The importance of energy dissipation in damper is recognized most in automotive suspensions, where ride comfort and vehicle handling are encountered. When designing a suspension damper, it is well-known that damper characteristics required for good handling are not the same as those required for good ride comfort. Any choice of damper characteristic is therefore necessarily a compromise between ride comfort and handling. Semi-active suspension damper are those which otherwise passively generated damping forces are modulated according to a parameter tuning policy with only a small amount of control effort. Semi-active suspension damper, as their name implies, fill the gap between purely passive damper and fully active suspension-control systems and offer the reliability of passive systems, yet maintain the versatility and adaptability of fully active devices. During recent years there has been considerable interest towards practical implementation of this semi-active suspension damper for their low energy requirement and cost. For this reason, this paper briefly reviews the basic theoretical concepts for semi-active suspension damper design and implementation, and surveys recent developments and control techniques for these systems. Typical design guidance and future trends of research in the semi-active suspension damper design are also discussed.
Keywords: ride comfort, handling, semi active suspension damper and control techniques
Introduction
Traditionally, suspension systems perform multiple tasks such as maintaining contact between vehicle tires and the road, addressing the stability of the vehicle, and isolating the frame of the vehicle from road-induced vibration and shocks, addressing the ride comfort of the vehicle. Wong [1], states that vehicles exhibiting good ride quality are characterized by suspensions with low damping rates, resulting in large suspension travel. Vehicles exhibiting good drive stability generally incorporate suspensions with high damping rates, resulting in small suspension deflections. This becomes more pronounced when the operational requirements of a vehicle are in conflict with the suspension design. The purpose of this paper is to investigate the suspension requirements for good ride comfort and good handling, respectively. Only the effect of changes in damper characteristics is analysed.
An overview of automotive suspensions
The paper by Kazuoka [2] presents that the refinement of passive damper for modern automobiles has resulted in acceptable levels of both handling and ride quality under common road conditions, these systems are not typically optimized for any particular type of terrain. Passive damper performance characteristics still represent a major compromise between ride quality and handling. The need to reduce the effects of the compromise has been the driving force for advanced suspension systems. Advanced vehicle suspension systems such as adaptive, semi-active, and active have extensively been used in most conventional ground transport fleets. Due to slow response time in adaptive systems and high energy consumption and cost in active suspensions, they are unlikely to survive in the future market [3]. Semi-active configuration addresses these limitations by effectively integrating a tuning control scheme with tunable passive devices. For this, active force generators are replaced by modulated variable compartments such as variable rate damper and stiffness [4, 5]. Much attention is being paid to these arrangements for their low energy requirement and cost. Recent advances in smart materials and adjustable dampers have significantly contributed to applicability of these systems [6-8]. Semi-active suspensions, which can achieve a ride comfort using less energy than active suspensions have been actively studied during the last decade [9], [10] and [11]. Another critical issue with active control is the stability robustness with respect to sensor failure; this problem is a big concern when centralized controllers are used. On the contrary, semi-active control devices are essentially passive devices in which properties (stiffness, damping) can be adjusted. The paper by Preumont [12] presents that semi active systems cannot input energy directly into the system being controlled and thus, they are robustly stable. From the paper by Song and Ahmadian [13], it is clear that the mechanical system presented is stable regardless of the damping tuning approach. One of the main features of the semi-active control systems is that they are fail-safe. This means that if the control system fails for any reason (including power failure and sensor failure), the system acts as a passive system. Thus Semi active suspension systems are becoming more popular because they offer both the reliability of passive systems and the versatility of active systems without imposing heavy power demands.
Adjustable semi-active suspension dampers for Vehicles
Adjustable semi-active dampers are mechanical (in most cases electro mechanical) control devices with the capability to vary the amount of energy they dissipate using a small source of power. Such systems extend the possible range of damping characteristics obtainable from a regular (passive) damper. The variable resistance is achievable in a variety of forms. The currently available semi-active damper technologies can be divided into three main groups.
1. Resistance to the damping fluid (by varying the orifice area of the damper)
2. Magnetic properties of the fluid (Magneto-rheological dampers)
3. Electrical properties of the fluid (Electro-rheological dampers)
The first uses controllable electromagnetic valves. According to paper by Milliken and William [14], Servo-valves can provide very high speeds, linearity, and accuracy of flow control at high operating pressures. Servo-valves have been used successfully in fully active suspensions which use a hydraulic pump, reservoir, accumulators and hydraulic cylinders connecting the wheels to the chassis. Unfortunately, servo-valves are very expensive and complicated devices. Solenoid valves provide an alternative to servo-valves. A solenoid valve does not have as fast or accurate response as a servo-valve; however, a servo-valve is much more expensive than a solenoid valve. The paper by Kim, Gillespie and Eslaminasab [15], reports that a solenoid valve is much simpler in design than a servo-valve, and it could possibly be manufactured in-house, bringing the added benefit of being able to optimize the valve design for the particular application. Solenoid valves have also been used in two- or three-state dampers that can change characteristics between hard damping and soft damping by opening or closing a bypass valve. These dampers have resulted in improvements in ride quality when appropriate control schemes are used. Although these dampers do not have the capability of being continuously variable, continuously variable solenoid valve dampers are also available. Solenoid valves of this type have been tested, with promising results by Kitching, Cole and Cebon [16]. The paper by Park Kim and Kim [17], presents Semi-active dampers with external solenoid valves have been modelled taking into account the flow forces on the valve, but not including variable fluid compressibility. Elmer and Gentle [18], Vaughan and Gamble [19] investigated more complex modelling of hydraulic control valves, including variable coil inductance and valve spool dynamics.
Recent advances in smart materials have led to the development of new semi-active dampers, which are widely used in different applications. In view of these semi-active dampers, the electro-rheological (ER) and magneto-rheological (MR) fluids probably serve as the best potential hardware alternatives for the more conventional variable orifice hydraulic dampers [20,21]. Both the electro-rheological (ER) and magneto-rheological (MR) effect are very fast (response time of less than a millisecond) and fully reversible. Winslow (1947) [22], was the first to patent the magneto-rheological fluid concept. It was not until the 1980's that the industrial applications of magneto-rheological fluids was developed by industry researchers Jazar and Golnaraghi [23]. Peel, Stanway and Bullough [24] demonstrated the applicability of MR fluid in vibration isolation. As a follow up, Jolly, Carlson and Munoz [25], Davis [26] introduced both a mathematical model and a mechanical model by using magneto-rheological fluids. This subject has resulted in an extensive number of publications. Walid [27] reports that magneto-rheological fluids are materials that exhibit a change in rheological properties (elasticity, plasticity or viscosity) with the application of a magnetic field. Typically, embedded electromagnets offer a simple means to alter the magnetic field in these damping devices [28, 29]. Dampers using magneto-rheological fluid as the working fluid are disclosed in various U.S. patents, including patent No.5,277,281 by Carlson and Chrzan [30], patent No.6,131,709 by Jolly, Carlson and Prindle [31]. Choi, Wereley, and Jeon [32], Liu et al [33] have developed semi-active isolators using magneto-rheological dampers. Anon [34] developed magneto-rheological fluid based adjustable shock absorbers for stock car and drag race vehicles were introduced in 1999. A recent and well-received model is that developed by Spencer et al [35], this model has used the Bouc-Wen model to analyze the nonlinear hysteresis behaviour of an MR damper.
Electro-rheological fluids exhibit rheological changes when an electric field is applied to the fluid. However, there are many drawbacks to electro-rheological fluids, including relatively small rheological changes and extreme property changes with temperature. The main difference between both fluids is that magneto-rheological fluids require large currents but low voltages, whereas electro-rheological fluids require low currents but high voltages. The main advantage of using magneto-rheological fluids is that the required electrical voltages can be generated directly from the car's own electric system without the need for invertors.
Vehicle models
According to Crolla [36], ride characteristics of passenger vehicles can be characterized by considering the so-called, quarter-car model which allows for only one-dimensional vertical motion. A quarter car model considers only one wheel and sprung mass bounce motion (in the vertical direction). For this model, more DOFs (degrees of freedom) other than the basic two degrees of freedom could be considered. However, each mass will only have DOF (degrees of freedom) in the vertical direction, and pitch and roll are neglected. The next category of model incorporates the vertical oscillation of the vehicle as well as the pitch of the vehicle body. This is often referred to as a half-car or bicycle model. The next modelling option incorporates vertical motion, pitch and roll which is referred as full car model. Thus greater accuracy is achieved by extensions to a half [37] or full car model [38].
Application of control techniques to semi-active suspension damper
As discussed in the preceding sections, the semi-active shock absorber generates forces passively but these forces are modulated continuously in accordance with some prescribed control law with only a small amount of external power. In other words, the semi-active shock absorber is basically a device with time varying controllable damping. The concept of semi-active control was first introduced by Karnopp in 1974 [39], and has since been developed and demonstrated to be a viable suspension alternative. This section briefly reviews the control techniques for semi-active suspensions.
The elementary semi-active controller design is the so-called ''on-off'' semi-active strategy, which was first proposed in 1975 [40]. It switches the damper off whenever sprung and unsprung masses move in the same direction and unsprung mass has a larger velocity. In any other situations the damper is set to the ''ON'' state. A somewhat more sophisticated approach is to change the damping from soft to firm and vice versa through a manual or slow adaptive control. This is referred to as ''on-off skyhook'' control policy. Conventional semi active damper control strategy may be of the on-off skyhook type or of the Continuous skyhook type [41]. The paper by Shamsi and Chouani [42] compared these two basic conventional control strategies in the vibration isolation. Discrete semi-active shock absorbers can only switch between an off-state and an on-state, corresponding to a minimum and maximum amount of damping. Miller [43] explored the effects of the levels of both on-state and off-state damping on the performance of the quarter car semi-active suspension system. It was found that the discontinuities from switching the damper on and off present in skyhook control policies caused some passenger discomfort. These effects also cause noise caused by the tire suddenly releasing stored energy as the damper is turned off. Methods for reducing or eliminating this noise are discussed in a paper by Miller and Nobles [44]. During recent years there has been considerable interest in the on-off semi-active concept. Further improvements and refinements of the concept were reported [45] and references therein.
The damping force provided by the skyhook damper is always opposite to the absolute velocity. The problem in implementing the Skyhook is that in some applications, especially in the automotive industry, the absolute velocity is impossible to measure. Many researchers, including Hedrick, Rajamani, and Yi [46], Shen, Golnaraghi and Heppler [47] have concentrated on this problem. Rakheja and Sankar [48] developed a control strategy which is based on a simple fact, resulting from governing equation of motion for the 1DOF (degrees of freedom) system. The equation developed by them shows clearly that the damping force in a passive damper tends to increase the mass acceleration when the damping force and the spring force are acting in the same direction. If an ideal semi-active damper is able to produce no damping force, or at least a minimum amount of damping force, when the spring and damping forces are acting in the same direction and produce damping force equal to the spring force when they are acting in different directions, then the mass acceleration and, as a result, the acceleration transmissibility will be minimum. Drawing on this concept, Alanoly and Sankar [49] have suggested an on-off control strategy, which for a 1DOF (degrees of freedom) system to gain a better understanding on the effectiveness of the Skyhook and R-S(Rakheja-Sankar) control strategies, the most widely used conventional semi-active strategies. Eslaminasab and Golnaraghi [50] developed a 1DOF (degrees of freedom) system and simulated in MATLAB® /Simulink. Relative displacement transmissibility and acceleration transmissibility are obtained and compared, which showed a significant reduction of transmitted acceleration in higher frequency ranges when the control strategies are in place. The paper by Yanqing et al [51] presents a two degree of freedom vibration isolation system with variable damping and stiffness on-off control. The damping and stiffness control was achieved using two magneto-rheological (MR) fluid dampers in series. According to the results, the damping and stiffness on-off controlled system has the best performance over a frequency range that encompasses its two structural vibration modes. In a study by Bellizzi and Bouc [52], an adaptive control scheme is developed for a variable orifice semi-active damper. The adaptive control scheme changes its feedback law parameters continuously in an effort to match the non-linear semi-active system to a reference linear system with the desired properties.
On-off semi-active suspension control strategies control the damping ratio of the system through a state-feedback-control. Although these control strategies might use different states and control rules, the resultant controlled system will be a frequency-dependent damping system. In the paper by Eslaminsab, Arzanpour and Golnaraghi [53], the frequency-dependent asymmetric damping systems are analyzed and studied. Despite different semi-active control methods and the type of actuators used, the response-time (delay) is an important practical aspect of all hydro-mechanical computer controlled systems. In the research work by Koo, Goncalves and Ahmadian [54], a comprehensive study on the magneto-rheological damper response-time has been presented. However, the effect of the response-time on the overall performance of the semi-active controlled system is not discussed, nor is the response-time of the solenoid valve controlled semi-active dampers studied. The paper by Eslaminsab and Golnaraghi [55], investigates the effect of response-time in an on-off controlled suspension system equipped with semi-active dampers. Specifically, the effect of the response time on the performance of a 1DOF suspension system controlled by conventional on-off control strategies such as Skyhook and Rakheja-Sankar (R-S) is studied. Finally, the test method and test results measuring the response-times of a number of semi-active dampers are presented.
In recent years, Shen [56] used the nonlinear approach to find a closed form solution for the conventional on-off control strategies to compare the performance of those systems. However, this study does not present a practical method for performance characterization. In the paper by Eslaminsab, Tom and Farid [57], the nonlinearity of on-off semi-active control strategies is discussed, and the nonlinear analysis methods are used to explore the problem. A method to characterize the performance of these systems is also established. In addition, the possibility of adverse effects of this nonlinear phenomenon on ride comfort quality of vehicles is discussed. The Root Mean Square (RMS) is a useful means to summarize and evaluate the performance of a suspension system and its control strategies. In addition, to address the issue, the Root Mean Square (RMS) has been used by many researches in this field, including Ahmadian [58], and Cole [59]. Lieh [60] explored the use of semi-active suspensions to control the dynamics of a full car model. He concludes that the use of the skyhook control policy reduces the root mean square (RMS) acceleration of the car body while increasing the root mean square (RMS) tire forces. The skyhook configuration is ideal if the primary goal is isolating the sprung mass from base excitations [61], even at the expense of excessive un-sprung mass motion. It has the disadvantage of increasing wheel hop, and it does not consider rattle-space. To minimize the wheel hop, it is possible to connect the wheel to a skyhook (in this case called a ground-hook), but this creates the opposite problem of increasing the vibration of the body, as now the body has effectively no damping. A hybrid method combining the best aspects of ground-hook and skyhook is discussed in a paper by Ahmadian [62] as a potential alternative to provide a more acceptable compromise for a wider range of applications. In practice the state cannot be measured online and it is more realistic to assume that only the body acceleration and rattle displacement are available for feedback control. Based on the Kalman filter theory Grewal and Andrews [63], suggest that an observer can be designed which optimally estimates the states from available measurement signals.
The continuously variable semi-active policy represents the next step up in the sophistication. It requires that the semi-active actuator continuously reproduce a linear quadratic (LQ) optimal control skyhook damping force whenever this is possible in view of the passivity constraint [64]. When this is not possible, the damper is simply turned off. The continuously-variable semi-active policy was subsequently extended to more complex model, which led to so-called ''clipped'' semi-active control [45]. The optimal semi-active control law was first studied in [65]. It was later proved that the clipped semi-active policy may often be very close to being optimal but not always [64]. Mostly linear quadratic (LQ) based optimal control concepts give useful insights about the performance characteristics and other requirements [45, 65].
Advanced Semi active Control Strategies
Although Zadeh developed fuzzy logic in 1965, only in recent years has it been applied to vehicle suspension dynamic control. Pedryez and Gomide [66], Klir and Folger [67] states that Fuzzy logic is based on the extension of two-valued logic (yes/no) to n-valued logic. Among the first application attempts, Holou, Joo and Shaout [68] adopted fuzzy logic to control the vibration of semi-active suspension systems. Although results obtained by fuzzy logic have shown improvement over passive suspension, not much work has been done to compare the results from fuzzy logic controllers with those from other conventional methods (introduced above). In the paper by Eslaminasab et al [69], a fuzzy logic controller, as well as Skyhook and Rakheja-Sankar controllers, is designed, and results are compared. In addition, since the suspension systems are complex and non-linear in nature, neural network is introduced as a tool to model the properties of an actual semi-active damper. Despite the large number of publications related to control of semi-active suspensions, most do not account for the nonlinear characteristics of vehicle suspensions. Techniques that do address nonlinearities include sliding mode control [70, 71] and the application of neural networks [70, 72]. However, these approaches typically require more accurate system modeling and development time than is suitable for commercial adoption. For these reasons, the application of fuzzy logic control (FLC) has become very popular for semi-active suspensions, as it is very intuitive for the user, easily implemented, and can account for system nonlinearities [69,70,73,74,75]. Yoshimura et al [76] developed fuzzy logic control methodologies to control both continuous and discrete adjustable dampers semi-actively. The paper by Michael Jacob [77] outlines the benefits of implementing real-time, fuzzy logic control (FLC) to a vehicle suspension equipped with commercial magneto-rheological (MR) shock absorbers. Lieh and Li [78] discuss the benefits of an adaptive fuzzy control compared to simple on-off and variable semi active suspensions. The intent of their work is to apply a fuzzy logic concept to control semi active damping that is normally nonlinear with stochastic disturbances. A quarter-car model was used to implement the fuzzy control rule. The paper by Mehrdad, Khajavi and Abdollahi [79], compares the ride comfort and drive stability performance of a specific automobile with passive suspension system to a proposed fuzzy logic semi active suspension system designed for that automobile. The result shows improvement over passive suspension method.
Conclusions
To summarize, this review has formulated the fundamental principles of semi-active suspension damper and control strategies which provide vibration suppression solutions for tonal and broadband applications with small amount of control and relatively low cost. However, it is quite a design challenge by using conventional technologies to build a practical semi-active suspension under the constraints of weight, size and cost. There are many important areas directly or indirectly related to the main theme of this paper such as practical implementation of semi-active suspensions, nonlinear control schemes, actual hardware implementation, actuator bandwidth requirements, reliability and cost. Furthermore, the design of semi-active suspension involves many mechanical and electrical components that puts limit on the tuning range of the resonance frequency of the device. Design guidelines to improve vehicle ride comfort and handling are also presented. Future trend in research have also been indicated.
Design Guidelines
The following is design guidance based on the literatures presented in this paper:
(a) A passive suspension system is a compromise between ride comfort and handling as the respective requirements for ride comfort and handling are at opposite ends of the design space.
(b) The standard way of choosing damper characteristics, is to decide whether to bias the suspension system towards ride comfort, or handling. This compromise is often improved with the use of semi-active suspension damper.
(c) A new approach, that seems feasible, is to use a controllable suspension damper that can have two discrete states - one giving best possible ride comfort and the other giving best possible handling.
To implement the new approach suggested in (c) above, the following is required:
(a) Two discrete damper characteristics namely:
High damping for best handling.
Low damping for best ride comfort.
(b) The capability to switch between the ''ride comfort mode'' (low damping) and the ''handling mode'' (high damping).
The literatures presented strongly suggests that a control strategy, which can switch between a ''ride comfort'' mode and a ''handling'' mode in a safe and predictable way, should result in significant improvements in both ride comfort and handling. Additional improvements and benefits may be obtained applying control to the individual dampers.
Trends and Future Research
Future work should include the following:
(i) Develop suitable semi active suspension damper to achieve the required characteristics.
(ii) Develop a control strategy to make the ''ride comfort vs. handling'' decision.
(iii) Implement on a test vehicle if suspension hardware and control is feasible.
Many leading automotive industries are carrying out these exercises in house, but these investigation results are not available in the public domain. Hopefully, they will be documented and made public very soon. Furthermore, in the process of designing a semi-active suspension, in practice, several critical criteria must be considered, which were not discussed here. These include weight, size, shape, center-of gravity, types of dynamic disturbances, allowable system response, ambient environment and service life. A designer cannot satisfy all these requirements simultaneously. Certainly future trends in research, in the present authors' opinion, will attempt to answer these questions scientifically through synthesis.
