Two objectives, namely power loss and total bearing weight, are considered one by one. 152 Figure 5.6 Optimized geometries of magnetic bearings for the objective function as. a) Minimization of overall weight (b) Minimization of power loss (c) Minimum weighted normalized distance of objective functions.

## Introduction

The term magnet is believed to be coined from the place called magnesia, where magnetic ore (ie the loadstone) is found in abundance. Apart from the applications for motors and generators, the permanent magnets were tried to levitate an object without contact.

Basic components of an AMB

## Classification of magnetic bearings

Hybrid: AMBs that also incorporate permanent magnets are called hybrid magnetic bearings (HMB) (Groom et al., 2000). Electromagnetic effect: This is due to the drag force that causes magnetic bearings.

## Advantages and applications of magnetic bearings

Kasarda (2000) discussed some of the advantages of commercial and research applications of magnetic bearing technology. The detailed advantages and applications of AMB technology were investigated by Schweitzer et al.

## Limitations of magnetic bearings and research areas

A sudden failure of some power amplifiers or coils due to power failure or damage results in the instability of the AMB system. In the worst case, the entire control system becomes destabilized due to the failure of all power amplifiers and coils or due to overload.

## Literature review

*Design methodologies**Configurations of magnetic bearings**Magnetic thrust bearings**Control system technology in magnetic bearings**Genetic Algorithms in magnetic bearing design optimization optimization*

Song et al. (2001) introduced a bidirectional bearingless axial-slot motor into a brushless direct current (BLDC) blood pump. Parametric and systematic approaches have been developed to optimize the power-to-weight ratio of magnetic bearings (Malone, 1993) and to achieve minimum power (Klesen, et al., 1999).

## Books, conferences, and journals

Schweitzer and Maslen (2009) compiled an expertise of nine pioneers in the field of magnetic bearings for rotating machinery. The book begins with an overview of the technology and an overview of the range of applications. At the end of the book, some special concepts and systems, including microscale bearings, self-supporting motors and self-sensing bearings are presented as promising directions for a new research and development.

## Aim and objective of the present work

### General Challenges in the Design of AMBs

The heat is mainly generated by the electrical resistance of the coil and is limited by the critical temperature of the insulation material and the heat absorption capacity of the cooling system used. With conventional mechanical bearings, some of the rotor's power is lost due to friction (i.e. damping). The compatibility of the power amplifier and the controller is another task when designing magnetic bearing systems.

### Objectives of the Design Optimization and its Importance Importance

However, the integrated design methodology implemented in this paper considers all components simultaneously to optimally design and analyze. The above discussions are valid for both radial and thrust AMBs; however, our main focus now is working on AMB propulsion without and with bias magnets. Most of the research in the design optimization of AMB systems has been carried out by single-objective optimization, especially of the controller unit.

## Organization of the thesis

Genetic algorithms are implemented with an objective to optimize (minimize) coil power loss and actuator weight, one at a time. The trade-offs, namely the minimization power loss and the actuator weight will be studied together. Performance parameters of double-acting actuators and magnetic bearing controller for various choices are presented.

## Introduction

For AMTB, these fundamental relationships can be obtained by eliminating terms corresponding to permanent magnets.

## Geometry of Magnetic Bearings and Fundamental Relations

### Geometrical Relations

Thicknesses are provided on the inner wall, back wall and outer wall of the actuator to support the flow of magnetic current in the stator iron. The permanent magnets in the case of the HMTB actuator will be attached to the inner and outer walls of the iron core to their thickness and length as shown in Figure 2.11. Figure 2.11 shows that the inner radius of the bearing can be expressed as

### Magnetic Circuit Theory

Magnetomotive Force (MMF): The force that drives the spread of the magnetic flux in the magnetic circuit is called the magnetomotive force. The one-dimensional path of the magnetic flux flow in a closed loop is called the magnetic circuit. Due to the electric current flowing in the coil, a magnetic flux flows in the magnetic circuit.

## Optimization Model for the Design of Actuators of Magnetic Bearings

*Objective functions of actuators**Power-loss**Weight of the bearing**The design vector of Actuators**Design constraints of Actuators**Load to be supported**Current density supplied in the coil**Magnetic flux-density in the stator-iron**Maximum power-loss allowed in the coil**Space available for the bearing**Influence of different parameters on objectives and constraints and constraints*

In addition, the power loss at the operating point (or at the steady state) of the bearing is determined. From equation (2.4), the area of the air gap at poles depends on the inner radii of the bearing and the coil. The volume of the coil affects the outer radius of the coil and the height of the coil in the design vector.

## Conclusion

The design optimization tool is one of the key elements to achieve a desired solution of the problem. The choice of optimization tool depends on many factors such as nature of the objective function(s), simplicity of implementation, reliability of the solutions obtained by a tool, ability to find a global optimum, etc. In the last chapter, the design optimization problem for the actuator of magnetic bearings is detailed.

## The Choice of the optimization tool

### Deterministic optimization methods

The condition that conveys the exact status of the solution point is called the sufficient condition. The necessary condition for the Newton's method is that the gradient vector of the objective function with respect to the design vector must be zero. Two of the popular methods are DFP (Davidon-Fletcher-Powell) and BFGS (Broyden-Fletcher-Goldfarb-Shanno) methods (Rao, 1996).

### Stochastic optimization methods

The next solution update is found based on the solution obtained by the quadratic subproblem. In the deterministic methods mentioned above, the next update of the solution must be found by updating the Hessian matrix. Some stochastic local search methods are stochastic hill climbing (HC), tabu search (TS) and simulated annealing (SA) (Deb, 2001).

Genetic algorithms as the optimization tool

## Details of Genetic Algorithms Implemented

*The general description of GA procedure**Chromosome (Representation of a solution)**Generating the initial population**Evaluation of**Ranking and sorting of the population**The selection operator**The crossover operator**Simulated binary crossover (SBX)**The mutation operator**Elitism operator**Key issues of convergence in SOGA*

For an individual, the sum of the dominance rank and the dominance number equals the population size. Fitness values are assigned, and the population is ranked based on the rank of individuals in the population. Generating the initial population involves generating the chromosomes of all individuals in the population.

## Conclusion

The problem of design optimization of actuators of AMTB and HMTB using MOGA will be described in Chapters 5 to 7. The implementation of MOGAs for design optimization of integrated systems of double-acting hybrid actuators, controllers and power amplifiers will be described in Chapters 8 and 9. The design optimization problem for magnetic thrust bearing actuators and the genetic algorithm as an optimization tool have been described in detail in Chapters 2 and 3, in which, geometries of the magnetic thrust bearing actuator, various objectives to be weighed, design vector and constraints have been explained (see Sections 2.2 and 2.3 ).

## Numerical Simulations

### Input Variables

Design variables for actuator optimization in a magnetic thrust bearing include coil dimensions as discussed in Section 2.3.2. The main limitations include the maximum allowable current density that must be supplied to the coil corresponding to a given load, the inner radius of the bearing and the air gap. Actuator input parameters: In the actual work, the inner radius of the bearing is 25 mm, and the operating gap is 4 mm with the operating load of 2025 N.

### Implementation of the algorithm

The optimization flowchart: A flowchart of the actuator analysis module in the GA implementation is shown in Figure 4.2 and briefly described here. Determine the pole area of the air gap, the volume of the coil, and the cross-sectional area of the coil. The outer radius and height of the bearing are then determined and the restriction values are checked for violations.

### Optimized Geometries of the Bearing

*Convergence issues**Comparison of results*

The second row of Table 4.8 shows the changes in values for AMTB and HMTB in rows 3 and 4 of Table 4.6 for the case of minimizing power loss. In the case of minimizing power loss (ie, the second row of Table 4.7), there is not much. Furthermore, rows 3 and 4 of Table 4.6 show that it is the maximum coil volume chosen to minimize power loss.

## Conclusions

This will be discussed separately for the AMTB and HMTB actuators in the following chapters. The multi-objective optimization of AMTB actuators will be discussed in Chapter 5 and that of HMTB in Chapter 6.

## Introduction

The inner radius, outer radius and height of the coil were considered as design variables. A new analysis methodology for Pareto-optimal designs of the final population has been introduced and is called Pareto-optimal design analysis methodology. The magnetic thrust bearing geometries and optimization model are discussed in Section 2.3 of Chapter 2.

## Pareto optimal front

Point 'U' (with objective values smaller by small quantities (ε ) than those of ideal point . 'I'), is called the utopian point. Point 'N' (where maximum values of the Pareto front ABC intersect) is called the Nadir point. Point 'W' which involves the intersection of maximum (or worst possible) feasible solutions in both objective functions of a minimization problem is called the worst point.

## Multi-objective optimization problem formulation for AMTB

*Fundamental relations**Objective Functions**Choice of the Design Vector**Constraints**Multi-objective optimization tool**Crowding selection operator*

The standard form of the coil power loss expression is given as From the thrust magnetic bearing geometry in Figure 2.5 in Chapter 2, the cross-sectional area and coil volume are expressed as. It has also been shown that for this case the other stator iron dimensions can be determined from the coil dimensions.

## Numerical simulations

### Input Variables

Actuator input parameters: As explained in Section 2.3.2, design variables for optimizing different elements of the actuator of AMTB include dimensions of the coil. Typical input parameters to be provided by the user for specific application of the actuator of AMTB will be [ , ,rlig F B, sat,αmax,αmin,Jsat, ,η γ γs, c,K K ri, a, omax,himax ,Vmax ,Pmax]. GA input parameters: For the optimization, SBX parameters, namely the crossover probability pc, the mutation probability pm, the crossover distribution index ηc, and the mutation distribution index ηm are assumed to be 0.9, 1 /n, 5, and 10, respectively ( Manos and Poladian, 2005); where n is the number of real variables, and n= 3 for the current problem.

### Optimized Geometries of the Bearing

*Comparison of Results from MOGAs and SOGAs**A Typical Choice of an Optimum Design*

The convergence of minimum weight and minimum power loss with production is shown in Figure 5.4. Compared to the case of the lowest weight, the inner wall thickness and the area of the air gap poles are increased. A typical choice of the optimal design closest to the utopian point in the Pareto front of the finite population is shown for comparison in the third row of Table 5.4.

## Sensitivity Analysis

It can be observed from cases 1 to 3 in Table 5.6 that all three design variables affect both objective functions, but the effect of the outer radius of the coil on the objective functions is almost twice that of the inner radius of the coil or the height of the coil. It can be observed from case 3 in table 5.6 that the height of the coil has no influence on dependent variables, but it only depends on the cross-sectional area of the coil. From cases 1, 4, 6 and 7 in Table 5.6; the magnetic flux density could be observed to be the function of only the inner radius of the coil.

## Pareto Optimal Design Analysis on the Final Population

Beyond point B, the area of the coil becomes constant with the increase in power loss as the current density in the coil becomes saturated. Graphs of coil height and coil thickness versus power loss are shown in Figure 5.10(f). After point B, the height of the coil decreases, while the thickness of the coil increases with further increase in the power loss as in Figure 5.10(f).