3.1. Time-Delay Independent Stability
A time-delay dynamical system is said to be time-delay independent stable if it is asymptotically stable for any given
. According to the theory of time-delay differential equations, we obtain the following characteristic equation [
36]:
where
is the characteristic root of the characteristic Equation (9); and
is the identity matrix. The characteristic Equation (9) is expanded into a polynomial form, which can be expressed as
and
are real coefficient polynomials of the sixth and fourth order. These can be expressed as
where
According to the stability theory, the critical time-delay stability is the characteristic equation with pure imaginary roots . For the time-delay independent stability of Equation (10), the following two conditions need to be satisfied simultaneously:
Condition 1. When, the characteristic polynomialis Routh-Hurwitz stable.
Condition 2. When, the polynomialhas no real root.
For Condition 1, when
, Equation (10) can be expressed as
where
,
,
,
,
,
, and
. According to the Routh–Hurwitz stability criterion, to ensure the asymptotic stability of the system, it is necessary to satisfy that the principal minor of each order of the determinant is greater than zero. The parameters of
Table 3 are substituted into Equation (12), and the Routh–Hurwitz stability dividing line is judged to be obtained, as shown in line R
1 in
Figure 3.
Figure 3 is divided into two parts by lines R
1. The orange and green areas on the upper are the Routh–Hurwitz stability area, the blue area below does not meet the vehicle selection parameters.
For Condition 2, judge whether
has real roots. To facilitate the calculation, the real and imaginary parts of the polynomial can be separated by bringing Euler’s formula
into Equation (10). We can obtain
where
,
,
,
.If Equation (13) is equal to zero, then the real and imaginary parts of the polynomial are zero, respectively. That is,
By solving Equation (14) and eliminating the trigonometric function terms, a polynomial without time-delay
can be obtained:
Let . If Equation (10) has a pair of purely imaginary roots under the condition of , it must be satisfied that has a positive real root. The polynomial does not include the time-delay . Through the above description, we can simplify the solution of whether the polynomial has real roots or not to the solution of whether the polynomial has real roots or not. By this method, the judgment of condition 2 can be completed.
Equation (10) can be expressed as
where
Equation (16) is a 12th order polynomial, and the number of real roots of the polynomial equation
can be determined according to the generalized Sturm criterion. Based on the theory of time-delay independent stability of two-degree-of-freedom vibration system, this paper extends it to the three-degree-of-freedom vibration system. We define the discriminant sequence
as
, i.e.,
. The formula
is shown in
Appendix A.
To draw the time-delay independent stable region, The parameters of
Table 3 are substituted into Equation (16); then, the coefficients
,
,
,
,
, and
in Equation (16) are substituted into the Sturm criterion. We can draw the curve of the discriminant sequence coefficient
. Taking the intersection with the curve satisfying condition 1, we can see that the selected parameter range is divided into 15 regions, as shown in
Figure 3.
Sign tables of the discrimination sequence of the polynomial
are shown in
Table 4.
is the number of non-zero terms of the discriminant sequence and
denotes the number of discriminant sequence number changes, then
has
distinct real roots. If
, the polynomial
has no real roots, which means that the system increases with time delay, and the characteristic roots do not cross the imaginary axis. The stability of the system does not change, so the stability of the system depends on the stability of the system when
. If
, it means that the polynomial
has real roots, i.e., the stability of the system falls outside the time-delay independent stability region, and the stability of the whole system switches a finite number of times with the increase in time delay.
In the regions s
1, s
2 and s
3, the number of non-zero terms in the discrimination sequence is 12 and the number of sign changes is 6, so the polynomial has no real roots. In the regions of s
4~ s
15, the number of non-zero terms in the discrimination sequence is 12 and the number of sign changes is 4, so there are two positive real roots of the polynomial. From the above analysis, it can be concluded that the blue region part of
Figure 3 does not meet condition 1, so it is not Routh–Hurwitz stable at
, and this region is unstable. There are real roots in the orange region, so it is necessary to determine the stability interval according to the stability switches. The green region is the time-delay independent stability. It can be found from
Figure 3 that the time-delay independent stability region is V-shaped. We find that that the time-delay independent stability region expands with the increase in
, which indicates a close relationship between damping and time-delay independent stability. The larger the damping, the larger the time-delay independent stability region.
3.2. Time-Delay Dependent Stability
From the damping determined in
Section 2.2, we obtain
. If the feedback gain
falls within the orange region of
Figure 3, the system switches in stability as the time-delay changes. Assuming that the polynomial
has positive real roots [
37], a set of trigonometric equations can be derived as
where
.
When
, we can solve for the relationship between
and
as follows:
When
,
The critical time-delay can be obtained:
where
i denotes the
i-th positive real root of the polynomial, and
j denotes the
j-th critical time delay corresponding to the
i-th positive real root
. In this way, the relationship between
and
can be obtained as
. Then, the stability interval is determined according to the stability switches theory. The relation between the gain
and
can be obtained from Equation (16),
. According to the above, the relationship between
and
can be obtained as
. Therefore, the time-delay dependent stability region can be drawn as shown in
Figure 4.