Abstract
Vibration control systems using smart dampers (SmDs) such as magnetorheological and electrorheological dampers (MRD and ERD), which are classified as the integrated structure-SmD control systems (ISSmDCSs), have been actively researched and widely used. This work proposes a new controller for a class of ISSmDCSs in which high accuracy of SmD models as well as increment of control ability to deal with uncertainty and time delay are to be expected. In order to achieve this goal, two formualtion steps are required; a non-parametric SmD model based on an adaptive neuro-fuzzy inference system (ANFIS) and a novel fuzzy sliding mode controller (FSMC) which can weaken the model error of the ISSmDCSs and hence provide enhanced vibration control performances. As for the formulation of the proposed controller, first, an ANFIS controller is desgned to identify SmDs using the improved control algorithm named improved establishing neuro-fuzzy system (establishing neuro-fuzzy system). Second, a new control law for the FSMC is designed via Lyapunov stability analysis. An application to a semi-active MRD vehicle suspension system is then undertaken to illustrate and evaluate the effectiveness of the proposed control method. It is demonstrated through an experimental realization that the FSMC proposed in this work shows superior vibration control performance of the vehicle suspension compared to other surveyed controller which have similar structures to the FSMC, such as fuzzy logic and sliding mode control.
Export citation and abstract BibTeX RIS
Nomenclature
Hyper-plane index vector | |
C | Number of data clusters |
Error vector | |
Damping force of SmD | |
, | Fuzzy approximation functions |
The desired current | |
Number of data directions | |
Number of data samples | |
Sliding surface | |
Number of samples in | |
Control force | |
State variable vector | |
Desired state variable vector | |
Cluster centroid | |
Input–output sample | |
Output of control system | |
Normalized data output | |
Desired boundary layer of | |
Membership degree | |
kth hyper-plane |
SMC | sliding mode control |
SmD | smart damper |
MRD | magnetorheological damper |
ERD | electrorheological damper |
ISSmDCS | integrated structure-SmD control systems |
ANFIS | adaptive neuro-fuzzy inference system |
FSMC | fuzzy sliding mode controller |
im-ENFS | improved establishing neuro-fuzzy system |
I-SmD | inverse smart damper model |
FL | fuzzy logic |
ANN | artificial neural networks |
UAD | uncertainty and disturbances |
ISS | integrated SmD system |
I-MRD | inverse magnetorheological damper model |
IDS | initial data space |
CDS | cluster data space |
JDS | jointed input–output data space |
LMSM | least mean squares method |
RMSE | root-mean-square error |
MISO | multi-input single output |
AD/DA | analog–digital/digital–analog |
LVDT | linear variable differential transformer |
ANFIS-PF | ANFIS based on a data potential field |
OHCS | optimized hyperplane clustering synthesis |
OD-T2FLS | optimally designed interval type-2 fuzzy logic system |
NFSmUoC | neuro-fuzzy sliding mode control enhanced by an uncertain observer |
1. Instruction
During the past few decades, smart damper (SmD), consisting of magnetorheological and electrorheological dampers (MRD and ERD), has been used widely for the vibration control in many application fields. The two most prominent application fields of SmD are vehicle and civil engineerings such as SmD suspension systems of road and railway vehicles [1–11] and seismic dampers [12–17]. In general, these applications can be classified as the integrated structure-SmD control systems (ISSmDCSs) whose operating principle and structure can be illustrated as in figure 1. The controller is a combination of an estimator, for calculating the desired control force and an inverse model of SmDs signed I-SmD, for specifying the desired current (or voltage U). Based on the plant dynamic response depicted via the state variable vector and the desired control force estimated by the estimator, the I-SmD specifies (or U) supporting the SmDs such that they can generate the real damping force The aim is the error which means that the states of the plant track well the desired states, In order to achieve the above control aim, the following aspects are focussed in this work. (1) Building a non-parametric SmD model is undertaken. The model must be able not only to characterize appropriately the intrinsic characteristics of the smart fluid via dynamic response of each SmD but also have a competence to express the interaction effects. Thes effects many be occurred during control action from the relationship among the SmD/SmDs, plant and other nonlinear control devices constituting the ISSmDCS in the presence of uncertainty related to the ISSmDCS model error and external disturbances. (2) Designing a controller for a class of the ISSmDCSs such that the unwanted impact of uncertainty coming from the ISSmDCS model error and time delay is undertkaen. By doing this, in a robustly stable process.
The SmD model is used to develop control algorithms taking full advantages of SmDs expressed via the multi-directional and nonlinear relation between dynamic responses and created damping forces of them. It should be noted that the hysteretic response shown in figure 2 as well as the parameter variation characteristic are the main attributes of SmDs model [18]. The techniques for this issue can be broadly classified into two categories: parametric and non-parametric model. Several parametric forward models described in terms of analogous mechanical elements have been developed [11, 18–22]. The non-parametric modeling techniques have also received considerable attentions recently, which mainly includes fuzzy logic (FL) systems [23], artificial neural networks (ANN) [24–26] and adaptive neuro-fuzzy inference systems (ANFISs) [7, 17, 27–33]. In [22], Nguyen et al presented an analytical approach for dynamic modelling of ERD for accurate prediction of the hysteresis behavior. In [19], Seong et al presented damping force control performances of a MRD via a control strategy considering the hysteretic behavior of the field-dependent damping force. In [20], Nguyen and Choi showed dynamic modeling of an ERD using a lumped parameter method, in which a quasi-static modeling of the damper has been conducted on the basis of the Bingham model of ER fluid. In [21], Choi et al presented a feedback control performance of a full-car suspension system featuring ERDs for a passenger vehicle. A cylindrical ERD was established by incorporating the Bingham model of ER fluid. It can be observed that many of models focused on the hysteretic behavior of the field-dependent damping force only. In fact, several other characteristics participating in the complicated relation between dynamic responses and created damping forces of SmDs are exhisted. In [18], Choi et al presented vibration control of a semi-active MRD suspension system considering two important characteristics of MR fluid; the field-dependent hysteretic behavior and the parameter variation.
Download figure:
Standard image High-resolution imageIn several cases, the model must have a competence to characterize the interaction effects among the SmDs, plant and the other nonlinear control devices constituting the ISSmDCS [30]. Moreover, uncertainty problem related to the model error becomes more complex in the systems subjected to severe external disturbances. Obviously, it is quite challenging to develop an accurate mathematical model using the parametric models as mentioned above since a group of SmDs, the control devices and the plant should be considered in the presence of uncertainty and disturbances (UAD). A non-parametric model, which has an ability to deal well with these aspects to express the relation between the input and output of the SmD or of the integrated SmD system (ISSs), hence, is really necessary. Especially, when the systems are under high impact loads [30], FL, ANN and ANFIS have own advantages compared with the traditional parametric methods [27, 29]. Therefore, many researches have focused on this trend [7, 17, 23–33]. In [28], in order to predict and analyze the highly nonlinear behavior of integrated structure-control systems, the ANFIS which can identify smart concrete structures equipped with MRDs under the variety of high impact loads was used. In [29], the semi-active control of dynamic response of civil structures with MRDs was emerged in which the recurrent neural network modeling approach was used to reproduce the hysteretic nonlinear behavior. In [30], a time-delayed ANFIS was proposed for modeling of the complex nonlinear behavior of smart structures equipped with MRDs dampers. In [31], in order to improve the modeling accuracy, a hybrid modeling strategy via an adaptive-network-based fuzzy inference system was proposed to express the inverse dynamic characteristics of MRDs.
In the previous research [27], Nguyen and Choi proposed an algorithm for establishing neuro-fuzzy system named establishing neuro-fuzzy system (ENFS) to identify SmDs. A given data set (or the initial data space, IDS) of input–output samples derived from an unknown mapping from the input data space X to the output data space Y, was approximated by the ANFIS. To establish this model, firstly, clustering the IDS was performed to build a cluster data space (CDS). Based on the created CDS, fuzzy sets were generated to take part in the fuzzy inference system as a framework to which establishing fuzzy laws and defuzzification were all organized, associated and operated. In the algorithm ENFS, clustering in space was performed by separating data samples in X and in Y distinctly with a mutual result reference. This way, however, results in difficulties related to deploy the fuzzy clustering strategy which causes high calculating cost. In many cases, a hard relation could not appropriately reflect database attributes. Related to controlling the ISSmDCSs, in general, the estimator shown in figure 1 always operates in conditions of existing uncertainty coming from the mathematical model error of ISSmDCSs and time delay. Many solutions have been provided based on FL [34–36] using the sliding mode control (SMC) technique [8, 37–41], or combination models [10, 42–48]. Typical advantages of SMC are simplicity of implementation, robustness, ability to deal with uncertainty aspects, insensitivity to external effects, and easy to co-ordinate with other mathematical tools [6, 48]. Reality shows that the combination of SMC and FL in the well-known structure named fuzzy sliding mode control (FSMC), can enhance the effectiveness of the control systems [45–48]. In FSMCs, advantages of FL and SMC are exploited at the same time. Thus, fuzzy systems can be used as powerful approximators while a sliding mode approach adds the possibility of thorough stability analysis to establish the adaptation laws [49]. In [41], an optimal adaptive FSMC for a class of nonlinear systems subjected to UAD was presented. In this controller, however, a difficulty comes from the use of the feedback linearization approach which exists latent uncertainty attributes. In addition, the calculating cost is really an issue of large network systems. It is known that FL systems frequenctly increases the calculating cost. Despite many works on the modeling and controlling dynamic systems subjected to UAD, the development on the both accurate model and robust controller to enhance vibration control performances of the ISSmDCSs is still needed.
Consequently, the technical originality of this work is to propose accurate SmD models as well as formulate a novel FSMC to enhance vibration control performance of the ISSmDCSs subjected to uncertainty and time delay of the controller. In order to achieve this goal, firstly, an improved algorithm named improved establishing neuro-fuzzy system (im-ENFS) for establishing ANFIS which derives from the previous ENFS algorithm [27] is established. It is noted that the ANFIS works as a tool to build non-parametric model of SmDs as well as the ISSs using the measured databases. A novel FSMC, which can weaken time delay and uncertainty, is then formulated as a second step. In the formulation of the FSMC, in order to reduce the calculating cost, an adaptive gain calculated directly and updated adaptively based on the convergent status of the sliding surface is used. A robust stability of the proposed control system is proved using the Lyapunov stability criteria. Subsequently, in order to evaluate control effectiveness of the proposed method, an experimental apparatus equipped with the semi-active MRD vehicle suspension is set up. It is then demonstated that the proposed FSMC shows better vibration control responses compared with other type of the FSMC.
2. Problem formulation
2.1. Inverse SmD model
In order to control ISSmDCSs, the I-SmD shown in figure 1 needs to be established. As mentioned in Introduction, it is quite challenging to develop an accurate mathematical model using the parametric models for SmDs and ISSs in the presence of UAD [11, 18–22]. Therefore, non-parametric models using ANFIS have been widely utilized for modeling of uncertain control systems [7, 17, 27–33]. In this work, the I-SmDs typed ANFIS-I-SmDs is built using the ANFIS. The ANFIS-I-SmD expresses the relationship between dynamic responses and applied voltage or current value of the SmD (or the ISSs). Building an ANFIS-I-SmD is performed via two steps. The first step is to measure dynamic responses of the damper to build an IDS and the second one is to identify the damper via the IDS and ANFIS training algorithm. Figure 3 shows the IDS for building I-MRD, where, is the relative piston-cylinder displacement of the MRD; is the acceleration of the piston; is damping force; and is current supporting the MRD. In this example, for the I-MRD, is input, while is the output signal which needs to be estimated to generate the desired damping force, The ANFIS is used to identify and hence the result of this work can be indicated by I = ANFIS-I-SmD (dre, a, fsd).
Download figure:
Standard image High-resolution imageGenerally, the IDS of input–output samples and expresses an unknown mapping f : X → Y from the input data space X to the output data space Y. The result of the training ANFIS based on the ISD is an approximation of f : X → Y by the ANFIS. To train the ANFIS, firstly, clustering the IDS needs to be done. This is acknowledged as an important step to analyze the IDS and build a CDS. Based on this CDS, fuzzy sets are created to take part in the fuzzy inference system as a framework to which establishing fuzzy laws and defuzzification are all organized, associated and operated. In the previous study [27], to identify dynamic characteristics of the SmDs, clustering in a data space via an algorithm for establishing ANFIS named ENFS was presented. Thus, the separating data samples in data spaces X and Y was carried out distinctly with a mutual result reference, step by step. This way, however, results in difficulties related to deploying a fuzzy clustering strategy along with the high calculating cost. To overcome this, in this work an algorithm named im-ENFS is proposed. First, a jointed input–output data space (JDS) deriving from the IDS is presented, and a kernel-based fuzzy clustering process using a new cluster-centroid update law is then carried out in the JDS. Thus, a close interaction between X and Y is established and always upholden during the clustering process to increase the accuracy rate of the ANFIS. Based on the created CDS and the im-ENFS, the ANFIS is generated. Subsequently, the ANFIS works as the ANFIS-D-SmD or ANFIS-I-SmD.
2.2. Controller for the ISSmDCSs
Consider the ISSmDCS shown in figure 1 which is subjected to uncertainty coming from the model error. The dynamic response of the ISSmDCS can be expressed as follows:
where, f(·), g(·) are unknown functions; is control signal; is the output; is the state variable vector. Figure 4 shows the operating principle of the proposed controller. Based on the plant dynamic response depicted by and the desired control force estimated by the estimator FSMC, the ANFIS-I-SmD specifies the corresponding current (or voltage U) so that the SmD can generate the damping force The aim of the FSMC is to specify the control law in presence of uncertainty such that tracks stably the desired reference state .
Download figure:
Standard image High-resolution imageLet is the error vector defined as follows.
A sliding surface is then defined using the as follows.
where, is the vector of Hurwitizian polynomial, to which, poles locate in the left half of the complex co-ordinate plane. By using a Lyapunov candidate function is defined as With reference to and it can infer that if the control law is used so that where is the positive coefficient, then is a stable Lyaponov process. Based on this relation together with the equations of (3), (2) and (1), the feedback control signal such that or when can be inferred as below
where,
In equation (4), the functions and need to be determined to specify the control law. For and approximation functions and via the FL presented in [50] are adopted. For estimating the calculating time, or time delay should be paid attention. Therefore, more attention is needed to estimate In [41, 42, 50], in order to avoid the chattering phenomenon, the fuzzy gains were used. In [45], the PI controller was established to keep the state in a limited boundary layer. However, in these previous works the increment of the calculating cost related to the fuzzy structures was occurred as a common difficulty to solve. To resolve this problem, in this work an adaptive gain and a desired boundary layer of are used as follows (refer to theorem 1 in section 4.2 for the details):
where, is the adaptive positive coefficient chosen by designer and is the adaptive coefficient. The update law of is to be performed based on the convergence status of which is addressed in theorem 1. It is noted that in order to avoid the chattering phenomenon, the saturation function defined in [46] with the desired boundary layer Φ of is used instead of the signum function in (5);
Remark 1. Parameter always varies exponentially compared with As a result, if then rapidly to prevent from the chattering phenomenon; inversely, if increases then increases exponentially to keep the stable status of the system.
3. Building of the ANFIS-I-SmD
This section presents the algorithm im-ENFS for establishing the ANFIS-I-SmD shown in figure 3 on the basis of the given IDS. Generally, the IDS has P input–output data samples expressing an unknown mapping f : X → Y from the input data space X to the output data space Y.
3.1. Establishing a JDS
In order to establish a close interaction between X and Y, and increase the accuracy rate of the ANFIS during the clustering process, in this work a new clustering process is performed using a new data space called JDS derived from the IDS The JDS is built as follows. Firstly, the output signals of the IDS are normalized:
Normalized values are then added to the IDS in a new column. The JDS has P rows according to that of the IDS and columns, in which the (n0 + 1)th one is the added column. Thus, each input data sample belongs to while the corresponding output data sample belongs to Based on a defined objective function and a clustering algorithm, the clustering process is then carried out in the JDS. As a result, a CDS, for an ANFIS is established in next subsection.
3.2. Cluster centroids in the JDS
The result of the fuzzy clustering process is the CDS with C cluster centroids in the JDS such that the following objective function is minimized [51]:
subjected to
where, is the jth data sample, and is the ith cluster center in the JDS; denotes membership degree of the jth data sample belonging to the ith cluster; is the fuzzy factor; denotes the squared distance between and calculated in the kernel space. By choosing Gaussian kernel function, and the following equations are obtained.
From (13), the optimal centers are the solution of the following equation:
As a result, the iterative function used to update can be expressed as follows:
Based on (11), (13) and the method of Lagrange multipliers, the optimal values of the memberships are identified. Thus, the following iterative function is obtained for updating:
where, expresses the membership of the jth sample belonging to the ith data cluster. The stop condition of the clustering phase is where index is calculated by
In the above, is the required value of r denotes the order of loops.
Remark 2. As above mentioned, the JDS is only used in the clustering process to determine the cluster centroids belonging to as below.
When the clustering process has accomplished, the (n0 + 1)th column (corresponding to (9)) of these cluster centroids is deleted and they resume to the :
All of the next steps will be carried out based on the cluster centroids (19) and the IDS in .
3.3. Establishing output hyper-planes
The output value of the ANFIS is calculated via output clusters typed hyper-planes, which is the kth hyper-plane constituted by an index vector Firstly, by using the obtained cluster centroids in (19), the membership of each data sample in the IDS belonging to each cluster can be determined by These memberships are then used to specify the hard distribution status of the data samples in each cluster deputized by cluster centroids Namely, the ith sample belongs to the kth cluster if the following equation is satisfied.
Let be the number of data samples being hard distributed to the kth cluster. By using the well-known least mean squares method for the index vector of the hype-plane will be specified (see [27] for more detail). Thus, the value of hype-plane associated with data sample is calculated as below.
3.4. Structure of the ANFIS
The result of the training process provides five layers as shown in figure 5; (i) the data (D) layer: this layer has n0 nodes. (ii) The input cluster layer: the result of the clustering process is C clusters with cluster centroids to which C fuzzy sets, ..., are established. The output of this layer is the membership values of calculated for each direction Based on (16), the membership value of belonging to is calculated as follows:
(iii) The product layer, ∏: each node in this layer calculates the product of membership values of its inputs corresponding to all directions for each fuzzy rule:
(iv) The hype-plane (Hp) layer: this is used to specify hype-planes mentioned in section 3.3. (v) The specifying (S) layer: this one is used to specify the output of the ANFIS via defuzzification functions (DF). DF can come from the well-known methods such as the center-average method or the 'the winner takes all'. If 'the winner takes all' is used, the output signal is calculated as follows:
where,
Now, the building alogoritm for ANFIS can be completed by separating it into three phases: establish a training data set JDS, building a CDS and building ANFIS.
Download figure:
Standard image High-resolution imageEstablish the JDS: this was mentioned in section 3.1.
Initialize: the initial index of loops, r = 1; the number of clusters the initial cluster centroids are chosen randomly in which is any real number larger than the desired value, .
The im-ENFS:
- 1.Calculate based on (16);
- 2.Update cluster centroids using (15);
- 3.Calculate the stop coefficient via (17);
- -Either or ( and go to Step 4;
- -If and setup r =: r + 1 and return to Step 1;
- 4.
- 5.Calculate root-mean-square error (RMSE) and evaluate:
- -If Stop and use the ANFIS as an optimal fuzzy system;
- -If and Stop (the ANFIS cannot converge to );
- -If and then (1) C =: C + 1; (2) establish a new cluster centroid at the kth data sample specified by and (3) return to Step 1.
4. Design of a new FSMC
4.1. Fuzzy appropriation
As given in equation (4), the functions and are approximated via MISO fuzzy systems; n input variables and m fuzzy laws [50]. The ith fuzzy law is written as below.
where, is the fuzzy set in the input space related to the physical parameter and the ith fuzzy law, while is the corresponding fuzzy set in the output space. Using the center-average defuzzification, the output is calculated by
In (28), is the value of the membership function in the input fuzzy space of If the product law of is used to equation (28) becomes as follows.
Value can be determined by the well-known methods via the fuzzy set in the output space, which is the singleton fuzzification. Then, equation (29) is shortly re-expressed as follows.
In the above,
4.2. The proposed control law
Consider the approximation of the functions in (4) by the fuzzy system (27) given in (30). The obtained functions, respectively, are expressed as follows:
where, and are specified by (31). Let and be the optimal vectors of and as follows.
where, and are design parameters. Let be a function defined by
where, is the control law. Assume that is a bounded function; .
Theorem 1. The ISSmDCS (1) is controlled by the following control law.
where, is estimated as in (6); and are specified in (33). Then, when Thus, the proposed control system satisfies the stable Lyaponov process if the following update laws are adopted:
where, is the bound value of defined in (36); are adaptive parameters chosen by designer.
Proof. From equations (3) and (1), the following is given.
Using the control law (37) with reference to (40), (36) and (33), the followings are obtained.
where, Using the following Lyaponov function
with reference to (6), (7) and (43), the following equations are obtained.
In (46), due to and the following equation can be derived.
Then, using update law (38), (47) becomes as follows.
- If from (8) and (48), the followings can be inferred.
In this case, based on (49), the update law of is derived as follows.
- If from (8) and (48), the followings can be inferred.
So, the update law of is obtained as follows.
Finally, deriving from (50) and (52), the update law proposed in (39) can be inferred. □
5. Experimental realization
In oreder to validate the effectiveness of the proposed controller, an experimental apparatus shown in figure 6 is built by integrating with the semi-active MRD suspension system working as an ISSmDCS.
Download figure:
Standard image High-resolution imageThe experimental apparatus consists of four main equipment groups; (a) the suspension system is constituted of the linear spring N m−1 and the MRD with a damping coefficient of Ns m−1 corresponding to the zero current, (b) the hydraulic excitation system is operated by the hydraulic unit, connected to the lower bed, (c) the mechanical structure is constituted of the upper bed, lower bed and four parallel vertical circle pillars used to fix the suspension system, wheel, sprung mass and sensors, (d) the control system consists of a computer, an AD/DA converter, an amplifier, sensors and so on. In this work, the sprung mass is kg is chosen considering the variation.
The excitation from the hydraulic exciter results in displacement of the lower bed which makes the sprung mass to vibrate vertically. The relative displacement between the sprung and unsprung mass is measured by a linear variable differential transformer, while acceleration of the sprung mass is measured by an accelerometer. The signals from the sensors are transmited to the computer via the AD converter. Conversely, the control signal from the computer goes to the MRD via the DA converter and current amplifier. The use of the experimental apparatus shown in figure 6 for the semi-active suspension system can be schemetically shown by figure 7, in which denotes uncertainty of the mathematical model expressing the system. It has two main parts. The first one is the semi-active MRD suspension using the MRD, the damper and the linear spring The second one is the controller used to control the MRD consisting of the ANFIS-I-SmD and the estimator FSMC (refer to figure 4). The chassis mass (or the sprung mass) consists of the load mass including the mass of the upper bed. The constant parameter expresses the unsprung mass consisting of the mass of the lower bed, the wheel, MRD and the equipment for fixing. The vertical displacement of the sprung and unsprung mass is signed and respectively, while that of the road profile is signed In this study, very high values for and are chosen so that .
Download figure:
Standard image High-resolution imageThe state space is expressed via the dynamic response of the sprung mass as follows.
The dynamic response of the sprung mass can be described via Newton Law as follows.
Expressions in (54) can be re-expressed as follows.
where,
The proposed controller is constituted of the estimator FSMC and ANFIS-I-MRD as shown in figure 4. For the FSMC, it is built using (55) in the form of equation (1) as follows.
where, and are unknown functions. These functions and the control law in (58) are estimated based on theorem 1 with parameters of the FSMC given by table 1.
Remark 3. Functions and in (55)–(57) are used to specify the varying ranges of fuzzy systems and (27) which approximate and in (58), respectively.
Table 1. Parameters of the FSMC.
Sliding surface | |
---|---|
k2 | 1.5 |
Ω0 | 75 |
Φ | 1.2 |
2 | |
0.001 | |
Number of fuzzy laws | 49 |
On ther hand, for the ANFIS-I-MRD an experimental apparatus is used to build data sets as shown in figure 8. Based on the measured database and the proposed algorithm im-ENFS presented in section 3, an ANFIS working as the ANFIS-I-MRD is established.
Download figure:
Standard image High-resolution image5.1. Identification results
Using the measured database and the im-ENFS, ANFISs working as the inverse MRD models are established. The results related to this work are presented in figures 9–13 and table 2. Figure 9 shows the time-series prediction output of the ANFISs which is the damping force at different current levels, from 0 to 2.4 A. Identifying an extraction from this database corresponding to I = 1.6 A is illustrated in figure 10. The results from this figure show that the accuracy rate of the ANFIS is high showing a very low RMSE of 0.1652 (N).
Download figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageTable 2. The identification RMSE (N) corresponding each method.
Data-10 | Data-11 | Data-13 | |
---|---|---|---|
OHCS | 1.5218 | 0.3649 | 0.1872 |
OD-T2FLS | 0.9117 | 0.1768 | 0.0871 |
ANFIS-PF | 0.9836 | 0.1495 | 0.0899 |
ENFS | 1.2419 | 0.3832 | 0.1978 |
im-ENFS | 0.1652 | 0.0802 | 0.0241 |
Figure 11(a) illustrates the damping force versus the relative MRD piston-cylinder velocity. This is an extraction from the inverse MRD model expressing The time-series prediction error using the im-ENFS and the data set corresponding to I = 0.8 A as is quite small as shown in figure 11(b). The RMSE in this phase is 0.0802 (N). Figure 12 displays an extraction from the inverse MRD model using ANFIS expressing corresponding to six current levels from 0 to 2.4 A. Figure 13 shows the time-series prediction error of the ANFIS built by the im-ENFS and the data set deriving from figure 12 when I = 0.8 A. The RMSE in this phase is 0.0241 (N). In order to more fully evaluate the effectiveness of the proposed im-ENFS, identification results coming from the im-ENFS and the other algorithms are considered. They are the ENFS [27], ANFIS-PF (synthesizing an ANFIS via the proposed data potential field) [52], optimized hyperplane clustering synthesis [53], OD-T2FLS (optimally designed IT-2FLS from the dataset) [54]. The data sets used for figures 10, 11, and 13 (called Data-10, Data-11, and Data-13, respectively) are re-utilized for these ANFIS training algorithms. The compared RMSEs are listed in table 2 which shows the best identification accuracy rate of the proposed im-ENFS algorithm. Specially, the im-ENFS can deal well with the very large data structure such as the 25.000-data-sample set (Data-10). This is the inherent feature for building inverse smart models typed data-driven ones.
By combining the databases deriving from the illustrated in figure 11 and shown in figure 12, a new database with the input to be and the output to be the corresponding current, is generated. Based on this and the algorithm of the im-ENFS, an inverse MRD model expressing is established which works as the ANFIS-I-MRD of the controller presented in figure 4. Using this ANFIS-I-MRD, the FSMC formulated in this work is experimentally relaized for vibration control of the semi-active vehicle suspension system.
5.2. Vibration control results
In this test, in order to demonstrate the effectiveness of the proposed FSMC, the passive result which is obtained without any control action and the control result achieved from one existing ANFIS associated with the sliding mode controller named NFSmUoC [6] are presented together. It is noted here that all MRD structures and experimental conditions for three cases are same. The aim of the control system is to support the control force u(t) such that vibration of the chassis mass is damped out as soon as possible. This means the required state to be To show quantitative results, the maximum magnitude of the chassis displacement and acceleration are defined as follows.
where, P is the number of samples; and are the vertical chassis displacement and acceleration.
As a first test, the bump road shown in figure 14 is applied. Obtained survey results related to the passive, NFSmUoC and FSMC suspensions are shown in figures 15–17 and table 3. The maximum vertical displacement of the sprung mass shows that the vibration amplitudes of the two controlled suspensions are much smaller than that of the passive one. And it is clearly seen that the amplitude of 0.0020 (m) achieved from the proposed controller FSMC is the smallest. The vertical accelerations of depicted in figure 16 also shows the smallest amplitude (0.5332 m s−2) for the case of the proposed FSMC, while the largest amplitude (4.8402 m s−2) is occurred for the passive case as expected. It is known that the evaluation of control force (or power) is equivalently important to the evaluation of control performances. Figure 17 compares control force applied to the suspension system to get the results shown in figures 15 and 16. It is clear observed that the proposed controller is rquired smaller control force, but provides better vibration control results than the NFSmUoC. The quantitative results for vibration control and required force are identified and lisred in table 3.
Download figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageTable 3. The maximum magnitudes.
(m) | (m s−2) | Control force (N) | |
---|---|---|---|
Passive | 0.0254 | 4.8402 | 0.0 |
NFSmUoC | 0.0025 | 0.9812 | 1225.51 |
FSMC | 0.0020 | 0.5332 | 956.79 |
As a second test, a sinusoidal road profile is adopted as follows.
where, denotes the vertical displacement of the road surface; is the amplitude; D is the traveling length of the vehicle along the road with the speed of corresponding to one sine-wave cycle; denotes a random value belonging to Related to this road profile, describes the road disturbance surface. This implies the unknown shake status of the lower bed in experimental apparatus shown in figure 6. In (60), by choosing and km h−1, the road profile is generated as shown in figure 18. Using this excitation, vibration control results are obtained and presented in figures 19–21. The compared results show that the smallest and of the chassis vibration are obtained from the proposed FSMC. The maximum displacement amplitudes of the passive, NFSmUoC and FSMC, respectively are identified by 0.0253, 0.0058, and 0.0030 m, respectively. The maximum acceleration amplitudes of them, in this order, are evaluated as 2.2297, 0.8067, and 0.4948 m s−2, respectively. In this case, the control force applied to MRD is almost same for the FSMC and NFSmUoC as shown in figure 21. The results presented in this tests indicate that the FSMC proposed in this work can provide the smallest displacement and acceleration amplitudes of the vehicle suspension system which are directly related to the ride comfort of the vehilce.
Download figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution image6. Conclusion
In this work, a new FSMC integrated with ANFIS was formulated for a SmD system featuring a vehilce MR damper in order to improve vibration control proformance of the suspension system subjected to parameter uncertainties, time delay and external disturbances. As a first step, a non-parametric SmD model using ANFIS was developed from which an improved algorithm to identify SmDs named im-ENFS was derived. As a second step, a novel FSMC, was formulated based on an adaptive gain and control laws obtained by the Lyapunov stability analysis. Subsequently, the proposed FSMC was applied to the semi-active MRD vehicle suspension system to evaluate vibration control performance. After verifying that the identification accuracy rate of the im-ENFS is quite good, control implementation was undertaken at two different road profiles; bump and sinusoidal roads. It has been demonstrated that the FSMC proposed in this work can provide better vibration control responses at two different road conditions than one of existing FSMC called NFSmUoC. The reduction of the vertical displacement and acceleration of the suspension system directly indicates the improvement of the ride comfort of the vehicles. The results presented in this work are quite self-explanatory justifying that the proposed FSMC is very robust against parameter variations and hence guarantees the stabilty during control action. It is finally remarked that cost effectiveness of the proposed controller compared with conventional simple controllers such as a sky-hook controller needs to be investigated in terms of the estimation time of uncertainties and the feedback loop time of control action.
Acknowledgments
This work was partially supported by National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (No. 2015R1A2A1A5054000). This financial support is gratefully acknowledged.