Using open source software in development of computer-aided design systems for radar

Relevance of work, research tools selection justification

The following example can illustrate the relevance of the solution of the problem related to computer-aided simulation of the displacement field of the radar station (RS) structure within computer-aided design (CAD) systems for radar. Nebo-M radar has been selected for research purposes. A specific structural model (Fig. 1, a) has been developed based on public domain pictures. As for materials, we have selected steel grade 40 and aluminum alloy. The estimates based on this model show that under the action of wind load the resulting total force (621 kgf) may lead to considerable structural deformation. The effect of such deformation on shaping of the antenna radiation pattern is significant enough to be taken into account. Fig. 1 (b) shows the results of a linear approximation based calculation carried out with the help of MSC.Nastran and Nastran SOL101 as an internal solver. The displacement pattern is represented at a scale of 1:10. Maximum displacement values are 153 mm (~4 % of the typical size of the model).

Fig. 1. Displacements of radar model elements: a - radar antenna structure; b - scaled field of radar antenna structure displacements

It is widely known that numerical modelling of air flows around complex bodies shall involve experts in aerodynamics, numerical methods, and the theory of turbulence. Besides, aerodynamic calculation requires skills in generation of high-quality computational meshes, the use of which will not result in considerable numerical errors in computations. Each geometrical model needs generation of its own computational mesh. Therefore, any aerodynamic calculation requires involvement of experts in the respective field. Stress-strain state (SSS) simulation requires knowledge in the fields related to stress- stain state mechanics and numerical methods. In a similar way to generation of computational meshes for aerodynamic calculation, the SSS simulation also requires a high-quality computational mesh to be generated by a researcher specializing in the SSS numerical modelling. Nevertheless, in order to take into consideration the effect of antenna deformations upon generation of the radiation pattern without assistance of experts in continuum mechanics to run a routine analysis, CAD systems for radars are used to develop software able to solve an adjoint problem related to aerodynamics, strength and heat transfer automatically: the user just has to enter the input data and set up the boundary conditions, while the computer system will automatically generate computational meshes and transmit data between solvers in an unattended mode. Therefore, a radar design engineer who is not an expert in continuum mechanics will be able to carry out a calculation of aerodynamics, heat transfer, and strength of the radar structure and to take into account displacements of structural elements related to external loads when forming the antenna radiation pattern.

We have selected open source software packages such as OpenFOAM (aerodynamics calculation), CodeAster (strength and heat transfer calculation), and Salome (graphical interface) as basic software tools to develop a system for simulating deformations caused by external impact.

These tools have been selected based on several factors. First of all, an open source code enables code manipulations, which allows to customize various components, such as solvers and mesh generators, for specific tasks, integrate such components in other software products and implement them into CAD systems for radar. Moreover, setting up automatic data exchange between various commercial software packages, for instance between Ansys CFX and MSC.Nas- tran, is a much more sophisticated problem than“opening” the source code of the Open- FOAM and CodeAster packages and interfacing the packages internally. Finally, if commercial software packages are selected for using CAD systems for radar at workstations, it would require purchase of licenses while open source software packages are available for free, and the performance indicators of open source solvers are competitive with those of their commercial counterparts [1-3].

Deformation simulation system architecture

The architecture of the deformation simulation system under development is based on the concept that, on the one hand, reflects the need for a common solution of the problems related to heat transfer, aerodynamics, and structural deformation under power and thermal loads by means of various computer aids and, on the other hand, depends on the task of implementation of efficient interaction between the operator, the software system, and the computer cluster.

The system functioning logic implies that operator (user) participation in the numerical modelling procedure will be minimized, provided that the input data and boundary conditions are entered correctly. It is assumed that the user can enter only the following three groups of input data:

• radar geometry;
• terrain and meteorological data;
• mechanical and thermal and physical properties of the structure.

The system features a two-level modular architecture (Fig. 2). The first level is intended to implement system-operator interaction:

• preparation and correction of input data;
• computation control;
• visualizing results.

Fig. 2. Architecture of deformation simulation CAD system for radar

The second level of the system implements its compute kernel that processes the following blocks of operations:

• input and output data analysis;
• generation of structure and environment computational meshes;
• numerical modelling of airflow dynamics;
• numerical modelling of the heat-stressed state of the radar.

The two-level modular structure of the system allows, first, to efficiently divide the processes of development and testing of system components associated with user input monitoring, as well as components intended for solving numerical modelling problems, and, second, to use components independently of one another, thus allowing to keep them up-to-date in case a new version of any software package is released.

The compute kernel units are based on the following libraries:

• radar computational mesh generator unit: Python, SALOME, Netgen,
• airspace computational mesh generator unit: Python, bash, OpenFOAM [4];
• heat transfer computation unit: Python, CodeAster [5];
• aerodynamics computation unit: Python, OpenFOAM,
• deformation computation unit: Python, Code_Aster.

The above-mentioned software products are based on universal methods of mathematical simulation (finite volume method and finite element method), which allow to simulate free form structures [6]. Each library listed above can be used for a wide range of tasks and has a complex and flexible syntax of input and output data.

Internal logistics of data flow among the deformation simulation system components is shown in Fig. 3. Intercomponent data exchange is an automated process: as soon as the user reaches the next computation stage, the previously entered data is transferred to subsequent components.

Fig. 3. Diagram of data flows between system components

Some libraries used for system development are currently showing consistent performance if running on Unix OS only. Consistent performance of the system running not only on Unix OS, but also on Windows OS has been stipulated as one of the requirements of the design specifications for the CAD system for radar. This problem has been solved with the help of virtualization technology [7]. The technology implementation in simplified form is shown in Fig. 4.

Fig. 4. Implementation of virtual machine for operation of deformation computation module

Using a virtual machine reduces the software operation speed but the loss of efficiency does not exceed 10 %, which appears an acceptable loss regarding the requirements for cross-platform capability of components that initially possess no such property.

Results of development

Based on the said concept, Salome GUI shell has been customized to meet the requirements of “Almaz - Antey” Air and Space Defence Corporation, and computational mesh generation has been automated (both finite volume and finite element mesh types).

Fig. 5 shows a customized GUI shell. The GUI shell has been customized by implementing tools for setting the input data such as model geometry, model materials, boundary conditions for aerodynamics, strength, and heat transfer calculation, parameters of computational meshes for control of calculation accuracy and duration. The lower section of the panel shows a diagram representing the current calculation status, in particular, the status of computational mesh generation and calculation progress is displayed.

Fig. 5. Salome-based GUI shell

The condition setup menu for computational mesh generation is of special interest. In the system under development, finite volume mesh parameters (Fig. 6) include the scale of the zone above the radar that describes the height of the computation area expressed in radar dimensions, plus the mesh resolution level and the maximum number of cells. The user has an option to enter the size of the cell located near a body, for example, an antenna or an antenna tower, which allows to monitor the resolution of individual structural elements. In particular, this option is very useful if the structure has small elements, and making a denser mesh near such elements is reasonable.

Finite element mesh parameters (Fig. 7) include a specific set of elements and optional element dimensions for particular bodies, as well as optional resolution of small structural elements of the radar.

Fig. 6. Finite volume mesh settings

Fig. 7. Finite element mesh parameters

It is worth mentioning that tools for monitoring the quality of meshes for aerodynamic and strength calculation have been implemented for the deformation simulation system under development. For example, the relationship of linear dimensions of finite element mesh edges is checked (5 max.); the maximum allowable angle between edges at vertex is 20°. For finite volume mesh, the quality check includes measurement of percentage of elongation of cells, mesh nonorthogonality, and relation of maximum and minimum cell sides. If an automatically generated mesh fails to meet quality requirements, the user will be warned about it and prompted to change mesh generation settings, for example, to set up the cell size for any structural element of the radar.

For comparison, commercial software packages have more consistent automatic tools for mesh generation. For example, to generate a finite volume mesh, Ansys has an option to set up the boundary layer size by selecting the number of layers and the cell growth rate from one layer to another. On the one hand, this approach gives the user more options and, therefore, allows to obtain more accurate results, but, on the other hand, it requires expertise in computation methods regarding continuum mechanics. Taking into consideration that the system under development is intended for engineers not specializing in computational aerodynamics and strength calculations, an insignificant (as described below) penalty in accuracy in exchange for enhanced user friendliness is acceptable.

Besides, it is worth mentioning that the current system version has an option for geometry import from standard software packages intended for structural design (AutoCAD, SolidWorks, CATIA, etc.); however, import of a computational mesh, for instance, is impossible. Mesh import tools may be developed at further research stages, if required.

To test the performance of the deformation simulation system, the integrity of data exchanged between components, the correctness and automation of a computation chain, Nebo-M radar has been selected, which was earlier used for manual calculations by means of commercial software packages. Fig. 8 represents the results obtained by the deformation simulation system. Comparing the final results of simulation, in particular, the displacement field computed automatically with the help of the deformation simulation system (Fig. 8, c) with the results obtained by means of a commercial software package (see Fig. 1, a), we may conclude that the difference does not exceed 7 %.

Conclusion

The paper shows the results of development of a computer-aided radar design software system component that provides computer simulation of the radar structure airflow and the heat-stressed state of the radar structure. We have managed to achieve the main objective of development, i. e., to give a radar design engineer not specializing in continuum mechanics a software tool for aerodynamic and strength calculations. In an overview, we have compared the developed system with commercial products regarding mesh settings convenience and accuracy of obtained results. The paper describes the solution to the problem related to support of cross-platform deployment of software, consistent performance of which is only ensured when running on Linux family operating systems.