Theoretical and experimental study on seismic response control on top of Three-Gorges ship lift towers using magnetorheological intelligent isolation system and its key technique

Abstract

A vertical ship lift under earthquake excitation may suffer from a whipping effect due to the sudden change of building lateral stiffness at the top of the ship lift towers. This paper proposes a roof magnetorheological (MR) intelligent isolation system to prevent the seismic whipping effect on machinery structures. Theoretically, the dynamic models of MR damper and the mechanical model of ship lift was established, the inverse neural network controlling algorithm was proposed and the fundamental semi-active control equation for the Three-Gorges ship lift where the MR intelligent isolation system was installed was deduced. Experimentally, the experimental model of the ship lift was given, the vibrating table experiment of the MR intelligent isolation system controlling the whipping effect was carried out and the results of the inverse neural network control strategy and passive isolation strategy were compared. In practical aspect, the large-scale MR damper (500 kN) and a sliding support with limited stiffness were designed and fabricated. It was proven that the MR intelligent isolation system with proper control strategy can greatly reduce the seismic whipping effect on the top workshop of the ship lift and be simple and effective enough to be applied to real engineering structures.

Keywords

Three-Gorges ship lift tower / whipping effect / magnetorheological (MR) intelligent isolation system / neural network / sliding steel supporting

Introduction

The Three-Gorges ship lift, with the basic function of lifting ships vertically and shortening the time to cross a dam, is a significant structure in the Three-Gorges water conservation works. It is constructed using four huge reinforced concrete towers and a thick reinforced concrete top platform with a single-story reinforced concrete workshop. Because of very heavy loads on the ship lift towers, their lateral stiffness should be much greater than that of the top workshop. As a result, the top workshop may suffer from a whipping effect due to the sudden change of lateral stiffness. Because of the interaction between the workshop and the reinforced concrete towers, the seismic response at the workshop is substantially greater than that of a building constructed directly on the ground. Therefore, the seismic response control at the large-span top workshop becomes imperative in the safe design of the ship lift structure [1,2].

According to the structure of the Three-Gorges ship lift, the top workshop is a single-layered structure with a large span and no supporting or energy-consuming system is installed on the brace, so that the cranes and big lifts can work properly. But it is hard to control the whipping effect on the top workshop with normal structural control method. In this case, designing a passive isolation layer between the pillar tip of the workshop and the bracket of the roof structure is a simple and convenient way to change the dynamics of the workshop structure and reduce the seismic energy on it; however, the isolation layer will produce great interlayer displacement which can destroy the isolator and cause big P-▵ effect.

In order to solve this problem, intelligent control method was proposed to mitigate the whipping effect on the large-span workshop at the top of the Three-Gorges ship lift. The MR dampers and seismic isolators were used to form the MR intelligent isolation system which was installed under the roof truss of the large-span workshop to prevent the whipping effect [3], as shown in Fig.1. The seismic isolator devised between the tip of the workshop pillar and the base of the roof truss can change the dynamic performance of the ship lift and decrease the seismic force of the pillar from the top workshop building. The MR dampers connected with the seismic isolator can dissipate part of the seismic energy of the top machinery structure and prevent the seismic isolator from breakage due to excessive relative displacement.

MR damper is one of the most available performance control devices which have the attractive characteristics of providing controllable damping force for structural vibration response reduction. To realize the vibration response reduction of the MR intelligent isolation system, the large-scale and high-performance MR damper must be manufactured. Over the last several years, a number of studies have been done for the laboratory experiments and application of MR dampers. The MR damper with a maximum damping force of 200 kN and a dynamic range equal to 10 at full-scale has been designed and fabricated for its application in civil engineering structural vibration reduction. Additionally, its performance in laboratory experiment has been tested in the University of Norte Dame [4,5]. In 2001, the Nihon-Kagaku Miraikan, the Tokyo National Museum of Emerging Science and Innovation installed two 300 kN MR dampers manufactured by Sanwa Tekki Corporation using Lord Corporation MR fluid between the third and fifth floors respectively to provide seismic protection and it was the first implementation of MR dampers applied in civil engineering structure [6]. In 2003, a 400 kN MR damper was emerging in a residential building in Japan to provide the best seismic protection [7]. Ni et al. constituted the first application of full-scale MR damper for bridge cable to provide the cable vibration reduction caused by wind combined with rain. A total of 312 SD-1005 MR dampers manufactured by Lord Corporation were installed on 156 stayed cables in the Dongting Lake Bridge in Hunan Province. Meanwhile, the MR damper fabricated by Ou et al. in Haerbin University of Technology in China was implemented on the Binzhou Yellow Bridge to reduce cable vibration excited by wind and rain [8]. However, at present, MR dampers have not yet been widely adopted in the world mainly due to some technological problems, such as anti-sedimentation problem of MR fluid material, manufacturing technique of large-scale MR damper, etc.

The problem of intelligently controlling the seismic whipping effect of the Three-Gorges ship lift was studied in three aspects: theoretical calculation method, vibrating table experiment and practical key technique. The mechanical models of the MR damper and ship lift were established by calculation method, and an inverse neural network controlling algorithm was proposed and the fundamental equation of semi-active control of the seismic effect at the ship lift was deduced. The experimental model of the ship lift was given. It helps to carry out the vibrating table experiment of the MR intelligent isolation system controlling the whipping effect and compare the results of semi-active control strategy with that of passive isolation strategy. To put the system into practice, the advanced technique to make MR fluid and the large-scale MR damper was studied, and the most powerful and stable MR damper (500 kN) was developed. Finally, an anti-aging vibrating isolation device — the sliding steel support with limited stiffness — along with the powerful MR damper, constitute the real-size MR intelligent isolation system which can be employed in the practical project.

Theory and simulation of MR intelligent isolation control on Three-Gorges ship lift

Controlling equation of MR intelligent isolation system

Because of the complication of the Three-Gorges ship lift, it is necessary to simplify the mechanical model of the structure before installing the MR intelligent isolation system and calculating the seismic response. A typical simplified model of the Three-Gorges ship lift is shown in Fig. 2, which is a one-dimension equivalent lumped-mass model. In order to reduce the whiplash effect on the single stratum workshop, a roof MR intelligent isolation system is mounted between the tip of the workshop pillar and the base of the roof truss.

Considering a seismically excited structure controlled with MR dampers, the equation of motion controlled by the MR intelligent isolation system can be written as

(1)$$\begin{gathered} \mathit{M}\ddot{\mathit{X}}\left(t\right)+\mathit{C}\dot{\mathit{X}}\left(t\right)+\mathit{K}\mathit{X}\left(t\right) \\ =-\mathit{M}\mathit{I}{\ddot{X}}_{\mathrm{g}}\left(t\right)+\mathit{S}{u}_e{k}_{\mathrm{d}}\left(\left(x_{n+1}-x_n\right)-e\right) \\ =C_{\mathrm{d}}\dot{e}+F_{\mathrm{d}}\left(E\right)\mathrm{sgn}\left(\dot{e}\right)=u_e, \end{gathered}$$

where BoldItalic is a vector of relative displacements of the structure’s floors, {\ddot{X}}_{\mathrm{g}}(t) is a one-dimensional ground acceleration, \mathit{S}={\left[0\cdots 0-11\right]}_{1\times \left(n+1\right)}^{\mathrm{T}} is a vector determined by the placement of the MR dampers in the structure, u_e is the control force generated by the MR damper, e is the sliding displacement, BoldItalic, BoldItalic and BoldItalic are the mass, damping and stiffness matrices of the structural system respectively. BoldItalic and BoldItalic can be written as

(2)\mathit{K}={\left[\begin{array}{cccccc}& & & & 0& 0\\ & {\mathit{K}}_1& & & ⋮& ⋮\\ & & & & 0& 0\\ & & & & -K_{\mathrm{n}}& 0\\ 0& \cdots & 0& -K_{\mathrm{n}}& K_{\mathrm{n}}+K_{\mathrm{b}}& -K_{\mathrm{b}}\\ 0& \cdots & 0& 0& -K_{\mathrm{b}}& K_{\mathrm{b}}\end{array}\right]}_{\left(n+1\right)\times \left(n+1\right)},

(3)\mathit{C}={\left[\begin{array}{cccccc}& & & & 0& 0\\ & {\mathit{C}}_1& & & ⋮& ⋮\\ & & & & 0& 0\\ & & & & -C_{\mathrm{n}}& 0\\ 0& \cdots & 0& -C_{\mathrm{n}}& C_{\mathrm{n}}+C_{\mathrm{b}}& -C_{\mathrm{b}}\\ 0& 0& 0& 0& -C_{\mathrm{b}}& C_{\mathrm{b}}\end{array}\right]}_{\left(n+1\right)\times \left(n+1\right)},

where K_n is the lateral stiffness coefficient of the single stratum workshop pillar, K_b is the lateral stiffness of the isolation spacer, C_n is the damping coefficient of the workshop pillar, C_b is the damping coefficient of the isolation spacer. Usually, the lateral stiffness of reinforced concrete tube towers at the underside of the ship lift is a few hundred times as much as that of the single large-span workshop pillar on top of the ship lift; therefore, the huge stiffness mutation will result in the violent whiplash effect on top of the workshop of ship lift under earthquake excitation.

Mechanical model of MR damper considering magnetic hysteresis

Based on the experimental consequence of the MR damper and taking into account the pre-yield zone of stress-strain relationship, the simple mechanical model in Fig.3 was developed and shown to accurately predict the behavior of the MR damper [9]. This model consists of a Bingham cell and a spring cell in series. The spring cell corresponds to the equivalent axis stiffness of the MR damper resulting from the accumulator’s stiffness and initial shear modulus before yielding.

The output force of the MR damper can be expressed as

(4)F\left(t\right)=C_{\mathrm{d}}\dot{e}+F_{\mathrm{d}}\left(E\right)\mathrm{sgn}\left(\dot{e}\right)-F_0=K_{\mathrm{d}}(x-e)-F_0,

where C_d is the viscous damping parameter, F_d(E) is the damping force caused by applied field, e is the displacement of Bingham cell, x is the displacement of MR damper, K_d is the equivalent axis stiffness which is the result from the accumulator’s stiffness and initial shear modulus before yielding, F₀ is the bias of damping force from accumulation.

In order to make this mechanical model valid when changing electric current, the relationship between the model parameters and the current can be determined as

(5)$$\begin{gathered} C_{\mathrm{d}}=C_{\mathrm{d}\mathrm{s}}+C_{\mathrm{d}\mathrm{d}}u, \\ {F}_{\mathrm{d}}+F_{\mathrm{d}\mathrm{s}}+F_{\mathrm{d}\mathrm{s}}u, \\ {K}_{\mathrm{d}}+K_{\mathrm{d}\mathrm{s}}+K_{\mathrm{d}\mathrm{d}}u, \end{gathered}$$

where C_ds, F_dsand K_ds is the viscosity, damping force and equivalent axis stiffness without field respectively, u is given as the output of the first-order filter and expressed as

(6)\dot{u}=-\eta \left(u-I\right)

where u is the intrinsic variable determining the relationship between the parameters and applied current I, η reflects the response time of MR damper which is inversely proportional with response time, I is the applied current. Therefore, eight parameters (C_ds, F_ds, K_ds, C_dd, F_dd, K_dd, F₀, η) must be determined for the proposed MR damper model.

Inverse neural network intelligent control algorithm

Fundamental principle of inverse neural network

If the input in the model of the controlled subject is control signal, the inverse neural network process should be used to simulate the inverse model of the controlled subject, namely, the neural network which can calculate the control signal according to the input and output value of the previous mechanical model. The input and output data set is necessary to establish the inverse neural network to model the inverse processing. For different processing models, a matrix with the dimension of N rows and two columns comprising abundant frequency is needed: one column is the input data and the other is the corresponding output data. A typical inverse neural network of the processing model is shown in Fig. 4 in which y is the output response and u is the control signal.

The purpose of an inverse neural network is to form the inverse model of a system through training data from a forward model of the system. For example, the dynamic behavior of MR dampers can be well described by some forward mechanical models such as Bouc-Wen hysteresis model, revised Bingham model and Dahl model, etc. These models estimate the damper force based on damper displacement, velocity and control current. However, in control applications, it is necessary to command the MR dampers to generate the damper force which can track the optimal force as close as possible. In such cases, it is beneficial to develop an inverse dynamic model using inverse neural network that estimates the required current which is to be input into the damper so that a desirable damper force can be produced.

Realization of inverse neural network control on Three-Gorges ship lift

The inverse neural network of MR dampers applied to the Three-Gorges ship lift can be defined as

(7)u_k=N{N}_{\mathrm{i}\mathrm{n}\mathrm{v}}\left(x_k,x_{k-1},\cdots ,x_{k-l},F_k,F_{k-1},\cdots ,F_{k-m},u_{u-1},u_{k-2},\cdots ,u_{k-n}\right),

where x is the displacement of MR dampers, F is the damper force and u is the control current. The training data set can be generated by revised Bingham model, which is shown in Fig.5.

The data used to train the inverse neural network can be obtained from the revised Bingham model which expresses the mechanical properties of real MR dampers. As a result, nonlinear least square algorithm is used for parameter identification and the identification results are as follows: C_ds=1.50 kN·s·mm^-1, F_ds=7.95 kN, K_ds=3.30 kN·mm^-1, C_dd=1.02 kN·s·mm^-1·A^-1, F_dd=739.99 kN·A^-1, K_dd=10.08 kN·mm^-1·A^-1, F₀=0.98 kN, η=3 s^-1.

In order to gain the training data of the inverse neural network model, the displacement x and applied current I in the revised Bingham model were selected as Gaussian white noise, in which the frequency of the displacement and that of the current ranged from 0 to 5 Hz, and from 0 to 1 Hz, respectively. Then the corresponding control force F could be obtained from the revised Bingham model and the training data of the inverse neural network model was built. The sampling frequency of the training data was 500 Hz, the sampling time was 20 s, and 10000 groups of data were built among which, the former 4000 groups were used to train the neural network, the latter 6000 groups were validating data.

The neural network structure is a fully connected, two layer, feed-forward network with 9 inputs, 12 hidden neurons and 1 output. All neurons in the first layer use hyperbolic transfer functions and all neurons in the second layer use linear transfer functions. The 9 inputs include 1 displacement, 1 control force at present time and 2 displacements, 2 control forces at two steps of previous time and 3 applied currents at three steps of previous time; the 1 output is the applied current at present time. After 97 times of training, the sum-squared error of the neural network reaches 0.05. This shows that the neural network can converge relatively fast and the output is closed relative to the inputs of the neural network model. Figure 6 depicts the target current of the revised Bingham model and the predicted current of the trained inverse neural network model, in which the predicted value can precisely follow the target value during the former 8 s of 4000 groups of training data, while the predicted voltage can also accurately follow the target current during the latter 12 s of 6000 groups of validating data. It is concluded that the network is well trained and has a good generalization. It can be applied not only to the simulation of training data exactly, but also to predict the non-training ones precisely. Therefore, the inverse dynamic characteristics of the MR damper can be simulated by neural network reasonably well and this trained neural network model can be successfully applied to the intelligent control of the Three-Gorges ship lift.

After training, the inverse neural network is used to form the inverse model of MR dampers, which can connect with other active control algorithms to form a close control loop. In this paper, fuzzy active control algorithm was used. The control process is shown in Fig.7.

Simulation results of inverse neural network control

To study the vibration reducing effect made by the MR intelligent isolation system on the top workshop of the ship lift, a simulation calculation with the same parameters is carried out on the original structure (no isolation system), passive isolation control, passive-on control and inverse neural network control. The results of all cases are compared. Figure 8 indicates the maximum seismic bending moments at the bottom of the top workshop pillars in all cases. Since the roof vibration isolation can effectively reduce the peak value of absolute acceleration at the pillar top and the roof layer, the maximum bending moments are greatly decreased in all cases, in which the result of the inverse neural network is the best. Figure 9 compares the maximum interlayer displacements of different layers of the ship lift. It can be learned that all three control methods can effectively restrain the seismic whipping effect of the top workshop by reducing its interlayer displacement. But serious distortion appears at the vibrating isolation layer under passive isolation. The reduction of interlayer displacement produced by the distortion of the isolator will lead to extremely great P-▵ effect and interlayer displacement which destroys the system. On the other hand, passive-on control and inverse neural network control can effectively reduce the interlayer displacement. By comparison, the latter is much better than the former, regardless of interlayer displacement or bottom bending moment. Figure 10 shows the comparison between the inverse neural network control force and active control force. It is noted that after adding the inverse neural network module, the control force can track the active control force, and its control effect can approach the effect of active control.

Vibrating table experiment of Three-Gorges ship lift

An experiment of the vibrating table on simulating the MR intelligent isolation control over the seismic whipping effect of the Three-Gorges ship lift is carried out to testify the performance of the MR intelligent isolation system. According to similarity principle, an experimental model of the ship lift is designed with the MR intelligent isolation system. An effective result is obtained by the control of dSPACE hardware system and inverse neural network control over the MR system.

Ship lift experimental model

The two-story building was designed and constructed as a ship lift model as shown in Fig.11. The steel frame on the upper story was made up of pillars, each with a cross section of 120 mm×160 mm and a height of 700 mm, which simulated the machinery building on the top of the ship lift; the reinforced concrete frame on the lower story simulated the reinforced concrete towers of the ship lift, made up of columns, each with a cross section of 200 mm×200 mm and a height of 1600 mm. The lateral stiffness of the reinforced concrete columns therein is considerably greater than that of the steel pillars, which creates a huge and sudden change of lateral stiffness. Not surprisingly, the machinery building may suffer from serious seismic whipping effect. The plates of the upper frame were divided into two mass pieces, in which the top mass piece simulated the roof truss of the machinery building while the bottom mass piece simulated the crane and the wall of the machinery building. In the modeling experiment with rigid connection, those two pieces were bolted without any relative movement between the two mass pieces; in the modeling experiment only with seismic isolator connection, the two pieces were connected by the seismic isolators; in the modeling experiment with the MR isolation system connection, the two pieces were connected by seismic isolators and MR dampers.

The MR intelligent isolation system used in the experimental model of the ship lift consisted of three sections: seismic isolator, MR damper and dSPACE real-time simulator (made in Germany).

The external diameter of the seismic isolator was 50 mm while the internal diameter was 24 mm. The isolator had 26 rub layers and 25 steel plate layers alternately; the thickness of each layer was 1 mm. The lateral stiffness of the isolator was 26 N·mm^-1 while the vertical stiffness was 7300 N·mm^-1. As a consequence, while the ship lift model vibrated, the seismic isolator would suffer from the large horizontal displacement; on the other hand, it could brace the weight of the top piece which simulated the roof truss of the machinery building.

The MR damper (RD-1097-01X) manufactured by the Lord Corporation was used to link the two mass pieces of the upper steel frame (as shown in Fig.11), which serves as a confinement to the relative displacement of the seismic isolator in this system, so that the isolator would not be damaged because of excessive deformation. The maximum allowable input current to the damper is 0.5 A and 1.0 A, respectively, for continuous and intermittent application. More than 150 N force can be produced by the damper at 1.0 A, whereas an inherent damper force at 0 A is less than 9 N. The MR damper was operated under the maximum allowable temperature of 70° and for no more than 30 s continuous application at 0.6 A in the experiment. To achieve the best performance of the MR damper, the Rheonetic Wonder Box device controller kit (RD-3002-03) designed by the controller kit (the current controller afterwards) provided a closed loop current control. The output current (i.e., the input current to the MR damper) was almost linearly proportional to the input voltage (i.e., the command voltage from the dSPACE) of the current controller. A 0-5 V input signal can be switched up to 1 kHz.

To obtain the seismic responses of the ship lift model, the signal from the signal conditioner was passed to the dSPACE real-time simulator system. The dSPACE real-time simulator system comprised a DS1005 PowerPC controller (PPC) processor board, a DS2003 multichannel analog to digital (A/D) board, a DS2102 high resolution digital to analog (D/A) board, and a DS4003 digital input-output (I/O) board. The control of the dSPACE CPU and the access to its memory were executed by the main program ControlDesk of dSPACE, which offered an automatic implementation of the MATLAB/Simulink block program on the host computer via real time interface (RTI) and provided a real time interactive data display and visualization. The RTI had compiling, linking, downloading, and configuring capacities. In this experiment, the analog signal of 30 s duration, transferred from the signal conditioner to the dSPACE real-time simulator system, was sampled at 1000 Hz by the DS2003 multichannel A/D board to obtain the PPC controller together with the MATLAB/Simulink program on the host computer via RTI.

For the semi-active control of the ship lift model connected by the MR intelligent isolation system, the relative velocity and the relative acceleration of the seismic isolator were measured and transferred to the DS2003 A/D board to have digital signals. The digital signals were then analyzed based on the semi-active control strategy with inverse neural network control algorithm and sent to the DS2102 D/A board in the dSPACE to apply the control voltage signal to the MR damper via the current controller so as to form a closed loop semi-active control system. The control algorithm which consisted of fuzzy algorithm and inverse neural network model of the MR damper was implemented by the PPC controller via the MATLAB/Simulink program. The DS2102 D/A board consisted of six parallel channels at 16-bit resolution. A block diagram for the overall real-time semi-active control process is depicted in Fig.12.

Experimental results

The following aspects are discussed in this section: 1) to validate that the whipping effect on top of the ship lift model can be effectively reduced using the MR isolation system; 2) to prove that the MR damper can be effectively controlled by the semi-active control strategy with the inverse neural network control algorithm. For those purposes, the experiments were carried out on the seismic simulator for the building model with rigid connection, only with seismic isolator connection, and connected by the MR isolation system, respectively. In the experiments, the Three-Gorges artificial earthquake with a peak acceleration of 0.1 g was used as input excitation; the inverse neural network control method was taken to control the MR dampers.

To validate the effective reduction of the whipping effect on the ship lift model only with seismic isolator connection, the relative displacement of the steel pillar of the second story and the absolute acceleration at the top of the steel pillar are shown in Figs.13 and 14, respectively, when the ship lift model was with rigid connection and only with isolator connection under the excitation of Three-Gorges artificial earthquake. It is seen that with the isolator connection, not only the relative displacement but the absolute acceleration are obviously decreased by 40% and 60% respectively, which means that installation of the seismic isolators between the top two pieces of the upper story of the ship lift model can decrease the natural frequency of the structure, and weaken the seismic whipping effect on the structure.

The seismic responses of the ship lift model only with the seismic isolator connection and with the MR isolation connection under the excitation of Three-Gorges artificial earthquake is listed in Table 1. It is seen that the displacement at the tip of the steel pillar with the MR isolation system connection is decreased by 50% under semi-active control with the inverse neural network control algorithm, compared with that with only the seismic isolator connection when the input excitation is the Three-Gorges artificial earthquake. The relative displacement of the isolator with the MR isolation system connection, when subject to Three-Gorges artificial earthquake, is decreased by 57% in the semi-active control mode, compared with that only with the seismic isolator connection. Therefore, the ship lift model with MR isolation system connection is superior to weaken the seismic whipping effect on the structure, compared with only seismic isolator connection, because the MR dampers can prevent the seismic isolator from breakage from excessive relative displacement. Figure 15 shows the time histories of the relative displacement of the steel pillar for the three cases. It is learned that the safety of the top piece of the ship lift model is threatened because of the large relative displacement of the isolator only with the seismic isolator connection. With the MR isolation system connection, however, the relative displacement of the isolator is significantly reduced when the MR damper dissipates part of the seismic energy of the top piece of the ship lift model. Furthermore, the MR damper works in a better way in semi-active control mode with the inverse neural network control algorithm.

Application of key technique of MR intelligent isolation system

Manufacturing technique of high-performance MR fluid

The key intelligent material of MR damper is MR fluid which directly influences the performance of the MR damper. To meet the need for great stability and high yield stress of MR fluid, a systematic research is conducted about the accessory-ingredient-reshaping oil-based MR fluid system with ferriferous oxide, ferrous cobaltic alloy particles, carbonyl iron dust, micron/nanometer composite particles, carbonyl iron/polymer composite particles as discrete phase, and with silicon oil, coal oil, polyether etc. as continuous phase. It analyzes the relationship between the stability of MR fluid and its preparation process, between the magnetic field yield stress of MR fluid and its preparation process. By optimizing the formula and technique, a method is obtained to improve the sedimentation stability and yield stress of MR fluid and then give a solution to the three key techniques. They are respectively, 1) the preparation technique of shell-like magnetic composite particles, which proposes to adopt heterocyclic surfactant to improve the surface performance of magnetic particles, systematically studies the synthetic technique of particular heterocyclic surfactant and the technique of reshaped magnetic particles, and succeeds in producing the magnetic particles which has perfect soft magnetism and can form a network with the carrying fluid; 2) the preparation technique of functional carrying fluid, which transforms the function of the carrying fluid from the traditional solvent function to multi-function of solvent, resistance to abrasion, resistance to rotting, resistance to deposit, and anti-oxidation, not only improving the sedimentation stability, resistance to abrasion, resistance to rotting, and anti-oxidation of magnetic particles in the carrying fluid, but also effectively settling the contradiction between the stability and high viscosity of MR fluid whose stability takes the leading position in the world; 3) the preparation technique of high-powered magnetic rheological fluid, which studies the performance and technology of the oil base MR fluid with ferriferous oxide, ferrous cobaltic alloy particles, carbonyl iron dust, micron/nanometer composite particles, carbonyl iron/polymer composite particles as discrete phase, optimizing the technique to produce the high-powered MR fluid and applying it from the small-scale laboratory use into a large-scale semi-industrial manufacturing.

With the above-mentioned techniques, the high-powered MR fluid with leading stability is prepared and whose response time is shorter than 50 ms (the international standard is longer than 100 ms) and the maximum magnetic field yield stress is 50-80 kPa (the international standard is 50-80 kPa).

Key technique of designing and producing large-scale MR damper

The MR damper in the Three-Gorges ship lift MR intelligent isolation system works in improving the performance of the passive isolator, consuming some seismic energy at the top workshop, effectively reducing the interlayer displacement between the vibrating isolation layers and avoiding destroying the isolation system on condition of ensuring the isolation effect. The stability, duration, and adjustability of MR damper is fairly important. Therefore, three key techniques are studied on adopting MR damper in the ship lift. They are: 1) a built-in plate-shaped spring accumulator is adopted to supply an adjustable working pressure up to 10 MPa to empty all the air bubbles remaining in the MR fluid in damper oil cylinder, shortening the response time of the damper and greatly improving the stability of MR damper; 2) since the effective magnetic poles of MR damper touch its liquid material and both materials are magnetoconductable, the magnetic field would leak out from beside the poles. Therefore, non-magnetoconductable baffles are devised at each side of the damper magnetic core to prevent the magnetic line of force from leaking there. This technique can reduce the damper size and the number of loops, then decrease the heat of the damper and increase the duration of the MR damper; 3) the techniques of amplifying electric current and neural network prediction are used to adjust the hysteresis effect of MR damper so that its effect on control can be reduced or eliminated to ensure the stable adjustability of a damper when exerting vibrational control over a large engineering structure.

The most large-scale and stable MR damper whose controlling force is up to 500 kN is developed successfully with the above-mentioned key techniques and methods.

Manufacturing technique of sliding steel supporting with limited stiffness

The other important component in MR intelligent isolation system is the passive isolator. Considering the big mass of the top workshop roof which results in the large seismic inertia force, there would be a disastrous result if the passive isolator does not have enough intensity and tenacity. Thus, a new sliding steel supporting it is developed with a limited stiffness. The supporting uses surface contact which means small and even pressure. It can undertake pressure, pulling force and shearing force from all directions. Compared with the ordinary rubber support, this durable steel support overcomes the problem of rubber aging. It is very suitable to be used in the very high and big-spanned works like the Three-Gorges ship lift.

Conclusions

The MR intelligent isolation system is a practical semi-active intelligent controlling device to restrain the seismic whipping effect at the top workshop of the Three-Gorges ship lift as it can effectively reduce the maximum interlayer displacement of the top workshop and the maximum seismic bending moment of the pillars. In the MR intelligent isolation system, the MR damper plays a key role in reducing the interlayer displacement between the isolation layers and protects the passive isolator from being destroyed because of great horizontal deformation. The inverse neural network controlling algorithm can control the MR damper effectively. All the key techniques of the MR intelligent isolation system have been studied and solved and this system can be employed in the practical project.

References

[1] Wu Jiefang, Chen Minzhong, Dai Xinhe, Niu Xinqiang. Whole structure aseismic computation of Three Gorges project’s shiplift. Journal of Yangtze River Scientific Research Institute, 1997, 14(3): 44-47 (in Chinese)

[2] Chen Houqun, Hu Xiao, Wang Ji, Li Deyu, Yu Ying. Dynamic test and analysis for shiplifter tower structure of the Three Gorge project on shaking table. Earthquake Engineering and Engineering Vibration, 1999, 19(1): 47-53 (in Chinese)

[3] Qu Weilian, Lv Mingyun, Li Xuean. Fuzzy seismic control of top building of lift-ship structure with roof intelligent isolation system. Earthquake Engineering and Engineering Vibration, 2002, 22(3): 129-137 (in Chinese)

[4] Yang G. Large-scale magnetorheological fluid damper for vibration mitigation:modeling, testing and conrol. Dissertation for the Doctoral Degree. Indiana: University of Notre Dame, 2001

[5] Housner G W, Bergman L A, Caughey T K, Chassiakos A G, Claus R O, Masri S F, Skelton R E, Soong T T, Spencer B F, Yao J T P. Structural control: past, present and future. Journal of Engineering Mechanics, 1997, 123(9): 897-971

[6] Ou Jinping. Structure Vibration Control: Active, Semi-active, Intelligent Control. Beijing: Science Press, 2003 (in Chinese)

[7] Fujitani H, . Development of 400kN magnetorheological damper for a real base-isolated building. In: Proceedings of SPIE Conference Smart Structures and Materials, Vol. 5052, Bellingham(WA): SPIE-International Society for Optical Engineering, 2003, 265-276

[8] Ni Y Q, Ko J M, Chen Z Q, Spencer B F. Lessons learned from application of semi-active MR dampers to bridge cables for wind-rain-induced vibration control. In: Proceedings of China-Japan Workshop on Vibration Control and Health Monitoring of Structures and Third Chinese Symposium on Structural Vibration Control, Shanghai, China. 2002

[9] Zhou Qiang, Qu Weilian. Two mechanical models for magnetorheological damper and corresponding test verification. Earthquake Engineering and Engineering Vibration, 2002, 22: 144-150 (in Chinese)

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg