mathematics//calculus//Laplace transform
The Laplace transform is an integral transform that maps a function of time \(f(t)\) into a function \(F(s)\) of a complex variable \(s\), and engineers use it because it turns linear differential equations into algebra: in the \(s\) domain a derivative becomes a multiplication, a loop becomes a fraction, and the roots that decide stability become the poles of that fraction. It is the frame in which classical control is written, where blocks are transfer functions and the closed-loop eigenvalues are called poles.
The Laplace transform is an integral transform that maps a function of time f(t)f(t)f(t) into a function F(s)F(s)F(s) of a complex variable sss, and engineers use it because it turns linear differential equations into algebra: in the sss domain a derivative becomes a multiplication, a loop becomes a fraction, and the roots that decide stability become the poles of that fraction. It is the frame in which classical control is written, where blocks are transfer functions and the closed-loop eigenvalues are called poles.
F(s)=∫0∞f(t) e−st dtF(s)=\int_0^\infty f(t)\,e^{-st}\,dtF(s)=∫0∞f(t)e−stdt
Nobody computes the integral in practice; what matters is the short table of what it does. A derivative f˙\dot ff˙ becomes sF(s)−f(0)sF(s)-f(0)sF(s)−f(0), an integral becomes F(s)/sF(s)/sF(s)/s, a delay of τ\tauτ becomes a factor e−sτe^{-s\tau}e−sτ. A drone's altitude, my¨=um\ddot y=umy¨=u, becomes ms2Y=Ums^2Y=Ums2Y=U, so the plant is 1/(ms2)1/(ms^2)1/(ms2); a PID is Kp+Ki/s+KdsK_p+K_i/s+K_dsKp+Ki/s+Kds; closing the loop and setting the denominator 1+CG1+CG1+CG to zero gives ms3+Kds2+Kps+Ki=0ms^3+K_ds^2+K_ps+K_i=0ms3+Kds2+Kps+Ki=0. That is the same cubic found by trying solutions eλte^{\lambda t}eλt in the time domain, with sss in the role of λ\lambdaλ, which is why the poles of a transfer function and the eigenvalues of the state matrix are the same numbers (characteristic polynomial).
Setting s=jωs=j\omegas=jω gives the frequency response. The transform evaluated on the imaginary axis says how a sine of frequency ω\omegaω is amplified and delayed, which is what a Bode plot draws; the Fourier transform is that special case.
The final value theorem answers where does it settle? without simulating: for a stable system, limt→∞y(t)=lims→0sY(s)\lim_{t\to\infty}y(t)=\lim_{s\to0}sY(s)limt→∞y(t)=lims→0sY(s). It is the quick way to show that a loop with an integrator has no steady-state error under a constant load (steady-state error).
Its reach ends at linear, time-invariant systems. A saturating actuator, a sticking valve or a gain that changes with load have no transfer function, and the analysis returns to the time domain; for sampled systems the counterpart is the z-transform, with z=esTz=e^{sT}z=esT (digital control).
The calculus underneath (integrals that converge, inverse transforms by partial fractions) is a tool most engineers delegate to tables and to software such as scipy.signal and python-control (integral).