Published in IET Control Theory and Applications
Received on 6th August 2010
Revised on 29th April 2011
doi: 10.1049/iet-cta.2010.0454
ISSN 1751-8644
Non-iterative identification and model following
control of Hammerstein systems with asymmetric
dead-zone non-linearities
X. Lv X. Ren
School of Automation, Beijing Institute of Technology, Beijing 100081, People’s Republic of China
E-mail: lv_xiaohua@yahoo.cn
Abstract: A novel non-iterative identification algorithm based on the parameterisation of unknown dead-zone non-linearity is
proposed for control of Hammerstein systems in presence of asymmetric dead-zone input. A canonical representation for
piecewise linear functions is employed to describe the dead-zone function such that a universal-type parametric model can be
established to approximate the entire system. Both the dead-zone parameters (thresholds and slopes) and the coefficients in
linear transfer function can be estimated simultaneously using the designed persistent exciting input. A modified model
following control scheme is then designed in such a way th at the plant output tracks the desired output with satisfactory
performance. This method can be applied without a priori knowledge of the dead-zone non-linearity. Numerical simulations
are presented to illustrate the effectiveness of the proposed scheme.
1 Introduction
The Hammerstein system is a cascade model with a zero
memory non-linear element followed by a linear dynamic
element, which has been a widely used block-oriented non-
linear systems [1]. Hammerstein system with asymmetric
dead-zone non-linearity is a useful model in many
mechanical and electrical plants such as dc servo motors
[2], control valves [3] and two-mass mechanical systems
[4]. It is well known that the asymmetric dead-z one is a
non-differentiable function which characterises certain non-
sensitivity for some control inputs. The presence of this
non-linearity can severely limit the performance of control
systems. As the parameters that represent the dead-zone are
usually unknown, system identification constitutes a crucial
part in such control designs.
The asymmetric dead-zone non-linearity, as shown in
Fig. 1, can be fully characterised by four parameters (l
1
, D
1
,
l
2
, D
2
), where l
1
and l
2
are corresponding segment slopes
and D
1
, D
2
,(D
1
, D
2
) are thresholds. Owing to lack of a
priori information on these parameters coupled with the
linear part, identification of Hammerstein systems with this
kind of non-linear element becomes complicated. The early
work to deal with the dead-zone non-linearity was proposed
in [5] where a scheme estimates alternately relevant
parameters and some auxiliary variables [6]. The model of
non-linearity was defined only for inputs of absolute value
greater than the unknown dead-zone and the parameter
estimation had to be performed twice using two
pseudorandom binary signals with different amplitudes [7].
A separable non-linear least square method [8] was
proposed to estimate the symmetrical dead-zone function,
which can be characterised by a single parameter. However,
this method cannot be applied to the non-linearity that is
parameterised by a high-dimensional vector (for instance
(l
1
, D
1
, l
2
, D
2
)) and minimisation of the objective function
is inconvenient and time consuming. A recursive
identification of dead-zone/preload was proposed in [7, 9],
where a decomposition algorithm was employed to obtain
the linear regression model of the system. However,
computation of the unmeasurable internal variables (v(t),
z(t) and x(t)in[9]) increases the complexity of the
algorithm and no consistency analysis for the estimates was
given. To overcome the aforementioned limitations, a new
non-iterative identification algorithm employing a canonical
parameterised model of the unknown non-linearity is
proposed in this paper for control of Hammerstein systems
with dead-zone input non-linearities. The algorithm is
capable of simultaneously estimating coefficients in the linear
transfer function and the dead-zone slopes and thresholds.
Consequently, the compensation of the dead-zone as well as
the non-linear controller can be designed using the obtained
estimates. It should be pointed out that, based on the
parameterisation of the dead-zone, a universal-type
parametric model can be derived to describe the
Hammerstein system without separating the non-linear part
from the linear part. The complexity of the proposed
algorithm is greatly reduced compared with the existing results.
The paper is organised as follows: Section 2 reformulates the
identification problem for the concerned system. The unknown
parameters are estimated in Section 3 and Section 4 discusses
the control scheme of the system. Numerical examples are
presented to illustrate the method in Section 5. Finally, the
paper is ended by some conclusions in Section 6.
84 IET Control Theory Appl., 2012, Vol. 6, Iss. 1, pp. 84– 89
&
The Institution of Engineering and Technology 2011 doi: 10.1049/iet-cta.2010.0454
www.ietdl.org