Tutorial 4 Report
Sonu Kumar(2302MT09)
09/09/25
Output:Part 1:-Simulate the data
`
Part 4: Why does simple linear regression perform poorly for this problem?
- Simple linear regression works on the assumption that the relationship between the
variables is linear. To apply it here, we had to linearize the model by taking the natural
logarithm: P(h) = a * e^(-b*h) -> ln(P) = ln(a) – b*h. While this makes the relationship linear,
it fundamentally changes the distribution of the noise. The original noise (ε) is Gaussian and
additive. When you take the logarithm, the noise term becomes ln(P(h) + ε), which is no
longer simple additive Gaussian noise." This transformation biases the linear regression,
especially where the noise is large or the pressure values are small, leading to less accurate
parameter estimates and a higher MSE
part 4 :- Which method performed better under low and high noise?
As shown by the MSE values: Low Noise: The **Nonlinear Regression** performed slightly
better than the Linear Regression. High Noise: The **Nonlinear Regression** performed
significantly better than the Linear Regression. This is because the nonlinear curve fitting
method directly fits the exponential model to the data, correctly accounting for the nature
of the relationship and the additive Gaussian noise, without the distortion caused by the
logarithmic transformation.
Ans e:-> The final answer is: σ(w)=6FLw/A^2
Ans b:-> the standard objective function is f(w)=39×109w.
Ans c:-> Optimum bending Stress (σ*): 0.04 Pa
Ans c and d :-> --- Optimization Results ---
Optimal Width (w*): -9.9000e-01 m
Optimal Height (h*): -2.0202e-03 m
Optimum Bending Stress (σ*): 0.04 Pa
Ans e:-
Part e: Comment on the practicality of the answer
The mathematical solution from the unconstrained optimization (as found in Parts c and d) is
that the optimal bending stress approaches zero as the beam's width (w) approaches zero
and its height (h) approaches infinity. While this is the correct mathematical answer, it is not
a practical engineering solution.
Instability and Buckling: A beam with an extremely large height-to-width ratio is
highly susceptible to lateral-torsional buckling, a type of structural instability where
the beam collapses sideways. This failure occurs at a much lower stress than what is
predicted by the simple bending stress formula.
Manufacturing and Handling: It is physically impossible to manufacture, transport,
or install a beam with such impractical dimensions.
Ans a:-> a) P_1(w)=max [0, g_1(w)]^2
Ans b:->option a;
Ans c:-> The new objective function, F(w), is formulated by adding the penalty term to the original
objective function, f(w). We also introduce a penalty parameter, lambda, to control the severity of
the penalty.
F(w)=f(w)+lambdaP_1(w)
From Part 1, the original objective function is f(w)=frac6FLwA2. From Part 2a, the penalty
term is P_1(w) = max [0, g_1(w)]2, where g_1(w)=0.02−w.
The reformulated objective function is: F(w)=frac6FLwA2+lambdamax[0,0.02−w]^2
ANS d:->
Ans e:->
The introduction of the width constraint had a significant and positive impact on the
optimization results compared to Part 1.
In Part 1, the unconstrained optimization yielded a mathematically optimal, but
physically impossible, solution. The dimensions were an infinitesimally small width
and an infinitely large height, leading to a theoretical minimum stress of 39 Pa.
In Part 2, the constraint forced the optimizer to find a practical solution that
respected a minimum manufacturable width. The resulting dimensions are a finite,
realistic width of 0.02 m and a corresponding height of 0.1 m.
While the new bending stress (7.8 GPa) is vastly higher than the theoretical
minimum, it represents the lowest possible stress for a beam that meets the practical
width requirement. This shows that real-world engineering design is a trade-off
between idealized performance and physical constraints. The penalty method
successfully "pushed" the solution away from the trivial, non-physical optimum
towards the boundary of the feasible region, yielding a useful design.
Ans a:->
Ans b:->
Ans c:->
--- Optimization with Width and Stress Constraints ---
Optimal Width (w*): 0.0005 m
Optimal Height (h*): 4.0000 m
Optimum Bending Stress (σ*): 19499805.10 Pa or 19.50 MPa
Ans d:-> The introduction of the maximum stress constraint had a profound impact on the final
design compared to Part 2.
Part 2 had only the width constraint (wge0.02textm). The optimization pushed the
solution to the boundary of this single constraint, resulting in a width of 0.02 m and
an associated stress of approximately 7.8 GPa, a dangerously high value that far
exceeds the strength of any typical material. The width constraint alone was not
sufficient to produce a safe design.
Part 3 added a more restrictive and physically meaningful constraint: the beam's
stress could not exceed 12 MPa. Since 12 MPa is significantly lower than the stress of
the Part 2 solution, this new constraint became the governing constraint.
The optimization now finds a solution that satisfies this new, stricter limit. The optimal
design shifted away from the minimum width and found a new, wider, and shorter geometry.
The final result of a width of 0.0513 m and a stress of exactly 12 MPa represents a physically
meaningful and safe design. This demonstrates how optimization for real-world engineering
requires realistic constraints to produce a viable solution.
Ans a:-
Optimal Width (w*): 0.0005 m
Optimal Height (h*): 4.0000 m
Optimum Bending Stress (σ*): 19499805.23 Pa or 19.50 MPa
Ans b:->
