear systems

Algorithms for adaptive networks based on adaptive resonance theory for classification and identification

Algorithms for fault diagnosis for linear and nonlinear systems using model based approach

Data Mining Tool

Tool for 2 stage tolerance optimization

Optimization formulation at stage one looks at the design in its true form as the tolerance allocation process is not constrained by manufacturing process capabilities. Selection of manufacturing process based on the capability is done as a second step that separates achievable from unachievable tolerances. A detailed formulation incorporating the process capability and standard part usage along with true design trade-off is solved to arrive at the final tolerance allocation in the second stage. In another approach an effort is put to bring in robustness in the design by changing the nominal design so that more tolerance is acceptable. The formulation not only allocates tolerances but also finds out best nominal design that will give most cost effective tolerances. The results obtained can be proved rigorously for their robustness and convergence properties. The methodology is demonstrated and validated using full scale industry problems.

Tools for Capital Maintenance Planning

The objective of the optimization based tool for capital maintenance planning is to maximize the benefits accruing as a resunvestment. While achieving the objective, the optimizer also ensures that various budgetary and other constraints are satisfied. The group is also working on development of repair rate approaches for analyzing the failures of equipment. A combination of repair rate and whole life-cycle costing methods have been used to determine the economic life of equipment groups, as well as to assess the condition of equipments.

Supply Chain Planning under Uncertainty

For supply chain planning under uncertainty, the group has developed a comprehensive multi-objective, multi-site, multi-product and multi-period planning framework that considers uncertainties among various process parameters (e.g. demand, machine uptime, inventory safety level, cost coefficients etc.) in a supply chain environment. Different cutting edge uncertainty programming paradigms like stochastic programming, fuzzy logic and probabilistic programming were used for finding the inherent theoretical lacuna in each one of them when adopted for supply chain planning. The framework when tested on large scale industrial supply chain planning problems shows very encouraging results as compared to many standard techniques that are most frequently referred in the open literature.

Tool for fixed-structure parametric identification

Optimization formulation at stage one looks at the design in its true form as the tolerance allocation process is not constrained by manufacturing process capabilities. Selection of manufacturing process based on the capability is done as a second step that separates achievable from unachievable tolerances. A detailed formulation incorporating the process capability and standard part usage along with true design trade-off is solved to arrive at the final tolerance allocation in the second stage. In another approach an effort is put to bring in robustness in the design by changing the nominal design so that more tolerance is acceptable. The formulation not only allocates tolerances but also finds out best nominal design that will give most cost effective tolerances. The results obtained can be proved rigorously for their robustness and convergence properties. The methodology is demonstrated and validated using full scale industry problems.

An object oriented module for online multivariable control and optimization of linear systems

This is a quadratic optimization problem formulation that uses a predictive model of a plant and achieves the optimal control and performance objective. This has resulted in many MPC (Model Predictive Control) strategy variants for a multivariable plant structure. 

Methodology for identification of linear and  nonlinear dynamic systems

A family of methods have been developed for modeling plants as linear or nonlinear dynamic systems of known structures. The parameterization of such structures is obtained by making use of the operational data, design of excitation design of experiments, or the combination of both. The tool then uses LS based and optimization based techniques to identify the model parameters. The models thus identified could be used for optimal control or for its predictive control.

A library of tools / methods for the development of Virtual Sensors using operational data

These techniques are applicable for inferential measurement of parameters that are either not measured using conventional sensors or less frequently measured due to the measurement process constraints. They address the measurement/estimation of derived parameters as well.
The lab has developed techniques for soft/virtual sensing based on the technology that spans over statistical methods, ANNs, PCA and LS methods and optimization techniques.

Algorithms for adaptive networks based on adaptive resonance theory for classification and identification

Adaptive networks for supervised and unsupervised learning for regression and classification (ART2, Fuzzy ARTMAP, GARTMAP  are developed to classify among the multidimensional  data. The classes extracted from the data are then identified against the database and are mapped for the class identifier.

Algorithms for fault diagnosis for linear and nonlinear systems using model based approach

These are the techniques based on analytical redundancy principle where the plant is modeled for additive and multiplicative faults occurring in it.  The challenge addressed here is to detect the occurrence of faults and isolate them from each other.

Data Mining Tool

Data mining tool consists of a number of knowledge discovery algorithms as a loosely integrated collection of algorithms. The algorithms are written in ANSI C and are portable across various platforms, including MS-WINDOWS and LINUX. The input data is usually taken from ASCII text files in comma separated (CSV) format. The output is also usually written to a text file in CSV format. The parameters to be provided need to be provided in an initialization file.
The algorithms included in the tool are for dependency analysis (association rule mining, sequence mining, frequent itemset mining), outlier analysis, clustering and graph clustering which are part of unsupervised learning techniques and classification (decision tree induction) which is supervised learning technique.