This paper investigates the field-dependent rheological properties of magnetorheological (MR) fluid used to fill in MR dampers after long-term cyclic operation. For testing purposes, a meandering MR valve was customized to create a double-ended MR damper in which MR fluid flowed inside the valve due to the magnetic flux density. The test was conducted for 170,000 cycles using a fatigue dynamic testing machine which has 20 mm of stroke length and 0.4 Hz of frequency. Firstly, the damping force was investigated as the number of operating cycles increased. Secondly, the change in viscosity of the MR fluid was identified as in-use thickening (IUT). Finally, the morphological observation of MR particles was undertaken before and after the long-term operation. From these tests, it was demonstrated that the damping force increased as the number of operating cycles increases, both when the damper is turn on (on-state) and off (off-state). It is also observed that the particle size and shape changed due to the long operation, showing irregular particles.

Magnetorheological (MR) fluid devices are now applied in various applications. Although a lot of studies have been discussed in long-term implementation of MR devices, only few studies have concerned on MR fluid application in valve operation mode, such as at MR valve. The experiments were conducted on a dynamic test machine in a custom MR damper with meandering MR valve. The experiment test was applied at continuous-load in long term-operation with parameters of 20 mm length of stroke, 0.4 Hz of frequency, 0.5 A of applied current and 175,000 cycles. The rheological properties of MR fluid were characterized using rotational and oscillatory shear rheometer. The result showed that the rheological properties of MR fluid changed after applied in long-term operation. The changed of MR fluid also investigated through morphological characteristics using SEM and EDX.

An experimental study was undertaken to evaluate the mathematical modelling of the magnetorheological (MR) damper featuring annular radial gap on its valve. The experiment was conducted using a fatigue dynamic test machine under particular excitation frequency and amplitude to get force-velocity and force-displament characteristics. Meanwhile, the mathematical modelling was done using quasi-steady modelling approach. Simulation using adaptive neuro fuzzy inference (ANFIS) Algorithm (Gaussian and Generalized Bell) were also carried out to portray the damping force-displacement modelling that is used to compare with the experimental results. The experimental characteristics show that amplitudes excitation and current input affect the result damping force value. The comparison of the experimental and mathematical results presented in this paper shows a significant difference in damping force value and that the quasi-steady modelling could not significantly approach the damping force-velocity results. Moreover, the semi-active damper is compared to the passive damper. The results show that a semi-active damper performs better than a passive damper because it only requires a little power. Based on the damping force-displacement modelling, it can be seen that Gaussian has a higher accuracy rather than Generalized Bell. Discussion on the energy dissipation and equivalent damping coefficient were also accomodated in this paper. Having completed in mathematical modelling and simulation, the damper would be ready for further work in-vehicle application that is development of control system.

The extreme learning machine (ELM) plays an important role to predict magnetorheological (MR) fluid behavior and to reduce the computational fluid dynamics (CFD) calculation cost while simulating the MR fluid flow of an MR actuator. This paper presents a logarithm normalized method to enhance the prediction of ELM of the flow curve representing the MR fluid rheological properties. MRC C1L was used to test the performance of the proposed method, and different activation functions of ELMs were chosen to be the neural networks setting. The Normalized Root Mean Square Error (NRMSE) was selected as the indicator of the ELM prediction accuracy. NRMSE of the proposed method is found to improve the model accuracy up to 77.10 % for the prediction or testing case while comparing with the linear normalized ELM

No abstract