Motivation behind matrix diagonalisation
One application which I think is quite interesting and basic is to calculate the closed form of a recurrence relation.
Let's say we want to find a closed form (Binet's formula) of the Fibonacci sequence $(F_i)_{i \in \mathbb{N}}$. For this we can look at the equation $$ A \begin{pmatrix} F_n \\ F_{n+1} \end{pmatrix} = \begin{pmatrix} F_{n+1} \\ F_n + F_{n+1} \end{pmatrix} = \begin{pmatrix} F_{n+1} \\ F_{n+2} \end{pmatrix}, \; A = \begin{pmatrix} 0 & 1 \\ 1 & 1 \end{pmatrix}. $$ Applying this for $F_0 = 1$ and $F_1 = 1$ repeatedly we get $$ \begin{pmatrix} F_n \\ F_{n+1} \end{pmatrix} = A^n \begin{pmatrix} 0 \\ 1 \end{pmatrix}, \; n \in \mathbb{N}. $$ We can diagonalize $A=T^{-1}DT$ with a diagonal matrix $D$ and get $$ A^n = (T^{-1}DT)^n = T^{-1}DTT^{-1} DT \dots T^{-1}DT = T^{-1} D^n T. $$ With simple calculations we get $D^n, T$ and $T^{-1}$, so we easily get $A^n$ and $A^n \begin{pmatrix}0 \\ 1 \end{pmatrix}$ gives Binet's formula.
I hope this is an interesting application.
Kind regards!
When addressing people who do not have the necessary mathematical prerequisites to understand the technical details, it is imperative that you locate the point of contact between their daily lives and the topic at hand. When you are successful, then it becomes impossible for them to question the relevance of your talk.
I would open with the famous movie of the Tacoma bridge suffering from wind induced oscillations which end with violent structural collapsed. I would mention that the problem could have been predicted and avoided through computation of the eigenvalues for the stiffness matrix.
I would mention signal analysis and digitalization of sound and images, a topic where Fourier analysis is central.
Time permitting I would mention research into the fundamental properties of matter, i.e., quantum mechanics, the study of which hinges on our understanding of unbounded linear operators, i.e., an extension of your topic.
Instead of technical details, explain that these topics can not be accessed, understood and advanced further, if they do not posses the mathematical skills that you can teach them.
When motivating the diagonalization of matrices, I like to emphasize the following points in order to build an intrinsic narrative of why we are doing such a thing:
A matrix $A$ is not a linear transformation---it is the representation of a linear transformation $T$ with respect to a given basis $\mathcal{B}$. Symbolically, $A = \mathcal{Mat}(T, \mathcal{B})$.
Given a different basis $\mathcal{B}'$, we get a different matrix representing this linear transformation, which is related by a change of basis matrix: $A = P^{-1} B P$ where $B = \mathcal{Mat}(T, \mathcal{B}')$ and $P = \mathcal{Mat}(Id, \mathcal{B}', \mathcal{B}) $. For a non-trivial vector space, there are many possible bases we could choose and many possible matrix representations of a linear transformation. For a freshman in engineering/physics/..., connect this with the idea of choosing different coordinate systems when studying a physical problem. There's no "correct" coordinate system to use, only ones that are more convenient.
A natural question arises: what is the best basis we can pick to study the linear transformation $T$, i.e., in which basis is the matrix representation of $T$ simplest? Ask the students which matrices are easiest to multiply/invert/apply to vectors/etc. Invariably, this will be diagonal matrices. In particular, it is easier to see how a vector will be transformed under the effect of a diagonal matrix.
Thus our goal is to find a basis $\mathcal{B}'$ such that the matrix representation of $T$ is diagonal: $D = \mathcal{Mat}(T, \mathcal{B}')$. Again, this corresponds to finding the "best" coordinate system with which to study a problem.
Combining points (4) and (2), we arrive at $$A = P^{-1} D P.$$
After introducing the eigenvector equation $A \vec{v} = \lambda \vec{v}$ and $\det(A- \lambda I) =0$ and working a few examples, be sure to note that not all matrices are diagonalizeable!
A youtube video by 3blue1brown gives a nice graphical representation of diagonalization (and his entire "Essence of Linear Algebra" series makes excellent recommended viewing for students).
If you prefer to work with specific motivating examples (as opposed to showing how diagonalization of matrices is a natural instrinsic question to ask), considering finding the steady-state solution(s) of a Markov process. In particular, Google's classical PageRank algorithm is a $700B application of finding such a steady-state solution.