The proof of Theorem is similar to those of Proposition 2.1 [
19]. Assume that
\((x_n)\) is a sequence of fuzzy numbers and satisfies FDE (
3) with initial values
\(x_{-1}, x_0.\) Consider the
\(\alpha -\)cuts,
\(\alpha \in (0,1], n\in N^+\)$$\begin{aligned} \left\{ \begin{array}{lll} {[}x_n]_\alpha &{}=&{}[L_{n,\alpha },R_{n,\alpha }], \\ &{}&{}\\ {[}A]_\alpha &{}=&{} [A_{l,\alpha }, A_{r,\alpha }], \\ &{}&{}\\ {[}B]_\alpha &{}=&{} [B_{l,\alpha },B_{r,\alpha }]. \end{array} \right. \end{aligned}$$
(10)
Applying Lemma
2.1, it follows from (
3) and (
10) that
$$\begin{aligned}{}[x_{n+1}]_\alpha= & {} [L_{n+1,\alpha },R_{n+1,\alpha }]=\left[ A+\frac{Bx_n}{x_{n-1}^2}\right] _\alpha =[A]_\alpha +\frac{[B]_\alpha \times [x_n]_\alpha }{[x_{n-1}^2]_\alpha }\\= & {} [A_{l,\alpha },A_{r\alpha }]+\frac{[B_{l,\alpha }L_{n,\alpha },B_{r,\alpha }R_{n,\alpha }]}{[L_{n-1,\alpha }^2, R_{n-1,\alpha }^2]}. \end{aligned}$$
Utilizing g-division of fuzzy numbers, one of the following two cases occurs.
Case (i)
$$\begin{aligned}{}[x_{n+1}]_\alpha =[L_{n+1,\alpha },R_{n+1,\alpha }]=\left[ A_{l,\alpha }+\frac{B_{l,\alpha }L_{n,\alpha }}{L_{n-1,\alpha }^2}, A_{r,\alpha }+\frac{B_{r,\alpha }R_{n,\alpha }}{R_{n-1,\alpha }^2}\right] . \end{aligned}$$
(11)
Case (ii)
$$\begin{aligned}{}[x_{n+1}]_\alpha =[L_{n+1,\alpha },R_{n+1,\alpha }]=\left[ A_{l,\alpha }+\frac{B_{r,\alpha }R_{n,\alpha }}{R_{n-1,\alpha }^2}, A_{r,\alpha }+\frac{B_{l,\alpha }L_{n,\alpha }}{L_{n-1,\alpha }^2}\right] . \end{aligned}$$
(12)
If Case (i) holds true, from (
11), one gets, for
\( \alpha \in (0,1], n\in \{0,1,2,\ldots \},\)$$\begin{aligned} L_{n+1,\alpha }=A_{l,\alpha }+\frac{B_{l,\alpha }L_{n,\alpha }}{L_{n-1,\alpha }^2},\ \ R_{n+1,\alpha }=A_{r,\alpha }+\frac{B_{r,\alpha }R_{n,\alpha }}{R_{n-1,\alpha }^2}. \end{aligned}$$
(13)
It is clear that there is a unique solution
\((L_{n,\alpha },R_{n,\alpha })\) for any initial value
\((L_{k,\alpha },R_{k,\alpha }),k=-1,0,\alpha \in (0,1]\). We only need to show that, for
\(\alpha \in (0,1],\ n\in N^+\),
\([L_{n,\alpha },R_{n,\alpha }]\) makes certain the positive fuzzy solution
\(x_n\) of FDE (
3) with the initial condition
\(x_{i}, i=0,-1\), and
$$\begin{aligned}{}[x_n]_\alpha =[L_{n,\alpha },R_{n,\alpha }]. \end{aligned}$$
(14)
From [
18], since
\(x_j\in \Re _F^+,j=-1,0\), for
\(\alpha _i\in (0,1], i=1,2,\) and
\(\alpha _1\le \alpha _2\), it has
$$\begin{aligned} 0<L_{j,\alpha _1}\le L_{j,\alpha _2}\le R_{j,\alpha _2}\le R_{j,\alpha _1},j=-1, 0. \end{aligned}$$
(15)
We declare that, for
\(n=0,1,2,\ldots ,\)$$\begin{aligned} L_{n,\alpha _1}\le L_{n,\alpha _2}\le R_{n,\alpha _2}\le R_{n,\alpha _1}. \end{aligned}$$
(16)
By mathematical induction. From (
15), for
\(n=0,1\), it is true. Suppose that, for
\(n\le k, k\in N^+\), (
16) holds true, then it follows from (
13) and (
16) that, for
\(n=k+1\),
$$\begin{aligned} L_{k+1,\alpha _1}= & {} A_{l,\alpha _1}+\frac{B_{l,\alpha _1}L_{k,\alpha _1}}{L_{k-1,\alpha _1}^2} \le A_{l,\alpha _2}+\frac{B_{l,\alpha _2}L_{k,\alpha _2}}{L_{k-1,\alpha _2}^2}=L_{k+1,\alpha _2}\\= & {} A_{l,\alpha _2}+\frac{B_{l,\alpha _2}L_{k,\alpha _2}}{L_{k-1,\alpha _2}^2}\le A_{r,\alpha _2} +\frac{B_{r,\alpha _2}R_{k,\alpha _2}}{R_{k-1,\alpha _2}^2}=R_{k+1,\alpha _2}\\= & {} A_{r,\alpha _2}+\frac{B_{r,\alpha _2}R_{k,\alpha _2}}{R_{k-1,\alpha _2}^2}\le A_{r,\alpha _1}+\frac{B_{r,\alpha _1}R_{k,\alpha _1}}{R_{k-1,\alpha _1}^2}=R_{k+1,\alpha _1} \end{aligned}$$
So (
16) holds true. Also, from (
13), it has
$$\begin{aligned} L_{1,\alpha }=A_{l,\alpha }+\frac{B_{l,\alpha }L_{0,\alpha }}{L_{-1,\alpha }^2},\ \ R_{1,\alpha }=A_{r,\alpha }+\frac{B_{r,\alpha }R_{0,\alpha }}{R_{-1,\alpha }^2},\ \ \alpha \in (0,1]. \end{aligned}$$
(17)
Since
\(x_j\in \Re _F^+, j=-1, 0\), and
\(A, B\in \Re _F^+\) , then
\(L_{i,\alpha },R_{i,\alpha },i=0,-1,\) are left continuous. So, one gets from (
17) that
\(L_{1,\alpha },R_{1,\alpha }\) are also left continuous. By inductively, we can show that, for
\(n\ge 1\),
\(L_{n,\alpha }\) and
\(R_{n,\alpha }\) are left continuous.
Secondly, we will prove that
\(\text{ supp } x_n=\overline{\bigcup _{\alpha \in (0,1]}[L_{n,\alpha },R_{n,\alpha }]}\) is compact. It need to show that
\(\bigcup _{\alpha \in (0,1]}[L_{n,\alpha },R_{n,\alpha }]\) is bounded. Since
\(x_j\in \Re _F^+,j=-1,0 \), and
\(A,B\in \Re _F^+\), there exist
\(M_A>0,N_A>0,N_B>0, M_B>0, M_j>0,N_j>0,j=-1,0,\) such that,
\(\forall \alpha \in (0,1]\),
$$\begin{aligned} \left\{ \begin{array}{l} [A_{l,\alpha },A_{r,\alpha }]\subset [M_A,N_A],\\ \\ {[}B_{l,\alpha },B_{r,\alpha }]\subset [M_B,N_B], \\ \\ {[}L_{j,\alpha },R_{j,\alpha }]\subset [M_j,N_j]. \end{array} \right. \end{aligned}$$
(18)
Hence from (
17) and (
18), one can get, for
\(n=1\),
$$\begin{aligned} {[}L_{1,\alpha },R_{1,\alpha }]\subset \left[ M_A+\frac{M_BM_0}{M_{-1}^2},N_A+\frac{N_BN_0}{N_{-1}^2}\right] ,\alpha \in (0,1]. \end{aligned}$$
(19)
From which, it follows that
$$\begin{aligned} \bigcup _{\alpha \in (0,1]}[L_{1,\alpha },R_{1,\alpha }]\subset \left[ M_A+\frac{M_BM_0}{M_{-1}^2}, N_A+\frac{N_BN_0}{N_{-1}^2}\right] ,\ \ \alpha \in (0,1]. \end{aligned}$$
(20)
Hence,
\(\overline{\bigcup _{\alpha \in (0,1]}[L_{1,\alpha },R_{1,\alpha }]}\) is compact, and
\(\overline{\bigcup _{\alpha \in (0,1]}[L_{1,\alpha },R_{1,\alpha }]}\subset (0,\infty )\). Deducing inductively, one can get that
\(\overline{\bigcup _{\alpha \in (0,1]}[L_{n,\alpha }, R_{n,\alpha }]}\) is compact, moreover, for
\(n=1,2,\ldots ,\)$$\begin{aligned} \overline{\bigcup _{\alpha \in (0,1]}[L_{n,\alpha },R_{n,\alpha }]}\subset (0,\infty ). \end{aligned}$$
(21)
And since
\(L_{n,\alpha }, R_{n,\alpha }\) are left continuous, from (
16) and (
21), we get that
\([L_{n,\alpha }, R_{n,\alpha }]\) makes certain a positive sequence
\(x_n\) satisfying (
14).