📚 Mathematical Modelling: Difficulty and Breakthroughs in Accurate Model Building | 数学建模：准确建立数学模型的难点与突破
Mathematical modelling is the art of translating real-world questions into mathematical language. It is used across science, engineering, finance, and policy, from forecasting weather to pricing options. Building an accurate model is challenging because reality is complex and mathematics is precise; the gap between them must be managed carefully.
数学建模是把现实问题转化为数学语言的艺术。它广泛应用于科学、工程、金融和政策领域，从天气预报到期权定价。建立准确模型并不容易，因为现实复杂而数学精确；两者之间的差距需要被谨慎处理。
1. What Is Mathematical Modelling? | 什么是数学建模？
A mathematical model is a simplified representation of a real system. It uses variables, equations, inequalities, and logical rules to describe relationships between key quantities. For example, a bank might model savings growth as A = P(1 + r)ⁿ, where A is future amount, P is principal, r is interest rate, and n is number of periods.
数学模型是真实系统的简化表达。它使用变量、方程、不等式和逻辑规则来描述关键数量之间的关系。例如，银行可以把储蓄增长建模为 A = P(1 + r)ⁿ，其中 A 是未来金额，P 是本金，r 是利率，n 是期数。
Accurate modelling does not mean making the model exactly like reality. It means capturing the important behaviours with enough precision for the intended purpose, while ignoring irrelevant details. A weather model does not track every air molecule; it tracks pressure, temperature, and wind speed on a coarse grid.
准确建模并不意味着让模型与现实完全一致。它意味着在允许忽略无关细节的前提下，以足够精度捕捉重要行为。天气模型不会追踪每一个空气分子；它只在较粗的网格上追踪气压、温度和风速。
The most successful models are built through a cycle of simplification, solution, validation, and revision. Understanding this cycle is the first step toward overcoming common modelling difficulties.
最成功的模型往往通过“简化—求解—验证—修正”的循环建立。理解这一循环，是克服常见建模困难的第一步。
2. Key Steps in the Modelling Cycle | 数学建模循环的关键步骤
Every modelling project follows a similar structure, even if the details differ. The standard cycle includes:
每个建模项目都遵循类似的结构，即使具体细节有所不同。标准循环包括：
- Define the real problem clearly, including objectives and constraints.
- 明确现实问题，包括目标和约束。
- Make simplifying assumptions and choose relevant variables.
- 作出简化假设并选择相关变量。
- Translate the assumptions into equations or algorithms.
- 把假设转化为方程或算法。
- Solve the mathematics, either analytically or numerically.
- 求解数学问题，可以解析求解或数值求解。
- Validate the model against data or known cases.
- 用数据或已知案例验证模型。
- Interpret the results and, if necessary, refine the model.
- 解释结果，并在必要时修正模型。
This cycle is not a straight line. A model that fails validation may require new assumptions, new variables, or even a new mathematical framework. The cycle should be repeated until the model is good enough for its purpose.
这个循环不是线性的。未通过验证的模型可能需要新的假设、新的变量，甚至新的数学框架。循环应当反复进行，直到模型足够满足其目的。
3. Difficulty 1: Defining the Real-World Problem | 难点一：界定现实问题
Real-world problems are often vague. A company may ask, “How many workers should we hire?” Before any mathematics can begin, this must be transformed into a precise question: over what time horizon? What level of service? What budget? The way the problem is framed completely changes the model.
现实问题往往是模糊的。一家公司可能问：“我们该雇佣多少员工？”在开始任何数学之前，必须先把它转化为精确的问题：时间跨度是多长？服务水平是多少？预算限制是什么？问题界定的方式会彻底改变模型。
Another common issue is confusing symptoms with causes. For example, rising traffic congestion is a symptom; the underlying causes may include population density, poor public transport, or road network design. A model built on symptoms alone will not produce reliable predictions.
另一个常见问题是混淆症状与原因。例如，交通拥堵加剧是症状；根本原因可能包括人口密度、公共交通不足或路网设计不合理。仅基于症状建立的模型不会产生可靠预测。
Breakthrough: write a short problem statement with three parts — objective, variables of interest, and constraints. Test the statement with a simple example. If the answer to a toy version of the problem makes qualitative sense, the framing is probably reasonable.
突破：编写包含三部分的问题陈述——目标、关注变量和约束。用一个简单例子测试该陈述。如果小规模版本问题的答案在定性上合理，那么问题界定可能是合理的。
4. Difficulty 2: Choosing Appropriate Assumptions | 难点二：选择适当的假设
Assumptions are the bridge between reality and mathematics. They can also be the source of serious error. A classic assumption is that demand is linear in price: q = a − bp. This makes calculation easy, but real demand may be nonlinear, seasonal, or influenced by brand loyalty.
假设是连接现实与数学的桥梁，也可能成为严重误差的来源。一个经典假设是需求与价格成线性关系：q = a − bp。这让计算变得容易，但真实需求可能是非线性的、季节性的，或受品牌忠诚度影响。
There are two opposite mistakes in assumption-making. The first is assuming away too much, producing a model that is mathematically elegant but practically useless. The second is trying to include every detail, producing a model that is computationally impossible and impossible to validate.
在作假设时会犯两种相反的错误。第一种是假设得过多，产生一个数学上优雅但实际无用的模型。第二种是试图纳入所有细节，产生一个计算上不可能且无法验证的模型。
Breakthrough: list every assumption explicitly and classify it as essential, convenient, or stylistic. Essential assumptions must be defended. Convenient assumptions can be relaxed later. Stylistic assumptions, such as using units of thousands rather than single digits, should not affect the result. Writing this list forces honesty and makes it easier to revise the model.
突破：明确列出每一个假设，并将其分为本质性、便利性或风格性。本质性假设必须被论证。便利性假设可以在之后放宽。风格性假设，例如以千为单位而不是个位数，不应影响结果。写下这份清单能促使自己诚实，也让修正模型更加容易。
5. Difficulty 3: Selecting the Right Mathematical Structures | 难点三：选择合适的数学结构
Once assumptions are made, the modeller must choose a mathematical structure. The common choices include linear versus nonlinear, deterministic versus stochastic, discrete versus continuous, and static versus dynamic. Each choice carries hidden implications.
一旦假设确定，建模者必须选择数学结构。常见选择包括：线性与非线性、确定性与随机性、离散与连续、静态与动态。每种选择都带有隐含意义。
Linear models are attractive because they are easy to solve and analyse. However, many systems in nature and society are nonlinear: population growth saturates, stress in materials is nonlinear at high levels, and financial returns often show feedback effects. A linear model may work well near a reference point but fail outside that region.
线性模型很吸引人，因为它们易于求解和分析。然而，自然界和社会中的许多系统是非线性的：人口增长会饱和、材料在应力较高时呈非线性、金融收益常常表现出反馈效应。线性模型可能在参考点附近表现良好，但在该区域之外失败。
Breakthrough: use the principle of parsimony — choose the simplest structure that can reproduce the essential qualitative behaviour. Use dimensionless parameters where possible to reduce the number of unknowns. For example, the Reynolds number in fluid mechanics combines several physical properties into one dimensionless value, revealing when flow becomes turbulent.
突破：使用简约原则——选择能够复现基本定性行为的最简单结构。尽量使用无量纲参数以减少未知量。例如，流体力学中的雷诺数把多个物理性质合并为一个无量纲值，揭示流动何时变为湍流。
6. Difficulty 4: Handling Data Uncertainty | 难点四：处理数据不确定性
Data are never perfect. Measurements contain random noise, systematic bias, and missing values. In economic and social models, people do not always behave as recorded, and future events are inherently uncertain. Ignoring uncertainty can make a model seem more precise than it truly is.
数据永远不完美。测量包含随机噪声、系统性偏差和缺失值。在经济和社会模型中，人们的行为并不总是如记录所示，未来事件本质上具有不确定性。忽略不确定性会使模型看起来比实际更精确。
There are several types of uncertainty. Parameter uncertainty arises when we do not know the exact value of a variable, such as the transmission rate of a virus. Structural uncertainty arises when the mathematical form of the model is wrong, such as assuming exponential decay when the true process follows a power law. Scenario uncertainty arises when external conditions change, such as a sudden policy shift.
不确定性有几种类型。参数不确定性出现在我们不知道某个变量的确切值时，例如病毒的传播率。结构性不确定性出现在模型的数学形式错误时，例如假设指数衰减而真实过程遵循幂律。情景不确定性出现在外部条件改变时，例如政策突然转向。
Breakthrough: use probability distributions instead of fixed numbers for uncertain parameters. Then perform a Monte Carlo simulation, drawing many random parameter sets and recording the range of outputs. A model whose predictions are stable across many simulations is more trustworthy than one that depends on a single lucky guess.
突破：对不确定参数使用概率分布而不是固定数值。然后进行蒙特卡洛模拟，抽取许多随机参数组合并记录输出范围。一个在多次模拟中预测稳定的模型，比建立在单次幸运猜测上的模型更可信。
Output = f(parameter₁, parameter₂, … , parameterₙ)
Each parameter can be sampled from a distribution, and the resulting spread of outputs provides a confidence interval for the prediction.
每个参数都可以从某个分布中抽样，得到的输出分布为预测提供了置信区间。
7. Difficulty 5: Recognising Model Limitations | 难点五：认识模型局限性
Every model has a domain of validity. A linear regression trained on data from 2010 to 2020 may not predict well in 2030. A model of a pendulum with small oscillations will fail for large swings. The danger is using a model outside the region where it was calibrated.
每个模型都有适用范围。用2010至2020年数据训练的线性回归，可能无法在2030年作出良好预测。小摆角模型的单摆模型在大摆角时会失效。危险在于把模型用于校准区域之外。
Overfitting is another common failure. If a model has many parameters, it can be made to fit historical data almost perfectly. But fitting noise is not the same as learning the true pattern. The model may then perform badly on new data.
过拟合是另一种常见失败。如果模型参数过多，它几乎可以完美拟合历史数据。但拟合噪声并不等于学习真实模式。该模型在新数据上可能表现糟糕。
Breakthrough: always validate on a data set that was not used to build the model. Use cross-validation or hold-out samples. When no new data are available, test the model on extreme cases: what happens when a variable goes to zero or infinity? Does the behaviour still match physical or logical intuition?
突破：始终在未用于构建模型的数据集上进行验证。使用交叉验证或保留样本。在没有新数据时，用极端情形测试模型：当某个变量趋于零或无穷大时会发生什么？行为是否仍符合物理或逻辑直觉？
8. Breakthrough: Iterative Refinement | 突破：循环改进
No model is perfect on the first attempt. Iterative refinement is the practice of making a rough model, testing it, identifying errors, and then making targeted improvements. This is more effective than trying to build a perfect model at the start.
没有模型能在第一次尝试时就完美。循环改进是先建立粗略模型，进行测试，识别错误，然后作出针对性改进。这比一开始就试图构建完美模型更有效。
A useful refinement strategy is the “skeleton to muscle” approach. Begin with a simple model that captures the core mechanism. For example, model population growth with a constant growth rate r. Once this skeleton behaves well, add a carrying capacity K to represent limited resources, then add random noise to represent environmental variation.
一个有用的改进策略是“从骨架到肌肉”法。先从捕捉核心机制的简单模型开始。例如，用恒定增长率 r 建模人口增长。一旦这个骨架运行良好，再加入承载力 K 来表示有限资源，然后加入随机噪声来表示环境变化。
Each refinement step should be accompanied by a validation check. If adding a new factor does not change the conclusions, that factor can be removed. If it changes the conclusions dramatically, it must be studied carefully.
每一步改进都应伴随验证检查。如果加入一个新因素不改变结论，该因素可以被移除。如果它显著改变结论，则必须被仔细研究。
9. Breakthrough: Sensitivity Analysis | 突破：灵敏度分析
Sensitivity analysis asks: which input parameters have the greatest effect on the output? This knowledge directs attention to the most important data to collect and the most important assumptions to verify.
灵敏度分析要问：哪些输入参数对输出的影响最大？这一信息引导我们把注意力集中在最重要的数据收集和最重要的假设验证上。
One common measure is the sensitivity index. For a model output y depending on parameter p, the index can be written as:
一种常见度量是灵敏度指数。对于依赖于参数 p 的模型输出 y，指数可以写成：
Sₚ = (∂y/∂p) × (p/y)
This dimensionless number tells us the percentage change in y for a one-percent change in p. For example, if Sₚ = 2, increasing p by 1% raises y by about 2%. Such information helps modellers decide where to focus their effort.
这个无量纲数告诉我们，p 变化1%会引起 y 变化百分之几。例如，如果 Sₚ = 2，p 增加1%会使 y 增加约2%。这样的信息帮助建模者决定应把精力集中在哪里。
There are different methods for sensitivity analysis. Local methods compute derivatives near a reference point. Global methods, such as Sobol indices, explore the whole parameter space. Global methods are more reliable for nonlinear models, but they are more expensive to compute.
灵敏度分析有不同的方法。局部方法在参考点附近计算导数。全局方法，如 Sobol 指数，探索整个参数空间。全局方法对非线性模型更可靠，但计算成本更高。
10. Breakthrough: Cross-Disciplinary Thinking | 突破：跨学科思维
Many modelling breakthroughs happen when ideas are borrowed from another field. For example, the famous “random walk” model of stock prices was originally used to describe the motion of pollen particles in water. The mathematics is the same, even though the systems are completely different.
许多建模突破发生在从其他领域借用思想时。例如，著名的股价“随机游走”模型最初用于描述花粉颗粒在水中的运动。尽管系统完全不同，但数学是相同的。
Economists use differential equations borrowed from physics to model economic growth. Biologists use game theory, originally invented for economics, to explain animal behaviour. Engineers use queueing theory, developed for telephone networks, to design airport check-in counters.
经济学家使用借自物理学的微分方程来建模经济增长。生物学家使用最初为经济学发明的博弈论来解释动物行为。工程师使用为电话网络开发的排队论来设计机场值机柜台。
Breakthrough: when facing a difficult modelling problem, ask what analogous systems in other fields have similar behaviour. Try to identify the underlying mechanism rather than the surface description. Drawing a diagram of cause-and-effect relationships can reveal hidden similarities to known models.
突破：当面对困难的建模问题时，询问其他领域中哪些类似系统具有相同行为。尝试识别底层机制，而不是表面描述。画出因果关系图可以揭示与已知模型的隐藏相似性。
11. Worked Example: Population Dynamics | 案例：种群动态模型
Let us apply these ideas to a classic problem: modelling the population of fish in a lake. A naive model uses exponential growth:
让我们把这些思想应用到一个经典问题：建模湖泊中鱼类的种群数量。一个朴素的模型使用指数增长：
Nₙ₊₁ = Nₙ + r Nₙ
Here Nₙ is the population in year n, and r is the growth rate. If r = 0.1, the population grows by 10% each year. This model is simple but unrealistic: it predicts unlimited growth, while the lake has limited food and space.
这里 Nₙ 是第 n 年的种群数量，r 是增长率。如果 r = 0.1，种群每年增长10%。这个模型简单但不现实：它预测无限增长，而湖泊的食物和空间有限。
First breakthrough: add a carrying capacity K. The logistic model becomes:
第一个突破：加入承载力 K。逻辑斯蒂模型变为：
Nₙ₊₁ = Nₙ + r Nₙ (1 − Nₙ / K)
When Nₙ is small, the term (1 − Nₙ / K) is close to 1, so growth is almost exponential. As Nₙ approaches K, growth slows and eventually stops. This simple change eliminates the unrealistic unlimited growth.
当 Nₙ 较小时，(1 − Nₙ / K) 接近1，所以增长几乎是指数式的。当 Nₙ 接近 K 时，增长减缓并最终停止。这一简单改变消除了不现实的无限增长。
Second refinement: include environmental noise. Weather, disease, and human fishing all affect the growth rate. We can replace the fixed r with a random variable:
第二个改进：加入环境噪声。天气、疾病和人类捕捞都会影响增长率。我们可以用随机变量取代固定的 r：
rₙ = r + εₙ, εₙ ~ N(0, σ²)
Each year’s growth rate is drawn from a normal distribution with mean r and standard deviation σ. The population now fluctuates in a natural-looking way, and the model can be used to estimate the probability of population collapse under different management policies.
每一年的增长率从均值为 r、标准差为 σ 的正态分布中抽取。种群现在以自然的方式波动，模型可用于估计在不同管理政策下种群崩溃的概率。
Finally, the model is tested against real fish counts. If its predictions fall within the observed range, it is accepted for limited use. If not, the assumptions must be examined again — perhaps migration from other lakes is important, or the carrying capacity changes over time.
最后，模型要用真实鱼群数量检验。如果预测落在观察范围内，模型就被接受用于有限用途。如果没有，需要重新审视假设——也许其他湖泊的迁移很重要，或者承载力随时间变化。
12. Conclusion | 结论
Accurate mathematical modelling is difficult because it requires balancing simplicity and realism, precision and uncertainty, and mathematical beauty and practical usefulness. The key difficulties are problem definition, assumption choice, mathematical structure, data uncertainty, and model limitations.
准确的数学建模之所以困难，是因为它需要在简单与真实、精确与不确定、数学之美与实际效用之间取得平衡。关键难点包括问题界定、假设选择、数学结构、数据不确定性和模型局限性。
There is no single formula for success. However, the breakthroughs described above — clear problem statements, explicit assumptions, parsimonious structures, uncertainty awareness, honest validation, iterative refinement, sensitivity analysis, and cross-disciplinary thinking — provide a practical toolkit. They transform modelling from a guessing game into a disciplined, learnable skill.
成功没有单一公式。然而，上述突破——清晰的问题陈述、明确的假设、简约的结构、不确定性意识、诚实的验证、循环改进、灵敏度分析和跨学科思维——提供了一套实用工具。它们把建模从猜测游戏转变为一门有纪律、可学习的技能。
Ultimately, a model is not a substitute for reality. It is a lens that helps us see patterns, test ideas, and make better decisions. The goal of modelling is not perfection, but understanding.
归根结底，模型不是现实的替代品。它是一副帮助我们观察模式、测试想法并作出更好决策的透镜。建模的目标不是完美，而是理解。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
屏轩国际教育cambridge primary/secondary checkpoint, cat4, ukiset,ukcat,igcse,alevel,PAT,STEP,MAT, ibdp,ap,ssat,sat,sat2课程辅导，国外大学本科硕士研究生博士课程论文辅导Cancel reply