Continuous vs. Discrete Systems: Models, Equations & Simulation

TAKEAWAYS:

• Continuous systems are modeled using differential equations (e.g., Newton’s laws, heat transfer), while discrete systems use difference equations (e.g., population growth, digital filters).
• State-space models (e.g., $\dot{x}=f(x,u)$, $y=g(x,u)$) unify both types, where $x$ = state variables, $u$ = inputs, $y$ = outputs.
• Simulation steps: Discretize time, solve equations numerically (Euler, Runge-Kutta), and validate with real-world data.
• Applications: Continuous = NTC’s power grid stability; Discrete = Pathao’s ride-matching algorithm.
• Trade-offs: Continuous models are precise but computationally heavy; discrete models are faster but may lose accuracy.
• Exam focus: Compare systems, derive equations, and interpret simulation outputs (e.g., stability, steady-state).

---

1. Continuous Systems: Modeling the Real World

Continuous systems evolve smoothly over time (e.g., temperature in a room, voltage in a circuit). Their behavior is described by differential equations (DEs), which relate rates of change (derivatives) to system states.

Key Concepts

• State Variables: Quantities that define the system’s behavior (e.g., position $x(t)$, velocity $v(t)$ of a pendulum).
• Inputs/Outputs: External forces (e.g., torque applied to a pendulum) and measurable responses (e.g., angle $\theta (t)$).
• Ordinary Differential Equations (ODEs): For lumped systems (e.g., $m\ddot{x}+c\dot{x}+kx=F(t)$ for a spring-mass-damper).
• Partial Differential Equations (PDEs): For distributed systems (e.g., heat equation $\frac{\partial T}{\partial t}=\alpha {\nabla }^{2}T$).

Example: Pendulum Motion

Real-world tie: A wind turbine’s blade (continuous rotation modeled as a pendulum with air resistance). Model:

1. State variables: Angle $\theta (t)$, angular velocity $\omega (t)=\dot{\theta }$.
2. Equation of motion (ignoring friction): $$m{L}^2\ddot{\theta }=-mgL\sin \theta$$ Simplified for small angles ($\sin \theta \approx \theta$): $$\ddot{\theta }+\frac{g}{L}\theta =0\text{(Simple Harmonic Oscillator)}$$
3. Solution: $\theta (t)=A\cos ({\omega }_{0}t+\varphi )$, where ${\omega }_0=\sqrt{g/L}$.

Simulation Steps:

1. Discretize time: $t_n=n\Delta t$, $\Delta t=0.01$ s.
2. Use Euler’s method to approximate derivatives: $${\theta }_{n+1}={\theta }_n+\Delta t\cdot {\omega }_n$$ $${\omega }_{n+1}={\omega }_n+\Delta t\cdot \left(-\frac{g}{L}{\theta }_n\right)$$
3. Plot $\theta (t)$ vs. time to visualize oscillations.

Visualization: Exam Tip: Always check units! Here, $g/L$ must be in ${\text{s}}^{-2}$.

---

2. Discrete Systems: Digital and Event-Driven Models

Discrete systems change in jumps (e.g., bank transactions, digital signals). They are modeled using difference equations or state transitions.

Key Concepts

• State Equations: $x_{k+1}=f(x_k,u_k)$, where $k$ is the discrete time step.
• Input-Output Relations: $y_k=g(x_k,u_k)$.
• Z-Transform: Tool to solve difference equations (analogous to Laplace transform for DEs).
• Applications: Digital filters, queueing networks, financial models.

Example: Bank Loan Repayment (Amortization)

Real-world tie: Nabil Bank’s home loan calculator uses discrete compounding to compute monthly payments. Model:

1. State variable: Remaining loan balance $B_k$ at month $k$.
2. Difference equation: $$B_{k+1}=B_k(1+r)-P$$ where:
   • $r=$ monthly interest rate (e.g., 0.5% = 0.005),
   • $P=$ fixed monthly payment.
3. Goal: Find $P$ such that $B_N=0$ after $N$ months.

Worked Example:

• Loan: $100,000, annual interest rate = 6% ($r=0.06/12=0.005$), term = 20 years ($N=240$ months).
• Derivation: The closed-form solution for $P$ is: $$P=\frac{B_0\cdot r(1+r{)}^N}{(1+r{)}^N-1}$$ Plugging in values: $$P=\frac{100000\cdot 0.005\cdot (1.005{)}^{240}}{(1.005{)}^{240}-1}\approx \$659.96$$
• Simulation: Use a spreadsheet to iterate:

  Month $k$	Balance $B_k$	Payment $P$	Interest $B_k\cdot r$	Principal Paid
  0	100,000.00	659.96	500.00	159.96
  1	99,840.04	659.96	499.20	160.76
  ...	...	...	...	...
  240	0.00	659.96	0.00	659.96

Visualization:

---

3. State-Space Representation: Unifying Both Worlds

State-space models represent both continuous and discrete systems in a unified framework: $$\{\begin{array}{ll}\dot{x}(t)=f(x(t),u(t))& \text{(Continuous)}\\ x_{k+1}=f(x_k,u_k)& \text{(Discrete)}\end{array}$$ $$y(t)=g(x(t),u(t))$$

Example: Temperature Control System

Real-world tie: NTC’s smart grid adjusts heater power discretely to maintain room temperature continuously. Model:

1. Continuous dynamics (heat transfer): $$C\dot{T}=-h(T-T_{\text{env}})+P(t)$$ where:
   • $T$ = room temperature,
   • $C$ = thermal capacitance,
   • $h$ = heat loss coefficient,
   • $P(t)$ = heater power (input).
2. Discrete control (thermostat):
   • If $T<T_{\text{set}}$, turn heater on ($P=P_{\text{max}}$).
   • Else, turn off ($P=0$).

Simulation Trace:

Visualization:

---

4. Comparison: Continuous vs. Discrete Systems

---

5. Simulation Techniques

A. Continuous Systems

1. Numerical Methods:
   • Euler’s Method: Simple but inaccurate for stiff systems. $$x_{n+1}=x_n+\Delta t\cdot f(t_n,x_n)$$
   • Runge-Kutta 4th Order (RK4): More accurate. $$k_1=f(t_n,x_n)$$ $$k_2=f(t_n+\Delta t/2,x_n+\Delta t/2\cdot k_1)$$ $$⋮$$ $$x_{n+1}=x_n+\Delta t/6(k_1+2k_2+2k_3+k_4)$$
2. Validation: Compare simulation output to real data (e.g., lab-measured pendulum period).

B. Discrete Systems

1. Iterative Methods:
   • Directly solve $x_{k+1}=f(x_k,u_k)$.
2. Z-Transform:
   • Convert difference equations to algebraic form (e.g., solve $y[k]-0.5y[k-1]=x[k]$).

---

## In the Real World

1. NTC’s Power Grid (Continuous):

   • Uses PDEs to model voltage distribution across transmission lines. Simulations predict blackout risks by solving ${\nabla }^{2}V=0$ (Laplace’s equation) with boundary conditions.
   • Why it matters: Prevents cascading failures like the 2015 Nepal blackout.

2. Pathao’s Ride-Matching (Discrete):

   • Treats driver locations as state variables in a discrete-event simulation. The algorithm:
     1. Models drivers as nodes in a graph.
     2. Uses priority queues to match riders to nearest available drivers.
     3. Updates states (e.g., "driver busy") after each ride.
   • Key equation: $\text{Match}(r,d)$ if $\text{distance}(r,d)<\text{threshold}$ and $d.\text{status}=\text{free}$.

3. Khalti’s Transaction Queue (Discrete + Continuous):

   • Discrete: Processes payments as atomic events (e.g., "transfer $500 from A to B").
   • Continuous: Models network latency as a Poisson process (average 200ms delay).
   • Simulation: Uses queueing theory (M/M/1 model) to predict peak-hour failures.

---

## Exam Tip

1. Derive Equations from First Principles:

   • For continuous systems, start with physics (e.g., $F=ma$) or conservation laws (e.g., mass/energy).
   • For discrete systems, define state transitions clearly (e.g., "if balance > threshold, pay interest").

2. Compare Systems in Tables:

   • Examiners love questions like "How would you model a traffic light system? Continuous or discrete? Justify." Use the comparison table above.

3. Show Simulation Steps:

   • For numerical methods, write out one iteration of Euler/RK4 with sample values (e.g., $\Delta t=0.1$, $x_0=1$).
   • For discrete systems, provide a state transition table (like the loan example).

4. Real-World Applications:

   • Link models to Nepalese contexts:
     • Continuous: "How would you simulate Kathmandu’s traffic flow?" → Use PDEs for vehicle density.
     • Discrete: "Design a queue for Daraz’s order fulfillment." → Use M/G/1 queueing model.

5. Common Pitfalls:

   • Units: Always check (e.g., $g/L$ must be ${\text{s}}^{-2}$).
   • Stability: Continuous systems may blow up if $\Delta t$ is too large in Euler’s method.
   • Discretization Error: Smaller $\Delta t$ = more accurate but slower.

---