XTREME PERFORMANCE ADVANTAGES

1. Fast and very efficient optimization algorithms

1 Xtreme was developed to target large optimization problems that are found in aeronautics, automotive and energy applications. The common features of all these applications are the large computational time required to run one simulation (typically more than 1 hour) combined with a number of design variables around 100. This is why Xtreme was developed by combining the qualities of genetic algorithms and artificial neural networks in order to largely increase the efficiency of the optimization algorithms. Compared to the most advanced genetic algorithm, Xtreme will be about 25 times faster. Xtreme includes dozens of algorithmic innovations that will outperform any other optimization algorithms.

2. Xtreme is capable to deal with complex optimization formulation

2 Genetic algorithms (GA) are known to be capable to deal with complex optimization applications including: highly nonlinear problems resulting from the physics or the constraints, noisy functions, mixed integer/real design variables, no access to the simulation software source code, ... In Xtreme, the usual drawback of GA (which is the large number of simulations required to reach convergence) has been solved with the use of an artificial neural network. All the other qualities of the GA have been kept.

The complexity of the optimization process can come from different sources. In all these cases Xtreme will be the best tool you will find to solve your applications.

1. The relation between the design variables and the responses are highly nonlinear.
2. The optimization formulation contains constraints that will create highly nonlinear problems.
3. The simulation tool is a black box : you do not have or do not want to have access to the source code.
4. The optimization formulation contains many objectives, many constraints or interval constraints.
5. The responses are noisy.
6. The simulation tool is not always able to provide you with a converged solution.
7. You have mixed real/integer design variables. The complexity of your optimization formulation can usually be decreased by a specific review of your optimization process. For example, looking at your choice of design variables, of responses, objectives and constraints and implementing specific solution can usually have a major effect on the performance of your optimization process.

OPTIMAL COMPUTING can help you to drastically simplify your optimization process and improve its performance. Do not hesitate to contact us.



3. Xtreme is robust and largely tested

3 The Xtreme software is tested on a large number of well known optimization problems. Moreover, the algorithms are tested a large number of times with different initial conditions in order to make sure that the rate of success of the optimization algorithm is maximized.

4. Both a parallel and a sequential version of the optimization algorithms are available

4 An important aspect is to make use of the parallel processing while evaluating the external simulation. This is possible with Xtreme for all 4 types of algorithms. We have developped highly efficient parallel optimization algorithm coupling the genetic algorithms and the neural networks

XTREME TECHNICAL ADVANTAGES

Xtreme is based on a number of innovative algorithms that makes the software unique on the market.
The advantages which make Xtreme unique are summarised hereafter.

1. Xtreme is a black-box optimizer

1 A black-box optimizer is an optimizer that works only with the responses provided by the standard version of the simulation chain. With a black-box optimizer there is no need to have access to the function derivatives and to modify the simulation source code to compute these functions derivatives.

Template based numerical optimization In a very large number of industrial applications, it is even not possible or not suitable to access the simulation source code to have access to function derivatives or some specific knowledge of the simulation. This is often not possible because parts of the simulation tools are commercial codes or are too complex to understand or modify. This is often not suitable because developing efficient method to access the function derivative will very often require a huge investment and in a vast majority of applications will not give any advantages compared to Xtreme algorithms.

By using the black-box optimizer inside Xtreme you have an incredible flexibility and efficient optimizer. You can use as many simulation tools as required in your simulation chain, you can use your preferred commercial simulation codes, you can perform multi-physics optimization, you can optimized coupled multi-physics simulations, ...



2. Xtreme is very simple to use

2 During the development phase of Xtreme, a specific quality criterion was to present to the user only the strictly necessary parameters required to set up the optimizer: design variables, responses, objectives and constraints and finally the number of design cycles. With Xtreme there is no numerical parameter to be chosen or tuned to get a correct performance of the optimizer.

3. Xtreme very efficiently solves optimization problem having a time consuming simulation

3 While running the optimization process, the numerical optimizer will request a number of executions of your simulation chain. If the computation time required by your simulation package is not negligible (> 1 sec) the number of calls to your simulation package must be reduced as much as possible. This is important in order to reduce the required elapsed time needed by the optimization process.

One of the most important strength of Xtreme is to drastically reduce the number of evaluation requested to your simulation chain. Typically, the number of function evaluation is reduced by a factor of 100 compared the most advanced genetic algorithms you will find in the research community or on the market.

If your simulation time is larger than 1 sec and you have several design variables, Xtreme is the tool you are looking for. Very often, the simulation time can be reduced in a number of ways.

OPTIMAL COMPUTING has a large expertise in all key aspects that will allow you to reduce the time required by your simulations.



4. Xtreme can deal with many type of design variables : double, integer and list of discrete values

4

5. The number of design variables is in the range 1 to 200

5 Xtreme is capable to deal with optimization applications having a large number of design variables. This is a unique feature of our software that you won't find in any other numerical optimization software based on genetic algorithms. The selection of the design variables is a key factor in the success of your optimization applications. It requires building a parametric model of your product or process.

OPTIMAL COMPUTING has a large expertise in this specific area. We help you in building the parametric model, selecting the most appropriate design variables or reducing the number of design variables from several thousand variables to a much smaller number.

FINANCIAL ADVANTAGES

1. Xtreme allows you to cut your license cost

1 With Xtreme the optimization software license price is not anymore a barrier and a difficult choice. The licenses are very cheap, they are flexible in terms of both the modules you want to use and the validity period of the license. Moreover you can easily get a trial version to simply test a first version provided in an Microsoft Excel spreadsheet.

Xtreme is our optimization software which will help you to improve your product or process. Our optimization algorithms have unique performance and robustness and will allow you to cut your license cost by up to 80% compared to traditional commercial solutions.