Engineering Optimisation

Highly specialised requests are first checked for feasibility from this information. Please do not send confidential data through the public form; describe it first and agree secure transfer afterwards.

1. 01Enquiry and requirement assessment
2. 02Technical scoping and proposal
3. 03Execution
4. 04Quality review and delivery, with an agreed round of revisions

Read the full project workflow.

In depth: how the work is done

1. Formulation and scaling

   We translate the engineering goal into an objective function, decision variables with bounds and constraints, distinguishing hard constraints from preferences. Variables are scaled to comparable magnitudes, which is important for numerical behaviour. Redundant variables and constraints are removed, and units are checked. We also ask whether the model is differentiable, noisy or discontinuous, and how long it takes to evaluate, since these properties decide what algorithms are practical.

2. Gradient-based methods

   When the model is smooth and gradients are available or can be estimated, gradient-based methods such as sequential quadratic programming or interior-point methods converge quickly, even for many variables. They find local optima, so multiple starting points are used to explore different regions. Analytical, automatic or adjoint gradients are preferred to finite differences, which can be noisy and costly. Convergence criteria and the interpretation of optimality conditions are documented.

3. Derivative-free and global strategies

   Where gradients are unavailable or the response is noisy, discontinuous or multimodal, derivative-free methods such as pattern search, Nelder-Mead variants and evolutionary strategies are used. Global strategies such as multi-start, branch-and-bound for suitable problems, or population-based search address multiple optima. They require more evaluations, so surrogate models can reduce the cost. The choice reflects the evaluation budget and the penalty for missing a better optimum.

4. Handling constraints

   Constraints are handled by methods that keep iterates feasible, by penalties that discourage violation or by repair strategies. Penalty parameters must be chosen with care to avoid ill-conditioning or solutions that slightly violate the constraints. Active constraints at the optimum identify the bottlenecks of the design, and Lagrange multipliers or sensitivity values show how much the objective would improve if a limit were relaxed, a valuable piece of design information.

5. Sensitivity and robustness of the optimum

   An optimum that is sensitive to uncertain parameters is fragile. We examine how the optimum and the objective change with assumptions such as material properties, costs or operating conditions, and where relevant formulate robust or reliability-based versions of the problem. Verification with an independent evaluation model guards against exploiting numerical artefacts. Results are presented with the trade-offs and assumptions needed to apply them responsibly.

Common pitfalls this service helps avoid

• Poor scaling. Variables of very different magnitude slow or derail algorithms and obscure convergence.
• Single starting point. Local methods may converge to inferior optima unless several starts are tried.
• Trusting model-optimal designs. Optimisers exploit modelling weaknesses; independent verification is needed.
• Treating soft preferences as hard constraints. Over-constraining reduces the feasible region and hides useful trade-offs.

What good reporting looks like

An optimisation report states the engineering goal, formulation with objective, variables, bounds and constraints, scaling, the evaluation model and its cost, the algorithms used with settings, convergence history, the optimum with active constraints and sensitivity, results from multiple starts or runs, verification and robustness checks, and limitations. It documents the model assumptions that the optimum depends on and avoids claiming global optimality unless it has been shown.

An illustrative scenario

Hypothetical example written for explanation. It is not a client case, a result or a testimonial.

Consider a researcher selecting operating temperature and flow rate of a process to minimise energy use subject to a purity requirement. The first formulation yields an optimum at the edge of the allowed temperature range, which the researcher recognises as an assumption about equipment limits. After checking the bound and scaling the variables, multiple starting points converge to the same optimum. Sensitivity analysis shows the optimum is robust to a plausible change in feed composition but not to a change in heat-exchanger efficiency, which the report flags for measurement.

Questions to settle before you begin

• Which variables can I really change, and over what ranges?
• Which limits are absolute and which are preferences?
• How expensive is one evaluation?
• Is the model smooth, noisy or discontinuous?
• Which assumptions would change the optimum if they were wrong?

Key terms

Objective function

The quantity to be minimised or maximised.

Feasible region

The set of designs that satisfy all constraints.

Local optimum

A solution better than all nearby ones but not necessarily the best overall.

Lagrange multiplier

A sensitivity of the optimum to a constraint.

Derivative-free method

An algorithm that does not require gradient information.

Working together on Engineering Optimisation & Decision Science projects

Optimisation engagements begin with formulation: objectives, decision variables, constraints and the model or experiment used to evaluate designs. Much of the value is created at this stage, because poorly posed problems give unusable optima. We ask for the evaluation model or data, bounds on variables, constraints and any manufacturing or operational limits. Algorithms are chosen to suit evaluation cost and problem structure, and results are checked against independent evaluations and robustness tests. Reports present trade-offs rather than a single answer where objectives conflict, and they state assumptions behind the model that an optimiser could exploit.

Few projects need a single service in isolation. These are the services most often combined with this one.

• Multi-Objective Optimisation

  Optimisation with conflicting objectives, such as cost, weight and performance, producing Pareto-optimal trade-off sets rather than a single answer.

• Genetic Algorithm Optimisation

  Application and tuning of genetic algorithms to engineering problems with nonlinear, discontinuous or mixed-variable search spaces.

• Particle Swarm Optimisation

  Use of particle swarm and related swarm-intelligence methods for continuous engineering optimisation and parameter estimation, with attention to tuning and statistical comparison.

• Bayesian Optimisation

  Sample-efficient optimisation of expensive or noisy black-box functions using probabilistic surrogate models and acquisition functions.

• Topology & Design Optimisation

  Computational determination of material layout and form within a design space to meet performance targets under constraints, with attention to manufacturability.

• Operations Research & Engineering Decision Models

  Mathematical-programming, simulation and decision-analysis models for scheduling, allocation, logistics and engineering decision problems under uncertainty.