·Table of Contents
·Methods and Instrumentation
A Computer Based Methodology for Better Welding Quality
Fairouz Bettayeb, Amar Benchaala
Scientific Research Center on Welding and Non Destructive testing.
C.S.C, Route de Dely Brahim, BP:64, Chéraga, Algiers. Algeria.
Tel/Fax: (213-2) 361850. Email: email@example.com, firstname.lastname@example.org
University of Bab Ezzouar.USTHB.BP:32 El Alia.Algiers
The most important factors for the industrial plants during their working and maintenance are the security and the quality. The assurance of the structures integrity, is confirmed by a lack of defects during the construction process and particularly during the service of the component. Indeed, the most manufactured processes are based on complex physical phenomena, for which it is difficult to lay down realistic analytical models for determining optimal manufacturing conditions. Fabrication by the use of the welding technology necessitate a reliable and powerful specification, and welding parameters are chosen to match the strength and toughness mechanical properties of the base material.
The drawing of a welding specification is based on experimental methods and human qualifications according to the following pattern: the parameters are determined in several phases by the use of fabrication standards and empirical rules. But in most cases, relations between parameters require a decision making for starting the reasoning, which can be challenged by other deducted parameters. Generally, this scheme of deduction produces an unreliable welding quality.
New modelling approaches which aim at the productivity and quality, allow the definition of better fabrication conditions, thanks to efficient methods that conciliate technology and economy. In this paper, we present a computer aided welding specification, modelled on combinatorial analysis approach. The system draws an expertise plan for an automatic determination of the variation laws of the welding parameters.
Key words: Fabricated process, welding technology, numerical analysis, modeling, welding parameters, structural optimization, simulated annealing.
The tendency for industry to require higher performance fabricated structures places an ever-increasing power on inspection methods to determine accurately the degree of integrity of such constructions. Nowadays, economic costs affecting manufacturing strategies contribute to promote the optimization of the joining methods with an increasing credit on mathematical modeling. This paper presents, the development of a computerized system for use in arc welding engineering. It describes how the system is modeled and how it is used to control the implementation of welding procedures of steel structures. The objective of the PC-based system called 'Optima' is to provide the welding engineer with sufficient informations to design the most economic and reliable welded component for a given set of the fabrication conditions.
Establishing the needs for the project and the complexity of the welding procedure specification, the requirements for welding documentation in the fabrication of steel structures shall contain :
The documentation shall be sufficiently outlined, identifying welding procedure specification, welders, inspection method and need repair reports . The modeling of welding by means of software technology must include the thermal properties of the base and filler metals. These properties are the thermal conductivity, specific heat and density of the materials . Each specification consists of data in 3 groups: joint geometry, welding heat treatment data and welding parameters.
- The welding procedure specification,
- The welding procedure qualification,
- Test documents,
- Inspection methods.
3. Description of the welding specification
Earlier, it was the welding engineer who makes the selection of the welding consumable, with help of other experts, catalogues from electrode manufacturers, and documentation from steel-works. Material and welding method have to be chosen and then the method dependent parameters have to be filled in .
Now it's possible to make this selection with the help of a computer, and complex simulations become an effective memory for choosing the welding parameters . Good quality welds rely on the correct weld pool size, geometry and position relative to the weld preparation. In the case of the manual metal arc welding process, complex simulations of arc welding power source, try to become a new means for better command of arc welding. Indeed, the simulation shows even effects of small variations of a design parameter with a very high resolution, not available by other means.
3.1 The welding parameters
Parameters for the welding of structural steels are chosen primarily to match the strength and toughness properties of the base material. Typical selection criteria may be: type and thickness of base material, chemical composition, required mechanical properties of filler metal, type of joint, welding position, welding technique, electrode type, diameter and position, hardness, preheating, and post weld heat treatment .
'Optima' has been developed on the study of the properties of the welding information system. This study has shown that the system can be represented by a decision graph, in which each welding parameter is a graph node connected to the other parameter by links that have influences on the optimum value. These influences have been implemented by weighting factors.
3.2 Classification of the optimizing parameters
The relevant parameters have been organized in 3 sets as follow:
- The basic parameters have discrete values without any variation during the welding operation, i.e.: electrode type, diameter and coating type...
- The operating variables affect the working of the process, and can be adjusted during the welding operation i.e.: intensity I, voltage U, welding speed V...
- And particular variables that are observable and cannot be modified by computer, i.e. welding position or modifications in the metal welding conditions
4. The resolving scheme.
The complexity of the combinatorial optimization trials, makes the necessity to link the artificial intelligence and the operational analysis tools, in order to find solutions for complex and imprecise problems that cannot be resolved by exact solutions . In reality, the most manufactured processes are based on complex physical phenomena for which it is difficult to lay down realistic analytical models.
However, many approaches as the neural networks, the genetic algorithms, the simulated annealing and others, will determine optimal manufactured conditions, but no one of these techniques has conducted yet to a universal methodology. In this paper, the simulated annealing solution (S.A) has been chosen as a resolving strategy, thanks to the convergence of the annealing and the welding principles, as being described ensuing.
4.1 Convergence of the annealing and the welding fundamentals.
The annealing is a thermal treatment that consists in the heating of the metal at a highest temperature, and the cooling of this metal gradually in order to obtain a homogeneous structure. The physical principle of the arc welding is based on the calorific energy produced by the electrical arc flashing between the electrode and the pieces, which creates a fusion bath that solidify after cooling. This fundamental similitude between the physical states, has guided to the choice of the simulated annealing algorithm for the optimization scheme.
4.2 The simulated annealing approach
The 'Metropolis' algorithm , which implements the S.A approach brings the analogy of a combinatorial optimization problem, with a physical system composed of a great number of particles in iterations. In such system, the objective function is simulated by the free energy of a physical system, the parameters are simulated by the particles coordinates, and the exploration of a good configuration by the study of the low energy structures. The simulated annealing variables are: the initial temperature, the initial solution, the choice of an objective function, the time of the relaxation, the decrease law, and the resolving space.
- Ck is the estimated configuration at iteration k
- E(Ck) is the corresponding energy & T is the temperature
- The transition to the state k+1 is performed in the following 2 steps:
Step1: the selection of a new configuration C' by the application of the cost function
E = E (C') - E (Ck)
0 then Ck+1 ¬
Else extract randomly PÎ
If 0 < P < exp (- D
E/T) then Ck+1 ¬
C' Else Ck+1 ¬
4.3 Mathematical formulation of the cost function
Optimum design is a key synthesis which collects all-important engineering aspect to develop modern structural constructions, not only safe but also economic. The economy is achieved by minimizing a cost function and the safety is guaranteed by implementing the design constraints. Therefore the problem can be defined mathematically as a constrained function minimization task, which may be solved by mathematical programming methods .
The complexity of the system 'Optima' is defined by a highest cardinal space and several constraints. Thus, the objective function has been defined by the sum of the relative errors near the neighboring of the required criteria of the system . In the following definition, an estimation is assigned to each welding parameter, with consideration of the constraints features of the requested quality. The mathematical optimizing formulation of the objective function is:
5. "Optima" strategy
The system leads to a near optimal solution and proceeds by generating a certain number of neighbors at each temperature, while the temperature parameter T is gradually decreased. T is the temperature in the physical annealing process. The initial temperature in the simulated annealing algorithm is kept high as well as the algorithm does not block itself in a local minimum. The algorithm chooses an initial solution at random, a neighbor of this solution is then generated by a suitable mechanism and the amount in the cost function of the neighbor is calculated. If a reduction in the cost function is obtained, the generated neighbor replaces the current solution. At this point, we must indicate that the most difficult step of the implementation was the initialization of the system, due to the absence of usable theoretical welding background. Thus, in order to determine the influence of the initialization on the system and to set the initial values, the following experimentation has been conducted and the results and comments, are summarized below.
1 - In order to improve and exploit all the potential of the S.A method, new procedures based on the 'Metropolis' philosophy have been implemented. The procedures are called MST, MSP, MSTD, MSPD, and MSTID . A testing of their performance has been conducted on several experiments, and on 10 essays the best one has been saved. These experiments have concerned the behavior of the procedures under different conditions dealing with the space size of the variation laws of the variables, the constraints severity, the nature of the objective function and the initialization type.
As a result, the diagram 1 drives the following interpretations:
- The solution is estimated to the optimum, for little sized configuration (a small ranged variation of the variables), whatever the initial solution type (random or not), and the objective function nature (the weighting factor of the variables and the target neighboring).
- For more greater configuration, the procedures have a slow execution time, due generally to the penalization of non-verified constraints. At this stage a good initialization is preferred.
2 - Since the dominant character of the ameliorative heuristics is the random selection, and because the indiscriminate property of the S.A method, we have worked by simulations on 5 examples dealing with the influence of the algorithm initializations on the finale solution. The initialized parameters are the initial temperature (ti°.), the iteration count (nt.) for each temperature scale, and the decreasing temperature ratio (rt.).
The diagram 2 indicates that:
The diagram 3, shows that:
- An expanded (ti°) do not gives an excellent solution, but increases the operating time.
The diagram 4 shows that:
- The solution is better when (nt.) is increased.
- A slow decreasing ratio (rt.), gives a good solution but with an important performing time.
On the entrance of the 21st century, the power of information technology and materials technology, including its transportation and recycling, determine precisely the expectations and perspectives of the modeling tools for better fabrication quality. An optimum design procedure can be devised into 3 main phases as ensue:
Following the 'Optima' pattern, it will be indicated a considerable reduction in engineering hours for making new welding procedures or revising old ones, because in most cases new procedures are based on old ones. And due to the feed back easiness, number of inspection resources can be avoided and a percentage of repair costs have to be economized.
- The preparation step by the selection of materials, type of structures, fabrication methods and cost function.
- The minimization step of the constrained function by computerized mathematical methods.
- And the engineering evaluation of the computed results.
- Rostadsand, A. & al. (1991). In: Proceedings of the first conference on joining technology, pp.103-112. Sacer-Is/Cattani (Ed.). France.
- Roelews, J.B. (1995) Welding in the world, Vol. 35, N°2, P.110.
- Blumschein, E. (1991). In: Proceedings of the first conference on joining technology, pp.59-66. Sacer-Is/Cattani (Ed.). France.
- Niset, M., Daemen, R. (1988) Soudometal, IIS (Ed.). Brussels, Belgium.
- Bettayeb, F. & al., (1998). In: Proceeding of the 7th European conference on non destructive testing. Vol. 2, pp.1258-1265. Force Institute (Ed.). Copenhagen, Denmark,
- Sakarovitch, M. (1989), Optimisation combinatoire, méthode mathématique et algorithmique, Hartman (Ed.). France.
- Colin, J.Y. (1991), Doctorat thesis, Paris V University, France.
- Siarry, P., Dreyfus, G. (1988). La méthode du récuit simulé, théorie et application, IDESET (Ed.), Paris.
- Bettayeb, F. & al, (1999). In: Proceedings of the 14ème Congrès Français de mécanique, Universal & ENSAE (eds.), Toulouse, France.