📚 Mathematical Modeling: Steps and Methods for Building Mathematical Models | 数学建模：建立数学模型的步骤与方法
Mathematical modeling is the art of translating real-world problems into mathematical language, solving them with mathematical tools, and then interpreting the results back in the context of the original problem. It is a core skill in applied mathematics, economics, engineering, and the natural sciences.
数学建模是将现实世界中的问题转化为数学语言，运用数学工具加以解决，再将结果还原到原始问题情境中的一门艺术。它是应用数学、经济、工程和自然科学领域中的核心技能。
For A-level and IB students, understanding the modeling cycle is essential not just for exams but for developing genuine problem-solving ability. This article walks through the complete modeling process step by step, with concrete examples and practical methods.
对于A-level和IB学生而言，理解建模循环不仅能应对考试，更能培养真正的解决问题的能力。本文将逐步讲解完整的建模流程，并配有具体实例和实用方法。
1. Problem Identification and Definition | 问题识别与定义
The first and most crucial step is to understand the problem thoroughly. A vague problem leads to a vague model. You must ask: What exactly needs to be determined? What variables are involved? What are the constraints? What level of accuracy is required?
第一步也是最关键的一步，是彻底理解问题。含糊的问题必然导致模糊的模型。你必须问自己：究竟需要确定什么？涉及哪些变量？有哪些约束条件？要求的精度是多少？
For example, if asked to "model the growth of a population," you must clarify: Is it a single species? Is there limited food? Is the time scale in days or years? These choices fundamentally shape the model.
例如，若要求"建立种群增长模型"，你必须明确：是单一物种吗？食物是否有限？时间尺度是天还是年？这些选择在根本上决定了模型的形态。
Key questions: What? Why? For whom? With what precision?
关键问题：是什么？为什么？为谁建模？精度要求如何？
2. Assumptions and Simplification | 假设与简化
No model can capture every detail of reality. Making reasonable assumptions is what turns a messy real-world situation into a tractable mathematical problem. Good assumptions are transparent, justifiable, and clearly stated before any mathematics begins.
没有任何模型能捕捉现实中的所有细节。做出合理的假设，才能将庞杂的现实情境转化为可处理的数学问题。好的假设应当是透明的、有依据的，并且在任何数学运算开始之前就明确陈述。
Common assumptions include linearity, homogeneity, constant rates, perfect competition, or negligible friction. For instance, the simple exponential growth model assumes a constant per-capita growth rate and unlimited resources — obviously unrealistic, yet useful for short-term predictions.
常见的假设包括线性关系、均匀性、恒定速率、完全竞争或忽略摩擦等。例如，简单的指数增长模型假设人均增长率恒定且资源无限——这显然不现实，但在短期预测中仍然有用。
- State assumptions explicitly before deriving equations
- Justify each assumption with data, observation, or accepted theory
- Rank assumptions by their impact on the result
- 在推导方程之前明确陈述假设
- 用数据、观察或公认理论为每个假设提供依据
- 按假设对结果的影响程度进行排序
3. Choosing the Mathematical Framework | 选择数学框架
Once assumptions are fixed, you must select the appropriate mathematical tools. The choice depends on the nature of the problem: discrete or continuous, deterministic or stochastic, static or dynamic, linear or nonlinear.
一旦假设确定，你必须选择适当的数学工具。选择的依据在于问题的本质：离散还是连续、确定性还是随机性、静态还是动态、线性还是非线性。
For A-level mathematics, the most common frameworks include algebra and functions, calculus (differentiation and integration), differential equations, probability and statistics, and geometry/trigonometry. More advanced modeling may draw on matrices, optimization, or numerical methods.
对于A-level数学，最常见的框架包括代数和函数、微积分（微分与积分）、微分方程、概率与统计，以及几何/三角学。更高级的建模可能借助矩阵、优化或数值方法。
| Problem nature | Typical framework |
| Growth / decay over time | Exponential or logistic functions |
| Rate of change relationships | Differential equations |
| Optimizing profit / cost | Differentiation / linear programming |
| Uncertainty and risk | Probability distributions |
| Periodic phenomena | Trigonometric models |
| 问题本质 | 典型框架 |
| 随时间增长/衰减 | 指数或逻辑斯蒂函数 |
| 变化率关系 | 微分方程 |
| 优化利润/成本 | 微分法/线性规划 |
| 不确定性与风险 | 概率分布 |
| 周期现象 | 三角函数模型 |
4. Formulating the Mathematical Model | 建立数学模型
This is the heart of the process: translating your verbal assumptions into mathematical statements. You define variables, set up equations, and specify initial or boundary conditions. Precision in notation at this stage prevents confusion later.
这是整个过程的核心：将你的文字假设转化为数学陈述。你定义变量、建立方程，并指定初始条件或边界条件。在这一阶段使用精确的记号可以避免后续的混乱。
Consider a classic example — a savings account with annual interest rate r compounded continuously. Let P(t) represent the balance after t years. The model states that the rate of change of the balance equals the interest earned:
考虑一个经典例子——年利率为 r 的连续复利储蓄账户。设 P(t) 表示 t 年后的余额。该模型表述为：余额的变化率等于所赚取的利息：
dP/dt = rP, P(0) = P₀
This ordinary differential equation is the model. It is simple, but it captures the essential dynamics of continuous compounding.
这个常微分方程就是模型。它虽然简单，却抓住了连续复利的基本动态。
5. Solving the Model | 求解模型
With the model formulated, the next task is to obtain a solution. This may involve analytical methods (exact solutions via algebra, calculus, or known formulas) or numerical methods (approximation using computers when exact solution is impossible).
模型建立之后，下一个任务就是求解。这可能涉及解析方法（通过代数、微积分或已知公式获得精确解）或数值方法（当精确解不可行时，借助计算机进行近似求解）。
For the differential equation above, separation of variables gives the exact solution:
对于上述微分方程，分离变量法给出精确解：
P(t) = P₀ · eʳᵗ
This is the familiar exponential growth formula. In other contexts — for example, the three-body problem in physics — no closed-form solution exists, so numerical methods like Euler's method or Runge–Kutta are used.
这就是熟悉的指数增长公式。在其他情境中——例如物理学中的三体问题——不存在闭式解，因此需要使用欧拉方法或龙格-库塔法等数值方法。
6. Verification and Validation | 验证与检验
A model is only useful if it accurately reflects reality. Verification checks that the mathematics is implemented correctly (no algebra mistakes, correct coding); validation checks that the model actually fits real-world data. Both are essential.
模型只有准确反映现实才有价值。"验证"检查数学是否被正确实施（没有代数错误、代码正确）；"检验"检查模型是否真正拟合现实数据。两者都至关重要。
Common validation methods include comparing model predictions with independent data, computing residuals (observed minus predicted values), and assessing the coefficient of determination R² for regression models. If the model fails validation, return to earlier steps and revise.
常见的检验方法包括：将模型预测与独立数据进行比较、计算残差（观测值减去预测值），以及评估回归模型的决定系数 R²。如果模型未通过检验，应回到前面的步骤进行修正。
Residual = Observed value − Predicted value
残差 = 观测值 − 预测值
7. Sensitivity Analysis | 敏感性分析
Sensitivity analysis examines how changes in model parameters affect the output. A model that produces wildly different results for tiny parameter changes is fragile and may be unreliable in practice.
敏感性分析考察模型参数的变化如何影响输出结果。如果参数的微小变化会导致结果的巨大差异，那么该模型是脆弱的，在实践中可能不可靠。
determines the long-term population ceiling. To test sensitivity, you might vary K by ±10% and observe the percentage change in the predicted population at a future time.
例如，在逻辑斯蒂增长模型中，环境容纳量 K 决定了长期的种群上限。为了测试敏感性，你可以将 K 变动 ±10%，并观察未来某一时刻预测种群数量的百分比变化。
Mathematically, sensitivity can be quantified as the partial derivative of the output with respect to a parameter, often normalised into an elasticity:
在数学上，敏感性可以用输出对参数的偏导数来量化，通常将其归一化为弹性：
S = (∂Y / Y) / (∂p / p)
Elasticity S tells you the percentage change in output Y resulting from a 1% change in parameter p. Understanding which parameters dominate the system helps prioritise data collection and model refinement.
弹性 S 告诉你：参数 p 每变化 1%，输出 Y 变化百分之几。了解哪些参数主导系统，有助于确定数据收集和模型改进的优先级。
8. Interpretation and Communication | 结果解读与沟通
The final mathematical solution is meaningless until it is translated back into the original context. This requires interpreting variables and parameters in domain terms, stating limitations, and communicating findings in accessible language.
最终的数学解只有在被转化回原始情境后才具有意义。这需要从领域角度解读变量和参数、说明局限性，并以通俗易懂的语言传达结论。
For instance, if the model predicts that a disease outbreak will peak in 23 days, the health authority needs that answer in plain terms — not just as a parameter of a differential equation. The modeler must also state assumptions clearly: "This assumes no vaccination and no behaviour change."
例如，如果模型预测疫情将在23天后达到峰值，卫生部门需要以简明的方式获得这个答案——而不仅仅是某个微分方程的参数。建模者还必须清晰陈述假设："这假设没有疫苗接种，也没有行为改变。"
| Mathematical result | Real-world interpretation |
| t* = 23 days (max of I(t)) | Outbreak peaks in 23 days |
| R₀ = 2.5 | Each case infects 2.5 others on average |
| 数学结果 | 现实解读 |
| t* = 23天（I(t)的最大值） | 疫情在23天内达到峰值 |
| R₀ = 2.5 | 每个病例平均传染2.5人 |
9. Iteration and Model Refinement | 迭代与模型改进
Mathematical modeling is rarely a one-shot process. The first model is typically too simple or too complex. Based on validation results and new data, you refine assumptions, add or remove variables, adjust parameters, and re-solve.
数学建模很少是一次性的过程。第一个模型通常过于简单或过于复杂。根据检验结果和新数据，你修正假设、增加或删减变量、调整参数，然后重新求解。
This iterative loop continues until the model achieves acceptable accuracy for its intended purpose. The key is knowing when to stop — over-refinement can make a model unnecessarily complex without meaningful improvement in performance.
这一迭代循环持续进行，直到模型达到其预期目的所能接受的精度。关键在于知道何时停止——过度细化可能使模型变得不必要地复杂，却没有带来实质性的性能提升。
- Start with the simplest reasonable model
- Test it against data
- Identify systematic discrepancies
- Refine only where discrepancies are significant
- Repeat until accuracy is "good enough"
- 从最简单的合理模型开始
- 用数据对其进行检验
- 找出系统性偏差
- 只在偏差显著处进行改进
- 重复直到精度"足够好"
10. Worked Example: Modeling a Cooling Coffee Cup | 实例分析：咖啡杯冷却模型
Let us apply the full modeling cycle to a concrete problem. A cup of coffee is initially at 90°C. Room temperature is 20°C. How long will it take for the coffee to cool to 60°C?
让我们将完整的建模循环应用于一个具体问题。一杯咖啡初始温度为90°C。室温为20°C。咖啡冷却到60°C需要多长时间？
Step 1 — Problem: Predict the cooling time from 90°C to 60°C.
步骤1—问题：预测从90°C冷却到60°C所需的时间。
Step 2 — Assumptions: The cup is well-mixed (uniform temperature); room temperature is constant; no evaporation or other heat loss beyond Newton's law; the cooling rate is proportional to the temperature difference.
步骤2—假设：杯子被充分搅拌（温度均匀）；室温恒定；除牛顿冷却定律外无蒸发或其他热损失；冷却速率与温差成正比。
Step 3 — Framework: Newton's law of cooling leads to a first-order linear differential equation.
步骤3—框架：牛顿冷却定律引出一个一阶线性微分方程。
dT/dt = −k(T − Tenv), T(0) = 90
Step 4 — Solve: Separation of variables yields the exact solution:
步骤4—求解：分离变量法给出精确解：
T(t) = 20 + 70 · e−kt
Step 5 — Calibrate: Suppose after 5 minutes the coffee is at 70°C. Then 70 = 20 + 70·e−5k, so e−5k = 50/70 = 5/7, giving k = −ln(5/7)/5 ≈ 0.0673 min⁻¹.
步骤5—校准：假设5分钟后咖啡温度为70°C。那么 70 = 20 + 70·e−5k，所以 e−5k = 50/70 = 5/7，得到 k = −ln(5/7)/5 ≈ 0.0673 分钟⁻¹。
Step 6 — Predict: Set T(t) = 60 and solve: 60 = 20 + 70·e−kt, so e−kt = 4/7. Thus t = ln(7/4)/k ≈ ln(1.75)/0.0673 ≈ 8.32 minutes.
步骤6—预测：令 T(t) = 60 并求解：60 = 20 + 70·e−kt，所以 e−kt = 4/7。因此 t = ln(7/4)/k ≈ ln(1.75)/0.0673 ≈ 8.32 分钟。
Step 7 — Validate: You would test the model against additional measured temperatures at later times. If the model consistently over-predicts, evaporation may be significant, and the model should be refined.
步骤7—检验：你需要在更晚的时间点测量更多温度数据来检验模型。如果模型始终高估，说明蒸发可能不可忽略，则应对模型进行改进。
11. Common Pitfalls | 常见误区
Students frequently make avoidable errors in modeling. Being aware of these pitfalls strengthens the quality of any model.
学生在建模时经常会犯一些可以避免的错误。意识到这些误区能够提升任何模型的质量。
- Overcomplicating: using advanced mathematics when simple functions suffice
- Under-specifying: failing to state assumptions clearly
- Ignoring units: mixing m and cm or s and min leads to nonsense results
- Overfitting: fitting a 9th-degree polynomial to 10 data points
- Extrapolating blindly beyond the data range without justification
- Confusing correlation with causation in statistical models
- 过度复杂化：简单函数已经足够时却使用高等数学
- 设定不足：没有清晰陈述假设
- 忽略单位：混用米和厘米，或秒和分钟，导致荒谬的结果
- 过拟合：用9次多项式拟合10个数据点
- 没有依据地盲目外推至数据范围之外
- 在统计模型中混淆相关性与因果关系
12. Tools and Technology | 工具与技术
Modern modeling relies heavily on computational tools. Spreadsheets are excellent for simple regression and iterative calculations. Graphing software (Desmos, GeoGebra) helps visualise functions and data. Statistical packages (R, Python, Minitab) handle more sophisticated analyses and simulations.
现代建模高度依赖计算工具。电子表格非常适合简单回归和迭代计算。绘图软件（如Desmos、GeoGebra）有助于将函数和数据可视化。统计软件包（如R、Python、Minitab）则处理更复杂的分析和模拟。
Students should be able to work fluently across these tools — from quick exploratory sketches in Desmos to scripting full simulations in Python. The mathematical thinking, however, always comes first; the tool merely executes the plan.
学生应该能够熟练地跨这些工具工作——从在Desmos中快速绘制探索性草图，到在Python中编写完整的模拟脚本。然而，数学思维永远是第一位的；工具仅仅执行方案。
Mathematics is the engine; technology is the steering wheel.
数学是引擎，技术是方向盘。
In summary, the modeling cycle consists of problem definition, assumptions, framework selection, formulation, solution, validation, sensitivity analysis, interpretation, and iteration. Mastery of this cycle distinguishes a successful modeler from a mere equation-solver.
总而言之，建模循环包括问题定义、假设、框架选择、公式化、求解、检验、敏感性分析、结果解读和迭代。对这一循环的掌握，能区分出真正成功的建模者与仅仅会解方程的人。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图