Abstract
In this paper, the design criteria, assembly practices and experimental analysis of a low cost micro generation system are presented. A high-power density microturbine is specially devised using a stock automotive turbocharger. The project concept is evolved to achieve satisfactory operating conditions. The main practical challenges to achieve the desirable output performance are discussed from the electromechanical perspective including an extensive vibration analysis during operation. The balancing procedure is proven to be a keystone requirement for safe operation at higher rotational speeds. Characterizing the electromechanical interaction between the components is a crucial requirement in the design process. The microturbine is tested using steady inlet compressed air flow, with accelerometers installed on each bearing. The described design and assembly procedures allow the unloaded microturbine to run safely up to 100000 rpm.
Introduction
Distributed power generation systems play an important role in supplying energy to remote areas [1–3]. Historically, distributed stations have been generally based on internal combustion engines (ICE) mainly due to their high power density and availability [4]. Nevertheless, these systems present stricter maintenance requirements, are fuel sensitive and can have a negative environmental impact. They are being phased out in favour of greener and more efficient alternatives.
Distribution systems dependent on Renewable Energy Sources, namely wind and photovoltaic solar power generation, present advantages in relation to ICE systems in terms of environmental sustainability. However, the intermittency of these renewable systems means its application must be associated to an Energy Storage Systems (ESS), such as batteries, Compressed Air Energy Storage (CAES) or flywheels [5 6].
The potential of microturbines for small scale power generation systems has been an important topic of research. Microturbines are currently available in a wide range of power ratings (10–400 kW) at operating speed ranges up to 240,000 rpm [7–12].
By offering high power density and low maintenance operation costs, these machines have been proposed as solution for distributed generation. Furthermore, microturbines can run using multiple fuels, such as natural gas, diesel, landfill gas, ethanol, industrial off-gases and other bio-based liquids and gases for the energy generation [2,7,12–15]. Their versatility is a major advantage in providing energy for remote areas.
The available literature on technology advancements in high-speed turbines operation shows a tendency to select Induction Machines (IM’s) and Permanent-Magnet Synchronous Machines (PMSM’s) [16–18]. PMSM’s have a superior performance from the electromagnetic point of view, presenting higher power density and efficiency [16,19–21]. The drawbacks are increased material cost and attention needed to avoid irreversible demagnetization when operating temperatures surpasses Curie temperature. IM’s are widely adopted in industry application due to its robust rotor structure, low maintenance and field weakening capabilities at lower costs. Main disadvantages are lower efficiency and power density [16,20,22,23]. As summarized by Arkkio et al. [24], electrical characteristics are in favour of PMSM’s, whereas IM’s are mechanically superior.
Two important mechanical properties to be considered in the preliminary design of high-speed microturbines are the mechanical torque and the temperature. The transmitted torque is proportional to the rotor diameter and length. Increasing the rotor diameter leads to higher peripheral speeds, which increase the material mechanical stresses. Increasing the rotor length leads to lower natural frequency, which can increase the vibration levels at the operating speed range. Rotordynamic analysis of the microturbine rotor is crucial to estimate the critical speeds and to investigate the coupling among the vibration modes. In supercritical regime, lateral and axial vibration coupled modes can lead to severe vibration and noise conditions [25]. Another issue common of high-speed operation is the increasing temperature of several microturbine electromechanical components, demanding special cooling procedures, and materials capable of withstanding high temperature cycles [26].
These technological difficulties usually limit the experimental studies of the complete setup, i.e. an electrical generator coupled with overhung turbine or compressor wheel. The complete system is subjected to axial and fluid dynamics loads, fluid film bearing instabilities and thermal loads. The entire system is normally presented at no-load or light load tests conditions [27–29]. Gerada et al. [30] successfully managed to test electrical generator at full load using a high speed dynamometer. However, the integration of the complete system constitutes a complex task involving electrical, mechanical and thermal design parameters and requirements [9]. Accounting for some of the published works performed on microturbines, Kolondzovski et al. [31] discuss the challenges associated with the electromechanical coupling of the compressor and the high speed electrical machine. Mazzoni et al. [32] present a thorough thermodynamics modelling for a micro gas turbine used in a Concentrated Solar Power plant (CSP). Tenconi et al. [16] and Gerada et al. [33] present surveys of electrical machines operating at high speeds, however generation behavior and driving difficulties are slightly discussed. De Moura et al. [34] and Zhang [35] studied the vibrational origin and response of wind and PHES systems, respectively. However, there are few studies regarding the vibrational response of turbochargers adapted as microturbine generation systems. In general, because that technology is mainly hold by industries.
The present work discusses some electromechanical design aspects of a microturbine based on a low cost OEM commercial automotive turbocharger. There are some works available in the technical literature that present high-speed electrical motors are presented, such as the motor developed by Noguchi et al. [36,37] and by Gerada et al. [30]. However, the problems associated with the coupled electro-mechanical design and assembly of a complete microgenerator based on automotive turbochargers are still in need of further analysis.The use of commercial turbocharger can lower the costs of the volute, propellers, bearings and other mechanical components.
In this paper, commercial turbocharger parts are used to design three different microturbine prototypes with a specially-designed generator. The purpose is to evaluate different assemblies that can be rearranged with commercial turbochargers. One of the prototypes is able to run at speeds up to 100,000 rpm under no-load and up to 70,000 rpm at a three-phase load of 3.5 kW. In order to understand important aspects of the dynamic analysis and design of microturbine components, this paper analyzes and discusses the designed microturbines. Mainly the high-speed balancing procedures and the manufacturing tolerances for its electromechanical components.
General description of microturbines
Microturbines are usually used as auxiliary power systems. When operating in a combined heat and power system, can present overall efficiency levels above 70% [38,39]. The thermodynamic cycle for energy generation in a microturbine starts by increasing the air pressure, and consequently temperature, in the compressor. The compressed air is then mixed with fuel in combustion chamber, where it is ignited. After burning, the combustion gases are exhausted to drive the turbine stage, which in turn, powers both the compressor and the generator through a single common shaft. In some cases, a heat exchanger can be used, transferring heat from the turbine exhaust gases to the compressed air, increasing thermal efficiency [3,8,9,14]. A typical configuration of a distributed power generation based in microturbine is presented in the Fig. 1. With a common design based on a single-shaft system, the operating speed ranges from 60,000 up to 250,000 rpm. Electrical machine can also be placed over compressor inlet, to promote cooling and avoid thermal flow from turbine side.
Basic configuration of a microturbine generation system.
Within the alternatives for distributed micro power stations, microturbines have not yet reached commercial success, mainly due to relative high manufacturing cost and great technical complexity. Due the lack of commercial models available, recent studies have been conducted aiming at the development of efficient and cheap alternatives. Ennil [40] designed, simulated and thoroughly tested a small scale 3D printed low temperature axial air turbine, with satisfactory results, reaching efficiency values as high as 82%. However, it is only a preliminary study, and it is not ready to be commercialized.
Maia [8] analyzed the use of an automotive turbocharger in a small scale CAES system, whose design concept is also applied in the prototype developed in this work. Katsanos et al. [41] performed a simulation on the effect of replacing the conventional turbocharger setup, in which the compressor is driven by the turbine, to generate energy in an electric generator, achieving a net power output up to 62 kW. Vásques et al. [42] analyzed an hydrokinetic turbine under mechanical and electric perspectives, modelling not only the turbine itself but the electrical behaviour under different loads and control systems. Gröman et al. [43] performed a thorough analysis comprising of electrical and mechanical design, contructive procedures and experimental analysis of a two-stage intercolled electrically assisted turbocharger. The scope of this study is similar to what is hereby presented, however in a larger scale, utilizing a 1,200 kW diesel engine. The results showed an increase in efficiency and pressure ratios.
Automotive turbochargers are used in internal combustion engines to increase the engine power density and volumetric efficiency. A wide range of available models, along with the possibility of matching different compressor and turbine elements, offers a wide range of possible and flexible solutions. There is a great potential for developing cheaper and accessible microgeneration systems by adapting a commercial automotive turbocharger to a generator unit.
Some basic concepts about the electromechanical components and the configuration design of a small microturbine are presented in an attempt of contributing to the technical literature on microturbine design for power generation.
The proposed microturbine has been designed from a commercial turbocharger system in which the compressor is directly coupled to the turbine. To operate as a generation system, its power output is shared between the compressor and an electric generator.
Shaft
Commercial microturbines usually are designed with rigid shafts which render high bending and torsional eigenfrequencies. Generally the operational rotating speeds are below the shaft first critical speed. However, rigid shafts imply larger cross-sectional area and consequently higher bearing and windage losses due to the larger tangential speeds. While bearing losses depend linearly on the shaft diameter, the windage losses present fourth-order dependence on this diameter [44]. Larger cross-sectional area also renders higher centrifugal forces requiring a thicker sleeve for the magnets. That increases the air gap and thus, lowers the power density.
On the other hand, flexible shafts can use smaller permanent magnets, resulting in lighter rotating systems and higher power densities [16,21,45,46]. Flexible shafts are built with smaller cross-sectional areas that cause smaller tangential speeds, which could avoid the use of retaining sleeve. Consequently, electrical output could be increased in function of the smaller air-gap. However, flexible shafts usually operate at speeds above the first, or even the second critical speed.
Bearings
Bearings must support the dynamic loads and provide stable operation conditions. The literature of high-speed micro machines indicates that foil bearings, magnetic bearings, and ceramic ball bearings [25] offer good supporting system designs. These bearing types present low-cost maintenance [16,31,33,47–50]. Foil bearings usually presents an excellent performance, attenuating vibratory responses at high speeds and safely operating at higher temperatures. However, at higher rotational speeds and load conditions, friction losses in foil bearings increase. In these cases, ceramic ball bearings become a possible alternative. They possess excellent resistance to high temperatures, but lack the capability of attenuating shaft vibrations due to high stiffness and low damping coefficient. Finally, magnetic bearings can provide stable operating conditions, minimize friction losses and heat generation over a wide range of rotating speeds, but their high costs restrict their application.
These bearing types also introduce a number of design complexities to the supporting system. Foil bearings, for instance, demand high resistant material and complex foil geometry [51]. Moreover, magnetic bearings rely on complex power electronics control system [52] whilst ceramic ball bearings require an external lubricant recirculating system for cooling.
Commercial turbochargers generally utilize fluid film bearings, with three types being noteworthy due to their wide application and mechanical properties: fully floating ring journal bearings, the semi-floating ring journal bearings and the ball journal bearings with squeeze film damper. Among them, floating ring bearings present the largest damping coefficients, smallest lubricant viscous dissipation, largest carrying-load capacity and the lowest design cost [53–56]. The high damping coefficients are very important to lower vibrations.
During the operation of high speed microturbines, the pressure differential between the turbine and compressor causes an axial load in the shaft. Floating ring journal bearings are only capable of supporting lateral loads, and consequently thrust bearings must be employed to support the axial loads on the microturbine shaft. This thrust bearing consists of a thrust sleeve (or collar) attached to the shaft and a collar fixed to the bearing housing.
Design of the electrical generator
The design of an electrical generator for high speed (HS) rotating systems is a multidisciplinary task. There is not an universally appropriate machine topology and the design guidelines are still subject to further investigation [57]. However, some basic design requirements serve as guidelines to choose a certain type of machine. Load, speed, power, size, operational environment, cooling system, cost and manufacturing complexity are some keystone requirements in the selection process of the most adequate electrical machine.
The strong interaction between the electromagnetic and mechanical behavior is one of the main challenges designing a HS electrical machine. The limitations are usually defined in terms of maximum dimension, power or torque ratings. For instance, as discussed in [16,58], the developed torque is proportional to the rotor volume. The shaft dimensions must be carefully estimated to avoid stress levels close to the allowable limits of the material strength. In order to ensure the machine safety and integrity, a retaining sleeve may be used, which also increases the effective air gap. Also, the sleeve material should be carefully selected, because it affects strongly the machine performance [59].
In the design of electrical machines usually the rotor diameter can not exceed a limit established by the design requirements. On the other hand, an increase in the torque output can be achieved by extending the shaft length. Nevertheless, longer rotors pose another technical challenge, since the shaft critical speeds decrease. Many authors recommend the operating speed below the first critical speed or between the two first critical speeds [16,26,57]. The thermal design requirements become very important in long rotors, since the electromagnetic losses increase as the rotor length increases. Temperature can be a limiting design factor for generators operating above 20,000 rpm [18].
The selection of materials must comply with both mechanical and thermal requirements. PM’s (permanent magnet) electrical machines in high speed conditions are usually manufactured with high-energy-density magnets from neodymium-iron-boron (NdFeB) or samarium-cobalt (SmCo). Sintered NdFeB have highest levels of magnetic energy, allowing better power density with smaller rotor diameter and lower peripheral speeds. However, its temperature operating range is limited below 250 °C and it is sensitive to oxidation and corrosion without coating [60]. For operation at higher temperatures, SmCo magnets remains the only commercially available material, but with lower remanence and energy, requires a bulkier assembly [33]. Nevertheless, both magnets require banding to prevent damage from centrifugal stress [61].
For the stator design, a distributed winding design is preferable since it provides air-gap flux density with less harmonic distortion, leading to lower torque ripple and lower rotor losses. As discussed in [62] the distributed wound design reduces losses in the rotor and also improves thermal path for the stator copper losses. But this may lead to very thin and mechanically weak stator teeth [33]. Slotless configuration provides lower torque ripple, but it increases the effective air-gap, resulting in reduced electromagnetic performance. Concentrated winding may offer advantages in terms of lightness, smaller losses due to reduced end winding length and increased slot filling factor, but they are characterized with higher torque oscillations [63].
Eddy current and skin effect losses can be reduced using Litz wire [26]. However, designed for high frequency operation, Litz wire reduces the slot filling factor and increases the machine size as consequence [64,65]. Stator core losses can be addressed with low flux density, therefore enlarging the machine volume, or with special core materials. For the lamination of the stator and the rotor, silicon-iron (SiFe) and cobalt-iron (CoFe) alloys can be considered, with the former providing higher saturation magnetization, whilst being significantly more expensive [33]. In terms of frequency and flux density, thinner SiFe lamination, amorphous alloys, nanocrystalline materials, and soft magnetic composite (SMC) are more adequate to reduce stator losses [64].
Magnetic Forces in a HSPM motor
One important advantage of high speed permanent magnet (HSPM) motor drives is the high torque density. From the mechanical design standpoint, the effects of the torque ripples should be evaluated in order to prevent high levels of noise and vibration, which can ultimately lead to premature fatigue failure. Also PM motors are subjected to two types of forces: tangential and axial magnetic force. The first is vastly investigated while the later has only been assessed in a few special cases.
Two kinds of torque ripple are present in PM synchronous machines - cogging and electromagnetic (or pulsating) torque ripple. The cogging torque is only associated with PM’s and stator slot reluctance, and it can be characterized by the trend of the alignment of the rotor in a position which maximizes the total magnetic permanence. There are several design propositions to minimize this torque ripple component, e.g. slot or magnetic skewing, magnetic segmentation or simply utilizing a slotless configuration [20,66].
Conversely, electromagnetic torque ripple is generated from non-sinusoidal magnetic flux distribution in the air-gap. The flux linkage harmonics are produced from non-ideal stator slot and winding distribution. Armature current harmonics are generated because of non-linear drive system characteristics, mainly due to the power electronics components. Some works propose current injection techniques to minimize torque ripple based on back electromotive force waveform [66,67].
It is uncommon for technical literature to address the occurrence of the axial magnetic force in HS electrical machines. This force can be caused by flux leakage, especially due to rotor and stator center misalignment, resulting in thrust forces [68]. Some works discuss these axial forces caused by the unbalanced motor system, such as skewing or unconventional PM shape [69,70]. From the rotordynamics viewpoint, an axial electromagnetic force can induce radial oscillations, similarly to a mass unbalance. At low operating speed range, the axial electromagnetic forces are not relevant in large inertia motors. However, for high-speed motors, large centrifugal forces can lead to high levels of vibration and noise.
Electromechanical analysis of the microturbine prototype
Thermal Analysis
The cooling of the components of the canned turbogenerator is performed by a forced oil injection system. There are injection holes on the generator casing to mount the connectors that allow the installation of an oil cooling circuit, which is fed by an external 45 to 80 psi pump and a refrigeration unit. The inlet and outlet temperatures of the cooling oil are monitored by thermocouples and the cooling power is set accordingly by an electronic temperature control system. Basically, the convective forced oil flow is expected to carry out mostly of the heat generated by the rotor-stator electromagnetic losses, other mechanical losses and some of the heat conducted from the turbine volute.
For the rotating components, the same oil injection system is used for lubrication and cooling. A simple thermal analysis is performed using the commercial finite element package Nastran to estimate the temperature field on the rotor shaft, which receives the thermal load of the high temperature gas inside the turbine casing. The boundary conditions of this thermal analysis are the oil temperatures in the cooling circuit and the gas temperatures of the flow through the turbine. The temperature field is predicted for the shaft at steady-state operating condition, which means that the rotating speed is constant and the gas and oil flows are assumed steady-state. The thermal conductivity of the turbine rotor, made of Inconel, is considered equal to 10 Wm−1 K−1 [71,72]. The PM coupling faces are cooled by an oil with convective heat transfer coefficient of 30 Wm−2 K−1. The inlet oil temperature and the ambient temperature are assumed constant and equal to 343 K and 313 K, respectively. The gas average temperatures at the turbine inlet and outlet are approximately equal to 1050 K and 850 K, respectively. Figure 3 depicts a photo of the micro-turbogenerator, indicating the three cooling oil feed connectors on the machine casing.
Figure 2 shows the temperature fields predicted by the finite element thermal analysis on the shaft and on the turbine wheel of the micro-turbogenerator. The simulation uses tetrahedral finite elements in the turbine wheel with coupled transition to hexaedral finite elements in the shaft and PM sleeve, resulting in a total of 350000 elements. Further refinement was done in the turbine wheel due to the aspect ratio of its blades. A maximum skewness factor of 0.8 and growth ratio of 1.25 are used as mesh quality control. The temperatures are shown in Celsius degrees. The maximum temperature estimated by the thermal analysis occurs at the inner radius of the turbine blades. The maximum temperature of the shaft surface at which the rotor is mounted is predicted below 69 °C, showing that the heat from the turbine is absorbed efficiently by the oil flow. Experimental evaluation in the same conditions are yet to be done.
Numerical solution to the shaft and rotors steady state thermal analysis.
Cooling ports connections highlighted.
The prototype developed in this work was designed to deliver 10 kW at 100,000 rpm. The literature related to high speed machines suggests that such operating range reduces the available options of electrical machines [9,33]. Gerada et al. [33] evaluate a figure of merit, namely the rpm
The electrical machine design encompasses different concepts and topologies, in order to meet the requirements of several machine configurations. Since the objective is to design a high speed electrical machine, the vibration analysis must be a priority to verify the machine safety and durability. The interaction between the mechanical and electrical sources of excitation is the starting point.
The electrical generator topology is based on a Radial Permanent Magnet Synchronous Machine. The NdFeB magnets are the best solution considering their high energy density [45]. Since these materials are manufactured by sintering, there is more flexibility in the magnet geometry.
The material selected for the high speed rotor must have good mechanical strength to support the centrifugal inertial forces. The constitutive and geometric characteristics of the rotor play a very important role in the safe operation of the electrical machine. In the design, ASTM 4340 steel and the NdFeB sintered magnets are selected to be used in the design of the rotor hub and the permanent magnet, respectively. Table 1 depicts some properties of the rotor. Using the finite element package NASTRAN, a linear stress analysis of the machine rotor is performed to investigate the more adequate geometry for the magnet and hub. The stresses are estimated for the rotating system using the default tetrahedral mesh offered by the linear stress analysis solver.
Rotor properties
Rotor properties
Prototype operational results
The surface PM machine topology was adopted to avoid the high levels of stress concentration, and a ring shaped parallel magnetized magnet, with a retaining sleeve, is used as illustrated in Fig. 4. This resulted in a pure sine wave flux density being generated in the air gap.
Final model of total flux density inside the electrical generator.
The manufacture of sintered magnets offers more geometric flexibility than other manufacturing processes do. But, in order to attain an acceptable metrological quality, sintered components usually need posterior machining processes to adjust their final geometry. Commercial precision manufacturing machines generally employ ferromagnetic materials and consequently are not recommended to finish sintered components. A procedure to overcome this problem is to demagnetize the sintered magnet by increasing its temperature above the Curie point before the machining process and, afterwards, to remagnetize it. However, the neodymium iron-boron alloy is highly susceptible to oxidation and is flammable, making the machining process even more complicated.
The use of an external stainless-steel sleeve can help to assemble the magnets and also to pre-stress them to improve their mechanical properties and integrity [33]. During operation, it would also work as a damping winding, attenuating vibrations coming from the cogging torque or any other source of harmonics. The problems associated with an external sleeve can be the air-gap increase, some extra losses and rotor overheating due to eddy currents. Also, machining a stainless steel can lead to an austenite(𝛾)-ferrite(𝛼) transformation, which induces ferromagnetic properties and reduces the effective gap and increase the mean density flux. The small thickness of the sleeve imposes saturation from the sides, reducing the leakage flux [73], as shown in Fig. 4.
A two-pole rotor was designed to run up to 100,000 rpm, which leads to 1.67 kHz electrical frequency. Electrical steel performance curves however are commonly build up to 400 Hz, so it must be evaluated at higher frequencies in order to identify magnetization behavior and losses. In this scenario, three main core materials are selected:
Fe-3%Si (0.5 mm) - Common and commercial available core material. Hoganas SMC (Sintered) - Promising material to work at higher speeds for loss reduction [74]. Aperam Steel (0.35 mm) - Special steel developed for high speed electrical machines for hybrid vehicles.
Two important criteria for the selection of the core materials are their commercial availability in Brazilian market and their cost. The Fe-3%Si (0.5 mm) is an ordinary and widely used material, been considered in this work as a low-cost option. The Aperam Steel (0.35 mm) is found in the Brazilian market as a special material for high efficiency electrical machines, and is considered a candidate for applications at high speed range. Hoganas SMC is also considered for its advantages [9]. Other materials are adequate for selection, e.g. the 0.1 mm JNEX-Core steel with 6.5 % silicon content [30], which consists of a high cost option.
For design purposes, a standard magnetization curve has been employed [75] as electrical steel material. Fig. 4 shows some numerical predictions of the flux density inside the electrical generator. The current geometry is the result of successive analysis aiming at higher flux-linkage using optimal slot geometry shape, including fully-open or semi-open slot configuration.
The simulation target is to generate a 220 V potential, which was obtained by winding 18 coils and using a 100 mm length stator. Also, at the rated frequency the skin effect and the copper eddy current losses become critical, requiring more parallel winding. The final prototype was built with 26 parallel circuits using 28 AWG copper wire (∅0.32 mm).
The electromagnetic performance only evaluates the agreement between the simulated and tested results for voltage. Different operation points are experimentally evaluated to build curves of voltage versus time and rotating speed. Figure 5 shows the comparative curves of experimental and numerical voltage in the generator, whilst Fig. 6 depicts the curves of induced voltage in frequency rendered experimentally and numerically. The induced voltage is not always constant for a high speed machine since its core loss increases with frequency and some magnetic skin effect starts to appear in the system [76]. The Aperam steel was used in the test.
Fundamental experimental and simulated induced voltage in the generator.
Curve representing constant induced voltage in frequency for the generator.
The overall analysis of electrical generator design, including the comparison of the analytic-numerical predictions with experimental data, is presented more minutely [8,9]. These references present the measurements and predict the values of no-load induced voltage, evaluate experimentally the load voltage and current, estimate experimentally the turbine efficiency and the overall efficiency, and evaluate other characteristics. Some technical challenges involved in the development of high speed micro-turbogenerators are also discussed in these references.
At high-speed operating conditions, the rotating machine is usually subjected to large mass unbalance and electromagnetic forces, and there are different sources of dynamic instability associated with the fluid film bearings and the turbine working fluid. The strong interaction between mechanical and electrical sources of excitation makes the microturbine modeling a hard task. In order to investigate the microturbine behavior, three different compressor-turbine assembly configurations were manufactured for testing. During the test of some prototypes, dynamic instabilities caused the failure of the turbomachine at high loads. The electro-mechanical problems observed in these failures provided important insights into the design practice and procedures. As discussed by Lim et al. [77], electrical machines for high speed application requires vast knowledge and experience in design complexities related to the electro-mechanical system. Therefore, the guidelines presented here are intended to illustrate the design experience acquired during the development of high-speed microgenerators.
Three microturbine prototypes general layouts were specially devised for this investigation. The first two arrangement employed the original turbocharger mounted with an external generator, whose shaft was supported by rolling bearings. Figures 7a and 7b show the two microturbine configurations mounted with external generator. In the first one, the generator was mounted with a single bearing, whilst the second assembly used two bearings for improved stability. In the third configuration, shown in Fig. 7c, the generator is installed between the compressor and turbine, being supported by fluid film bearings. This last configuration presented the best performance, eliminating the need of shaft coupling and reducing misalignment excitations.
The first microturbine configuration (Fig. 7a) employed the original turbocharger coupled directly to the generator shaft. This configuration is called “single bearing solution”. A rolling journal bearing was installed with press fit on the generator shaft. This prototype was able to run up to 20,000 rpm, but could not pass this speed due to the high vibration levels. The direct coupling between the turbocharger and the generator shafts is very rigid and poses technical difficulties to attain good concentricity tolerances. Besides, extending the turbocharger shaft resulted in synchronous unbalanced load.
To overcome the vibration problems associated with rigid coupling between the turbocharger and the external generator, a flexible mechanical jaw coupling was designed to operate at high speeds, as presented in Fig. 8. This flexible coupling improved the microturbine assembly dynamic response, but required two rolling journal bearings at the generator shaft ends. This configuration was hence called “double bearing solution” (Fig 7b), and was able to reach rotational speeds up to 40,000 rpm. To further improve dynamic response, mechanical parts of the “double bearing solution” configuration were remanufactured with stricter tolerances, and a mass balancing was performed. With these actions, the modified prototype was able to run up to 50,000 rpm. The attempts of improving the behavior of the second microturbine configuration indicate that the dynamic response of high speed machines is very sensitive to the machining process quality and to the shaft coupling.
Evolution of the mechanical assembly solutions.
The third microturbine assembly was designed with the generator internally mounted. This configuration was named “internal journal bearing” (Fig. 7c) and due to its better dynamic performance, it has been chosen for the analysis presented in the following section. The compressor, generator and turbine are mounted inline on a single common shaft. The generator rolling bearings have been replaced by oil film journal bearings. This microturbine with internal generator is presented in Fig. 9. The mass balancing was performed at rated speed with the mounted bearings. After this mass balancing, the prototype could run up to 100,000 rpm, but the bearings and the balancing procedures must be carefully analyzed. Furthermore, placing the electrical stator between the bearings can lead to thermal problems. The high temperature from the turbine rotor cancause high heat flux directly into the shaft. The permanent magnet must be actively cooled below 120 °C, otherwise it will demagnetize. Thus, a pump has been emplyed to force the flow of cooling oil through the bearings, the stator and the permanent magnet.
View presenting the coupling from the generator and turbine.
External (left) and internal (right) generator prototypes mounted on the test bench.
Figure 10 shows a photograph of the microturbine prototype with the internal generator. Details of the microturbine internal rotary components, including rotor, seals and bearings are depicted in Fig. 11. A preliminary analysis of the vibration modes of this prototype is performed using a commercial finite element package. In the solution of the eigenvalue problem, the mass distribution associated with the mounted rotating components has been accounted for [78]. The three first natural vibration modes of the microturbine shaft, with the respective values of natural frequencies, are shown in Fig. 12.
Finished prototype ready for balancing.
Rotors, shaft and bearings distribution for test.
Mode shape for the first three natural vibration modes for the shaft, with their respective frequencies.
An experimental vibration analysis of the microturbine prototype was performed using a High Speed Vibration Measuring Machine (HSVMM), in which the vibratory response was measured at rotating speeds ranging from 30% to 110% of the nominal speed (100,000 rpm). The turbine was driven by compressed air and the vibrations measured with a piezoelectric accelerometer, mounted on the turbine bearing. The mass correction was performed on the compressor head nut. Only the synchronous whirling response was measured in the vibration tests.
In order to assess the influence of each component on the dynamic behaviour of the microgenerator, some vibration tests were performed. Localized discrete masses and magnetic fluxes were added to the machine for all configurations, starting with a simplified rotor, without magnetic components until and ending with the complete system. The vibratory responses of these configurations were captured response throughout the operation range. By analyzing the vibration signals from the simplest to the more complex machine configurations, it was possible to evaluate the influence of each variation in the machine components on the overall vibrational performance.
From all configurations experimentally evaluated, the following study cases were chosen for further analysis:
Case A - Compressor and turbine rotors with stock low-speed balancing; Case B - Assembly with high-speed balanced compressor and turbine rotors; Case C - Compressor and turbine rotors with some residual mass imbalance; Case D - Assembly of balanced compressor and turbine mounted with a metal sleeve to simulate the magnets mass; Case E - Compressor, turbine, magnet rotor and stator, with different machining techniques and magnet geometry; Case F - Compressor, turbine, magnet rotor and stator with addition of static preload.
In order to classify the fore-mentioned cases based on the vibrational response, three levels of acceleration magnitude were selected. The first level rotating systems with maximum acceleration under 1.5 g, which was considered adequate. Systems with vibration levels between 1.5 g to 2.5 g are named intermediate result and above 2.5 g require some corrective action for the high vibratory amplitudes. These acceleration levels follow the classification commonly used by industrial vibration standards.
The compressor and turbine wheels are manufactured with cast steel. Even though foundry is a very cost-effective process, the cast parts can present elevated mass eccentricity, especially when compared to components made with more advanced manufacturing processes. Then, an evaluation of the mass balance efficacy, the prototypes were subjected to rigid balancing (5 to 10,000 rpm), whose vibratory response after the mass balancing for case A is shown in Fig. 13. As expected, the low speed balancing is not enough to guarantee successful runs at nominal speed of 100,000 rpm. However, once the balancing at high speed is performed, noticeable improvements in the vibration levels are observed, which can be seen in Fig. 14 for case B after high speed mass balancing.
HSVMM result for case A.
HSVMM result for case B.
Cases C and D were performed to assess the influence of adding the mass of the PM rotor to the balanced system from case B. The shaft, turbine and the compressor wheels were assembled and high speed mass balancing was performed, resulting in case C, whose vibratory response is presented in Fig. 15.
Then a metal sleeve, with the same mass and shape of the PM rotor is added to the configuration of case B, generating the configuration of case D, whose vibratory response can be seen in Fig. 16. The vibratory responses for cases C and D present adequate acceleration levels.The slight variation of the vibration amplitude observed in Fig. 16 can be associated with the the individual dynamic contributions of several machine components with different mass distribution. Nevertheless, the bearing damping is able to attenuate the vibratory response, sometimes masking or eliminating some oscillation mode of the rotating system. The metal sleeve was manufactured by turning and reaming, which are processes with low residual mass imbalance. In this sense, it can be concluded that, as long as the PM rotor is manufactured with similar characteristics, these lower vibrational levels can be attained.
Cases E and F are finally obtained by installing the PM rotor and stator in the previous balanced configuration, which can be the case C. In these cases, the rotating systems start to present electromagnetic cross dependencies with the dynamic response.
The addition of a permanent magnet in the shaft introduces a different source of vibration, generated by a thrust bearing load emerging from the magnetic interaction between the stator and the rotor. This interaction increases at higher rotational speeds, leading to an asymptotic vibration peak (see Fig. 17). Despite many corrective actions to reduce the vibration caused by this magnetic interaction, it was not possible to control this vibration only by mass balancing, as presented in Case E. At this condition, the thrust bearing was so heavily loaded that oil thermal degradation occurred even for short periods of operation in a flow test performed on the machine testing bench stand, and Fig. 18 depicts a photo of the faulty bearing.
HSVMM result for case C.
HSVMM result for case D.
HSVMM result for Case E.
A viable solution to operate the complete assembly with acceptable vibratory response levels was to preload the axial bearing in the opposite direction, resulting in Case F, which is discussed next.
After identifying the axial load caused by the magnetic interaction, a finite element analysis was performed to identify the flux leakage and the load origin in the microturbine. The magnetic flux leaves the generator permanent magnet, goes through the ferromagnetic outer structure, seal and thrust collar to return through the shaft, represented in Fig. 19(a). On the right hand side of Fig. 19(a), a close up shot of the return path is depicted.
Thrust bearing oil thermal decomposition.
The results of the FEM analysis are presented in Fig. 19(b). On the left hand side of Fig. 19(b), it is possible to identify the magnetic flux flowing through the shaft, going through the seal and closing the magnetic circuit on the compressor back-plate. Some smaller secondary leakage is also present in the thrust sleeve left side. This flux leakage, however, has not been identified during the electromagnetic design. Even though its magnitude is much lower than the equivalent air gap, for higher shaft speeds, this small flux leakage can be enough to cause high vibrations and failure.
If the seal material was replaced by aluminium, some leakage would still be present in the thrust sleeve, which can be machined to increase the distances or replaced by Inconel, as shown on the right hand side of Fig. 19(b). An aluminium spacer can be also be installed between the stator core and the casing to avoid further magnetic loops.
A palliative solution is to change the stator axial position to counteract the high speed loading, inducing an opposite static axial load. The effects of this change can be observed in Fig. 20. This figure depicts the vibratory response of the case F, described in the previous item. The final vibration peak is still present but can be reduced in the near future. This solution is considered acceptable for energy generation trials, so the system is mounted in a test bench, with two piezoelectric accelerometers on each housing of the floating ring bearings.
Magnetic flux which origins the axial force. (a) Cross section of the model depicting the magnetic flux (Φ
B
) circuit, flowing from the permanent magnet, going through the seal, thrust collar and returning through the shaft. Left: General flux overview. Right: Detail of the seal, thrust collar and shaft. (b) Cross section simulation from flux density in the thrust bearing position.
A varying speed vibration test is performed on the microturbine (case F) to render the waterfall diagram for the microgenerator. Several frequency spectra are shown in Fig. 21 from zero speed to the nominal rotating speed of 100,000 rpm. In Fig. 21(a), the vibratory response of case F seems to be unsatisfactory, since it can be oberved sub, super and synchronous high-amplitude vibrations for the whole spectrum. In order to improve the vibratory response, the machine underwent a high-speed mass balancing. Then, Fig. 21(b), shows the cascade plot for the case F after the mass balancing, and it can be observed that the synchronous response reaches acceptable vibration levels. Besides, no noticeable supersynchronous vibration is detected, but the expected subsynchronous vibrations, due to oil bearing action is present.
HSVMM result for Case F.
Microturbine Waterfall diagram (Rotation × Vibration frequency × Vibration amplitude) from internal generator system.
In this work, an investigation over the electromechanical behavior of high speed microturbines for power generation is presented. Several experimental analysis were evaluated during the design of the machine components. The final experimental setup assembled uses a permanent magnet synchronous generator. The proposed system is tested with several layouts, from which a more adequate microturbine configuration is selected from extensive vibration tests. The presented test procedure can render acceptable levels of vibrations, which is critical for high speed operation range.
A mass balancing process is performed, in which rotor components are added gradually to the rotating system, and their effects are criteriously evaluated. It has been observed that not only the individual component balancing and the assembly sequential steps influence the vibratory response of high-speed microturbines, but also unforeseen and non-trivial interactions play a major role in the assembled machine dynamic behaviour. An important characteristic observed in the vibration tests of the microturbine prototype is the occurrence of axial vibrations caused by the magnetic leakage.
One of the main challenges in the design of high speed electrical machines is related to the elevated number of design parameters and characteristics involved. variables. Mechanical and electrical properties and characteristics are strongly linked and the design process depends on the interaction of different areas of knowledge. It is important to notice that the design process can be understood from different theoretical perspectives,and the results hereby presented are intended as one to provide basic insights and guidance for the design of high-speed microgenerators to guideline. The analysis performed in this paper can be situated stands in the borderline between the mechanical and electrical engineering, and from this interdisciplinarity stems its importance and relevance.
Footnotes
Acknowledgements
The authors gratefully acknowledge the financial support received from the Pró-Reitoria de Pesquisa da Universidade Federal de Minas Gerais (PRPq-UFMG).