Applications of Simulation in Supply Chain Management

Summary

本次直播活动由麻省理工学院（MIT）的供应链管理硕士课程主办，Miguel Rodriguez Garcia和Paulo Sosa共同主持。活动邀请了拥有超过八年供应链管理经验的工业工程师Yashar Ahmadov作为嘉宾演讲者。Yashar目前是亚马逊的高级仿真数据科学家，他利用仿真、数学优化和数据科学工具解决复杂的供应链问题。直播内容包括Yashar关于仿真的介绍、在库存管理、运输和系统动力学中的应用，以及一个实时演示，展示了如何在短时间内创建供应链的可视化仿真。此外，还讨论了何时使用以及何时避免使用仿真，并提供了相关工具和技能的建议。

Takeaways

• 🎓 这是麻省理工学院（MIT）供应链管理微硕士课程的第二次也是最后一次现场活动，由Miguel Rodriguez Garcia和Paulo Sosa主持。
• 👨‍🏫 Miguel和Paulo分别负责SCx供应链基础和SCx供应链动态课程，并鼓励参与者完成微硕士项目以申请MIT或其他大学的供应链管理混合硕士项目。
• 🤖 嘉宾演讲者Yashar Ahmadov拥有超过八年的供应链管理工作经验，目前在亚马逊担任高级仿真数据科学家，使用仿真、数学优化和数据科学工具解决复杂的供应链问题。
• 🔍 Yashar介绍了供应链中仿真的概念，包括确定性与随机性仿真、静态与动态仿真、连续与离散仿真等不同类型的仿真。
• 📈 仿真的优势在于增加现实感、研究危险系统、瓶颈分析、回答“如果……会怎样”的问题、结果可复现以及易于与管理层沟通。
• 🚫 然而，在某些情况下不应使用仿真，例如当可以使用常识分析，或者当系统行为无法验证，或者过度承诺而无法满足项目期望时。
• 🛠️ Yashar推荐了一款名为Analogic的基于Java的仿真软件，它提供了GIS地图、空间标记工具、2D/3D仿真以及特定行业的库。
• 📊 在库存管理中，仿真可以帮助进行需求预测、安全库存优化、订单点、提前期和ABC分析等任务。
• 🚚 仿真还可以用于交通流量建模、机场、港口运营、公共交通、交通安全等领域，许多仿真软件包含GIS地图，方便模拟实际路线。
• 💻 Yashar通过一个实际的供应链仿真案例，展示了如何快速创建供应链的视觉仿真，包括制造点、配送中心、零售商和卡车的移动。
• 🔧 他建议学习Java编程基础和特定仿真软件的操作，这对于从事供应链仿真领域的职业发展至关重要。

Q & A

MIT的供应链管理微硕士课程包含哪些课程？

MIT的供应链管理微硕士课程包含五个课程，其中包括供应链基础（SCx）、供应链动态（SC3x）、供应链分析（SCx Supply Chain Analytics）、供应链设计（SC2x Supply Chain Design）以及供应链管理的混合硕士课程。

Yashar Ahmadov在亚马逊担任什么职位？

Yashar Ahmadov在亚马逊担任高级仿真数据科学家，他使用仿真、数学优化和数据科学工具来解决复杂的供应链问题。

什么是数字孪生技术？

数字孪生技术是一种创建物理实体如工厂、港口、仓库、零售商等的数字表示的方法，这样我们就可以在数字环境中进行实验和分析。

在供应链管理中，模拟的主要作用是什么？

在供应链管理中，模拟主要用于理解、预测系统行为和评估不同的选择方案。它可以帮助进行'what-if'分析，即在不同情况下系统将如何表现。

模拟可以分为哪些类型？

模拟可以分为确定性模拟和随机性模拟。确定性模拟中没有随机变量，输入相同总是得到相同的输出；而随机性模拟中包含随机变量，更贴近现实生活的不确定性。

为什么在供应链管理中使用基于代理的模拟？

基于代理的模拟允许模拟从非常低的抽象级别到非常高的抽象级别，可以模拟个体行为规则和不同代理之间的交互，提供了极大的灵活性和复杂性处理能力。

模拟的优点有哪些？

模拟的优点包括增加现实感、可以模拟现有或不存在的系统、可以研究危险系统而无需风险、瓶颈分析、回答'what-if'问题、结果可复现、易于与管理层沟通等。

在什么情况下不应该使用模拟？

如果可以通过常识分析解决，或者当没有资源去建立和验证模拟模型，或者不能满足项目期望，或者系统行为结构不良时，就不应该使用模拟。

在供应链模拟中，如何定义和使用代理？

在供应链模拟中，代理是代表现实世界中的实体，如制造站点、仓库、卡车等的对象。你可以在代理内部定义它们的行为和参数，这类似于面向对象编程中创建类和对象的概念。

模拟中常见的陷阱有哪些？

常见的陷阱包括试图一开始就模拟整个复杂性，而不是逐步增加复杂度；以及在开发过程中缺乏与利益相关者的沟通，没有让他们了解模拟的工作原理和结果。

如何将模拟与优化结合起来使用？

模拟和优化可以结合使用，例如，先通过优化算法确定供应链网络的最优结构，然后通过模拟来验证这一结构在不同场景下的表现。此外，机器学习模型可以用于为模拟提供决策参数。

Outlines

00:00

🎓 开场介绍与课程概览

MIT微硕士供应链管理课程的直播活动由Miguel Rodriguez Garcia主持，他是MIT运输物流中心的研究员，也是供应链基础课程的负责人。他感谢参与者加入，并介绍了与Paulo Sosa共同主持此次活动。Paulo是S3X课程的负责人。他们介绍了活动的日程，包括客座演讲者的演讲和问答环节，并鼓励参与者使用Zoom的问答功能。此外，他们提醒参与者这是MITx微硕士课程的一部分，并鼓励大家完成课程。

05:01

🤖 供应链模拟的重要性与应用

客座演讲者Yashar Ahmadov介绍了供应链模拟的重要性，他是一名拥有超过八年工作经验的工业工程师，目前在亚马逊担任高级模拟数据科学家。他使用模拟、数学优化和数据科学工具来解决复杂的供应链问题，如网络设计、设施选址、资源规划、库存管理和散列等。Yashar拥有MIT供应链管理硕士学位，是微硕士课程的毕业生。他解释了模拟是如何模仿真实系统的行为，包括确定性和随机性模拟，并强调了数字孪生的概念，即创建物理实体的数字表示以进行实验。

10:02

📊 模拟的类型与工业模拟的范式

Yashar详细讨论了不同类型的模拟，包括确定性与随机性模拟、静态与动态模拟、连续与离散模拟。他介绍了工业模拟中的三种范式：系统动力学、离散事件模拟和基于代理的模拟，并解释了每种范式的特点和应用范围。他还强调了基于代理的模拟的灵活性和优势，可以模拟从微观到宏观不同层次的细节。

15:03

🚀 模拟的优势与局限性

Yashar讨论了模拟的优势，如增加现实感、研究危险系统、瓶颈分析、回答“如果...会怎样”的问题、结果的可复现性以及解释性。他还提到了模拟的易沟通性，特别是通过动画与管理层沟通。此外，他也指出了模拟的局限性，比如在可以进行常识分析的情况下不需要模拟，或者当系统行为无法验证，或者项目期望过高时不应该使用模拟。

20:04

🛠️ 模拟软件工具与应用实例

Yashar介绍了模拟软件AnalogX，它是基于Java编程语言的，并提供了GIS地图、空间标记工具和2D/3D模拟等功能。他还提到了软件中的特定行业库，如物料搬运、行人、道路交通和化工制造等。他通过一个例子展示了如何使用模拟软件来创建一个小型供应链的可视化模拟，并解释了如何使用软件中的拖放功能来构建模型。

25:05

🚚 供应链模拟的实际操作演示

Yashar通过一个实际操作演示，展示了如何快速创建一个供应链模拟。他使用了AnalogX软件，并在短短几分钟内建立了一个包含制造点、分销中心和零售商的模拟。他还解释了如何为不同类型的节点添加动画和行为，以及如何创建卡车代理来模拟货物的运输。这个演示强调了模拟的可视化和沟通优势。

30:05

🔍 模拟与优化的结合

在问答环节中，Yashar讨论了模拟与优化的结合使用。他指出，根据问题的类型，可能需要使用不同的工具，如数据科学、机器学习或数学优化。他解释了在优化复杂供应链后，可以使用模拟来验证其有效性，并测试不同场景下的表现。此外，他还提到了机器学习模型可以集成到模拟中，以帮助做出决策。

35:06

🚧 模拟实践中的常见陷阱

Yashar分享了在模拟实践中应该避免的常见陷阱。他建议初学者不要试图一开始就模拟整个复杂系统，而是应该从一个小的原型开始，并逐步增加复杂性。他还强调了与利益相关者保持沟通的重要性，确保他们理解模拟的过程，并得到他们的批准。

40:08

🎉 活动总结与未来课程预告

Miguel和Paulo对Yashar的演讲表示感谢，并总结了此次活动。他们提醒参与者，这是秋季系列直播的最后一次活动，并预告了SC2X和FORX课程将在圣诞节后开放。他们鼓励大家在假期后继续完成微硕士课程，并感谢所有参与者和观众。

Keywords

💡供应链管理

供应链管理是指对供应链中的物流、信息流和资金流进行整体协调和优化的管理过程。它涉及到从原材料采购到产品制造，再到最终用户交付的整个流程。在视频中，供应链管理是整个讨论的核心主题，特别是在讨论模拟技术如何帮助解决复杂的供应链问题时。

💡模拟技术

模拟技术是一种通过创建现实系统模型来预测系统行为和评估不同决策的技术。在视频中，模拟技术被用来解决供应链问题，如网络设计、设施位置、资源规划、库存管理和运输等。

💡数字化双胞胎

数字化双胞胎是指创建物理实体的数字表示，以便在数字环境中进行实验和分析。视频提到，通过模拟技术可以创建供应链中各种实体的数字化双胞胎，如工厂、港口、仓库等，以进行'what-if'分析。

💡随机性

随机性是指在模拟中引入的不确定性因素，它反映了现实世界中变化无常的特点。视频中强调，供应链中的许多变量如订单量、交货时间等都是随机变化的，模拟技术通过引入随机性来更真实地反映这些变化。

💡库存管理

库存管理是供应链管理中的一个重要组成部分，涉及到需求预测、安全库存、补货点和ABC分析等。视频中提到，模拟技术可以用来测试库存管理策略在面对不确定性时的有效性。

💡系统动力学

系统动力学是一种用于理解复杂系统行为的模拟方法，它通常用于宏观经济政策和系统整体行为的建模。在视频中，系统动力学被用来展示如何通过高级别的抽象来模拟供应链的动态变化。

💡离散事件模拟

离散事件模拟是一种模拟方法，它通过安排离散事件（如货物到达时间）来模拟系统的行为。视频中提到，这种模拟方法在90年代和2000年代非常流行，用于模拟仓库操作和港口卡车等。

💡基于代理的模拟

基于代理的模拟是一种模拟方法，它允许模拟个体行为规则和交互。视频中提到，基于代理的模拟因其灵活性而受到推崇，可以模拟从微观到宏观不同层次的复杂性。

💡优化

优化是指在给定的约束条件下，找到最佳解决方案的过程。视频中提到，优化通常与模拟结合使用，例如，先通过数学优化方法找到供应链的最优配置，然后通过模拟来验证这一配置在面对变化时的表现。

💡人工智能和机器学习

人工智能和机器学习是数据分析和决策支持的工具，可以用于模拟中的参数设置或决策制定。视频中提到，机器学习模型可以与模拟结合，为模拟提供实时的决策支持，如选择哪个供应商。

Highlights

MIT微硕士供应链管理课程的直播活动，由Miguel Rodriguez Garcia和Paulo Sosa共同主持。

介绍了matx微硕士项目，包含五门课程，目前有课程开放报名。

嘉宾演讲者Yashar Ahmadov是亚马逊的高级仿真数据科学家，拥有超过八年的供应链管理工作经验。

Yashar Ahmadov拥有MIT供应链管理硕士学位，并且是微硕士项目的校友。

讨论了模拟在供应链管理中的应用，包括库存管理、运输和系统动力学。

模拟可以用于理解、预测系统行为和评估不同选择。

介绍了不同类型的模拟，包括确定性和随机性模拟。

讨论了工业模拟中的三个范式：系统动力学、离散事件模拟和基于代理的模拟。

模拟的优势包括增加现实感、研究危险系统、瓶颈分析和回答“如果...会怎样”的问题。

模拟可以提高决策的可解释性，与管理层的沟通更为容易。

提供了模拟软件Analogic的介绍，它支持GIS地图和2D/3D模拟。

库存管理中模拟的典型任务包括需求预测、安全库存优化和ABC分析。

展示了如何使用模拟软件快速创建供应链的视觉模拟。

讨论了模拟在运输流量建模、机场和港口运营中的应用。

提供了关于如何结合使用模拟和优化的工具和过程的见解。

强调了在进行模拟时需要避免的常见陷阱，例如不要试图一次性模拟整个复杂系统。

直播活动是MIT运输与物流中心举办的秋季系列的最后一次活动。

鼓励参与者完成微硕士项目，并预告了下一批课程将在圣诞节后开放。

Yashar Ahmadov分享了他在供应链模拟领域的经验和见解，并鼓励参与者提问和交流。

Transcripts

[Music]

hi everyone welcome to another live

event of the MIT microm Masters in

Supply Chain management I'm Miguel

Rodriguez Garcia a researcher at the MIT

Center for transportation Logistics and

I'm the course lead for sex supply chain

fundamentals first I just want to say

thank you to everyone for joining us

today h this is the second and final

life event of the full series a series

of crosscourse Life events for se1

supply chain fundamental

and s3x supply chain Dynamics and that's

why I'm really happy to be co-hosting

this live event with uh my colleague

Paulo Sosa junr course lead of s3x hi

Paulo how are you hey Miguel how are you

thank you for the introduction hi

everyone it's great to be here with you

all we are excited to share some great

insights about supply chain in this live

event today uh in our agenda for today

is the following first our guest speaker

will give us a presentation that will

last around 25 minutes

then we will have some time at the end

uh when he will answer questions from

the audience so we encourage you to

participate and use the Q&A feature in

Zoom U not the chat box but the Q&A

feature and then Miguel and I will take

those questions and channel as many as

we can to our um speaker but before we

introduce our guest speaker we want to

share some um something with you all

right Miguel yeah that's right Paulo so

we just want to remind everyone that

this event is part of the matx

micromasters program in Supply Chain

management a program that we develop

here uh at the center for transportation

and Logistics at MIT and as well as

supply chain fundamentals and supply

chain Dynamics uh the micromasters

program includes three other courses uh

so five courses in total and some of

them are currently open for enrollment

so don't hesitate to check them out

we'll be posting the link in the chat

group in case you guys are interested in

completing the uh the program which

which of course we encourage you to do

so without further Ado uh let's introd

uh introduce our guest speaker Paulo all

right so today we are honored to have

yashar ahmadov as our guest speaker

yashar is an Industrial Engineer with

more than eight years of work experience

in Supply Chain management and he is

currently a senior simulation data

scientist at Amazon he uses simulation

mathematical optimization and data

science tools to Sol complex supply

chain problems this include Network

design facility location resource

planning inventory oranization and

scatterling among others yashar holds a

master's degree in Supply Chain

management from MIT he was part of the

2021 Blended cohort um he is also a

microm masters Alum which means he

passed all courses from the micromasters

program like many of you are doing right

now uh we always like uh to remind the

audience that one among many other

benefits from earning the microm Masters

program credential is that you become

eligible to apply to the Supply Chain

management Blended Masters program at

MIT just like Yar did and to other

universities around the globe so welcome

Yar hello um thank you thank you Paulo

and Miguel I'm happy to be here and I

greet all the um all the people who are

watching this live video event I'm going

to start sharing my screen um so today

we're going to dive deep into

simulations and what they are used for

what are they good and in which

situations should we use simulations so

mainly I will focus on inventory

transportation and System Dynamics

aspects of of simulations so here we go

the overview is what is simulation and

then I I will talk um about applications

in inventory management Transportation

System Dynamics and the most exciting

part will be a live demo and I I will my

target is to show you that within a very

short time frame let's say five minutes

you can create a very visual simulation

of your supply chain and that's going to

be um the last part so first of all what

are simulations we hear this word a lot

in different contexts but it's a

collection of methods and applications

to mimic the behavior of real systems we

it can be any system supply chain is one

of them fulfillment networks

warehouses and in this picture you see

the the truck simulator it is also a

simulator because we are trying to mimic

the behavior of some real systems so a

lot of things go under simulation

however um in this context in supply

chain we're talking about industrial

simulation and why we do so we want to

uh understand predict the systems

behavior and evaluate various

Alternatives so the word digital twin

emerged in the last years which means we

have physical factories we have ports uh

we have vessels trucks warehouses

retailers and so on we want to create

their visual representation and digital

representation so that we can experiment

on top of it and the world is changing

fast the situation is changing fast and

we want some kind of tool that would let

us to do what if anal analysis so there

are many different types of simulations

within industrial simulations also some

point the at the beginning when we did

not have very strong computers the

simulations were mainly deterministic

meaning that there is no random

variables inside that simulation and

whatever input you give you always take

you always get a unique output on the

other hand we have stochastic

simulations which is the widely used in

this domain and here you can have random

variables and the beauty of having

stochastic simulations

is we know that in real life nothing is

deterministic right uh one day we

receive orders maybe for thousand units

the next day 1,100 the next day 900 it

changes all the time between days

Seasons uh weeks the lead times are also

stochastic that's also one of the major

things taught at mitx micromasters in in

Supply Chain management uh we have some

kind of probabilistic distributions most

of the time we can approach it as normal

distribution and we want to optimize our

policies our resources under this

probabilistic environment so the lead

time could be one week eight days 5 days

but it's never a stable number and

sometimes there are static simulations

that have no time Dimension um for

example Monte Carlo simulations this is

also told you open an Excel file

generate some random variables and

experiment on top of it and like we we

mainly Focus the ones that are

stochastic Dynamic that is the system

Behavior changed over time right we have

let's say thinking of ocean

transportation we have vessels with a

number of containers moving the state of

the system changes every minute every

second and the third dimension is

continuous versus discrete so continuous

is System state changes continuous on a

continuous basis and discrete is

basically you have discrete points in

time and then that's it it's um there

are some defined points that you jump

from one state to another an example of

continuous system is an altitude of an

airplane and as long as it's flying

there is a number that's evolving it

goes up when we take off and Landing it

goes down and it can be zero when it's

on ground for some time but that's okay

it's also continuous thing or discrete

events like customers visiting a

supermarket right the the customers if

you try to take notes when they enter to

supermarket it's it's like um some

points in time these is discreet so

mainly the most advanced methods focus

on stochastic dynamic and continuous

types of simulations okay this brings us

to the next slide and in the world of

industrial simulation we have three

paradigms three different

worlds and this started with um System

Dynamics and discrete event simulations

and this type of discrete event

simulations became very popular in the

90s and 2000 years and now we have

evolved a new stage where we have agent

based simulations and I'm big fan of

agent based simulation because it it

lets you model all kinds of complexities

so System Dynamics if you have a look at

this chart um the y axis it goes from

low abstraction to high abstraction

meaning that how detailed you want to

model your system when you're doing a

modeling a System Dynamics that's at

high level macroeconomic policies the

overall behavior of the system you're on

the top right corner it's mostly

continuous system and you have a very

high

abstraction and you have very less

details and discrete event simulations

are located around this region um they

low you modeling average number of

details like you can model your

Warehouse um operations you can model

your uh port and trucks or arriving and

so on and so the discrete event allows

you within let's say a little bit above

micro level until to mid level and this

was dominating for two decades now we

have agent based simulations which let

us model from a very low abstraction to

a very high abstraction it gives you

basically the whole um a bunch of

opportunities you can model active

objects individuals Behavior rules

interactions between the different

agents and models this is mostly um what

I I prefer because of the flexibility

that it gives me discret event

simulations were working like um the

model was scheduling discrete events

like the track arrives that time is

equal to zero it gets loaded at at time

is equal to 1 hour so that's why it was

called discrete events

simulations what are the advantages of

simulation there are plenty increased

realism existing or non-existing systems

can be started let's say you have a

certain supply chain but you want to

transform it you're going to buy from

suppliers that maybe does not exist in

your supply chain right now maybe you

are going to get new customers that are

not existing now you can model both of

them hazardous systems can be studied

without risks bottleneck analysis

usually if you have read the the book

The called the goal where the theory of

constraints um are explained there the

main idea is to find the bottleneck in

your system understand it better make it

lighter you know because bottleneck is

the problematic part where which defines

the throughput of our system and you can

do this in a digital environment what if

questions can be answered like for

scenario analysis you can say what if I

change this what if I add a new

Warehouse what if I get a new retail

customer and so on results are

reproducible as long as you keep

although it it does a stochastic or

probabilistic simulation but as long as

you keep the random seed the same every

time with the same inputs you will get

the same outputs and one of the things

that I love about simulations is there

explainability nowadays we have ai

Solutions ml Solutions mathematical

optimization Solutions and so on right

everybody is now most of the people are

using for example cheat GPT when you ask

like what is Supply Chain management it

generates an answer but why does it use

certain words but not the others nobody

can answer that because it is how it is

trained based on the uh data it has been

trained and there is um a complex neural

network behind and it's really difficult

to explain uh complex models why they

make certain decision but not the other

so this always comes as a question uh

when using other types of solutions but

with simulations you can say hey and

here is the truck here is the customer

and at that point in time this was the

cheapest option that's why I chose this

route for example ease of communication

with the management especially with the

help of animation and this has helped me

a lot when uh talking to customers to

the leadership management like instead

of um like some some

theoretical tables or data you just open

and show what's going on there are also

some cases where you should not use uh

simulation right if you can do with a

common sense analysis you don't need

there are some simple queuing systems in

the literature if you can use it like

for for driving restaurants and so on

there is no need to set up um spend a

lot of time and energy to build the

simulation models when you don't have

resources if you cannot validate or

verify the behavior of this system if

you can't ex you can't meet the

expectations of the project you don't

you should not overpromise or the system

behavior is ill structured so basically

nobody knows uh what to expect from the

system or how it behaves but this is as

a as a side not and within this agent

based simulations there are multiple

providers the one that I use is software

called analogic it's written in Java and

it offers GIS maps space markup tools 2D

3D simulations many industry specific

libraries for process modeling Material

Handling pedestrian Rail and road

traffic libraries fluid library for

chemicals manufacturing for example and

they also have um a section called

System Dynamics where you can take it

and use right so the conveyor system for

example the transportation system this

is not unique to every company some

things are generic and already these

packages are created for you where you

can just drag and drop and and use them

and spend your energy to fine tune um

the model towards the details of your

system what the things that are not um

general or applies to everybody for

inventory management typical tasks

demand forecasting stock Safety stock

optimization order points lead time abc

analysis often when you're managing

inventory you use some kind of again

probabilistic model you can use um for

example economic order quantities you

can use order use computations for order

up to points or the minimum stocks again

you can compute things but will it work

in reality under the uh probabilistic

behavior of the system so this is a good

place to test what is going on and um

what is happening and you can do

scenario analysis so one of the things

that the coid period taught us is that

instead of forecasting the future better

way is to do scenario planning and

prepare accordingly what if World War II

starts tomorrow the the worst case and

what if everything goes perfectly fine

and um the interest rates again go down

and you know the shipping rates are

affordable you can Define certain number

of scenarios and prepare accordingly

there this is where um the simulation

comes in handy and again for example for

for the uh the the warehouse simulations

and I I do have some simple examples on

on how to model this or or visual

simulations so as I said this is a

sample Warehouse simulations here you

have racks you have employees you have

um forklifts you have trucks coming in

they bring goods for you and some of

them uh these blue ones they take and

take it to your customers and like here

you can do a lot of different type of

experimentation and um you can change

for example let's say you want to know

how many forklifts you you need you can

change some figures from 8 to nine and

see what is their utilization factors

and different number of employees so it

is helping you to make decisions on how

many resources you need this is usually

the case when make when we need to make

a decision on the resources uh we don't

want to overshoot and also underestimate

so we we're trying to find the golden

meane all right and one of the examples

that I like the System Dynamics that's

like a very high level uh

formulations um they used this agent

based modeling to predict the coid

infections and this is one example of

the of that and there were many

proposals and there was not enough data

to validate which

systems give you the best projection to

the Future and there were many different

methodologies proposed and the agent

based simulations outperformed others in

terms of like how many P people um are

suspectable they exposed what is the

infection rate and then if they get

infected how many of them get recovered

how how many of them uh lose their lives

and based on this system dynamic

simulations in in the hint sight now we

see that the this type of agent based

simulations yielded uh the most one of

the most accurate projections on what's

going on in Transportation traffic flow

modeling airport Port operations public

transportation PR Traffic Safety many

things are possible by using um um

simulations again most of the simulation

packages come with GIS maps which means

they that already contains the

information about the railways the uh

highways and you don't need to guess the

transit times it already comes in a

package you just tell the origin and

destination and then it is going to tell

you what's going on the system and for

the um for the uh System Dynamics again

supply chain Dynamics policy modeling

environmental systems Health Care

Systems and also the co analysis that I

showed you are some examples of System

Dynamics right now I will I will stop

this and jump onto U on onto this

simulation software just to show you

that in in a few minutes in some minutes

you can create a a simulation that is

very Visual and you it is basically once

you master the basics it's it it will

take you less and less time here is the

question uh I have one manufacturing

site in Albany New York I have two

distribution C centers in Springfield

Massachusetts and hardford Connecticut

and I have two retailers in Boston and

Providence so the aim is to create the

the simulation of this small supply

chain so I picked these as as an example

so and I I will show you here on this

software uh I don't expect you to follow

all this steps I'm going to go fast just

to show that it works and if you later

want to follow I you can watch the

recording or in a slow mode or I can

share some examples of of of this step

by step so uh the first thing here is

I'm going to create a new model I'm

going to call it supply chain

simulator and then I'm going to set the

model time to hours and it creates a

blank model

for me the first thing I'm going to do

is to draw um to drag and drop the gis

map so on on the left side there are

different libraries that you can use and

one of them is space markup and there is

GIS map and this map as I said contains

all the information about um you know

basically Google Maps but it's coming

from uh open street map provider so it's

a different provider but um it it

already contains all the routes and

highways and you don't need to to to

tell what is the exact rout so in our

example we have one manufacturing site

in Albany so I'm going to just double

click and zoom on around around Boston

to to show it

easily and if I search Albany uh it's

popping up here I'm going to convert it

to GIS map and and then REM remove all

other elements so this threed point here

it's going to be our manufacturing

facility and then I will I will also

locate to the others the Springfield

Massachusetts I'm going to type Spring

Field Massachusetts and it's going to

give me multiple options so I'm going to

convert this also which is located here

and then the remove all other

elements and here I will Zoom it a bit

to see

better in this region so here we go and

then the other one is in hardford I'm

going to search for hardford and then

the the others I'm going to remove all

the elements the next step is two

retailers one let's say in Balon

Massachusetts and then I add it here and

remove the others then the last one is

is in

Providence it gives me this option I add

it here and remove all the elements now

I have all the all the noes located here

right this is going to be my

manufacturing facility and these two are

going to be my um distribution centers

and these two places are going to be my

um retailers and once I do this I can

create some kind of collection again on

the on the left hand side you can create

different Collections and I will use one

for

manufacturing site

location right and it's going to

include the um the gis

point and once we once I add it into

this collection it is

uh giving me to the option to iterate

over um over this set and then you can

easily create other collections for um

the let's say um distribution centers

right and then you can also select here

it's going to be other type it's going

to be a

GIS

point and here I will add the

Distribution Center which was in one was

in Hartford and the other one was in and

Springfield and then I will create

another another collection for the

retailer

locations and I will add

here the other two points which is

Boston and I will put plus sign here I

will add Boston and Providence now this

this

map contains most of the information I

need if I run this simulation it's just

going to stay there and no movement or

anything per se but in then now I need

to tell uh what this is going to look

like it's just plotted in the in the the

locations and that's it for now and then

uh now I need to create actual agents

for for different types and in in this

case I will have one man uring uh site

location

manufacturing site it's going to be a

single and let's select an animation for

this let's call it

warehouse and then finish now once we do

this uh here we need to to tell the the

model where it's located it is located

in a note and it's called it's located

in in Albany now if I rerun the

simulation it's going to to pop up in

the right place with the right animation

right you see the factory sign here

which means um everything is fine I need

to to do the thing for the other two I

need to create the respective agents for

the population agent na agent and then

this is going to be the um

distribution

center and then it's going to have also

2D animation I'm going to use this

warehouse and then

finish and then these are also going to

be located in in the node and that node

is defined by The Collection here which

is Distribution Center

location. get

index so this is going to to put in in

the right place initial number of agents

and this is going to be um this many do

size and when I run it now so we we got

these two also located here and

last uh last one is the the retailer

part I need to to do that also I put

here and I'm going to collect population

of Agents these are are going to call uh

retailer and then next it's going to be

a retail store sign and then finish I

will do the same thing here contains

this

this

retailer we have two of them right now

retailer location do size which means it

will take it from there and they are

located in the node and this is going to

be uh

retailer location do

get

Index right now if if I do this it's

going to also plot the last piece I have

only one thing to create the trucks and

then um I'm going to finalize just to

show you how it moves now on the map we

see um all our nodes uh right now I need

to also create um the the the trucks and

for that I

will go to the main pal and then uh

bring agent

here and it's going to be population of

Agents it's going to be used in

flowcharts and this is called truck I

will select um a sign from here which is

this one next and it will have a a

client which is of

type manufacturing centers will send to

distribution centers and then finish

right and and right now we we can create

initial number of Agents let's say uh

100 and

it's it's going to be um um the the

trucks if I if I run this over we see do

we see the truck also located here but

it's huge so we need to to make it make

it smaller I will go to the truck

section and then

reduce the uh the scale I will put Maybe

0.5

0.5 then at the end this is this is how

it's going to to look like the

simulation and if I run this we will see

that all the all the the trucks are

moving in the in the the right direction

so when you create the truck agent

select them here is our manufacturing

facility these are the two distribution

centers and these are the retailers

now with just a few commands I was able

to create this simulation right and I

don't care about the roads and so on the

trucks are already following the actual

routes between the cities and why is

this beautiful because it's easier to

communicate it is Visual and there are

tons of things that you can add this was

the thing that I did in just five or six

minutes but you can you can add tons of

other things on top of this

um different types of kpis

visualizations and um any type of thing

like time stack charts plots bar charts

histograms you can use entire library

for System Dynamics for card Library

there is an entire thing designed for

you here and for example for uh

warehouses conveyor you don't need to

Define it's already here you you drag

and drop this conveyor object and tell

what is the size what is the speed and

so on here I finish my part now it's the

Q&A session that's correct thank you so

much Yar for walking us through so many

examples of applications of simulation

and Supply Chain management and also for

um sharing this live demo which is great

I'm pretty sure the audience appreciate

this as well by the way we have a great

audience today we have many questions

and we will share some of those right

now I want to encourage you if you have

a question please use the Q&A feature uh

and we will Channel it to yashar so let

me start with two question the first one

I can take myself so are there AI is

asking are there any mitx classes that

focus on supply chain simulation and

optimization the answer to that is yes

we do have so you have content on scx

supply chain analytics you have content

on sc2x supply chain design and also

sec3x supply chain Dynamics we cover

optimization and simulation content

there so feel free to enroll in one of

the links that Emma is sharing right now

in the chat and the the question that is

addressed to your Shar so darl Fernandez

is asking what skill sets do you

recommend we concentrate to learn in

order to have a career in supply chain

simulation field and he's also and the

learner is also asking about toos that

we should be well ver to be relevant in

this field yes so the simulation tools

that I use as of now uh Java they are

based on Java programming language you

don't need to be an expert just

understand how it works the

objectoriented programming how you

create classes and basic syntax and

there is an software that I use today is

called any logic but you can also look

at the market if there are other agent

based simulation providers

you can stick to any one of them but the

ones that I prefer is analogic and the

thing is I've tested this in in very

complex environments right today I had

just one manufacturing two uh

distribution center and two retailers

what if I had hundreds of suppliers

thousands of delivery stations and

millions of customers so this this

methodology would work in that case from

my experience but the other types of

approaches don't work because when it's

too complex it takes you like 40 hours

to run the whole simulation which nobody

is willing to wait for So my answer for

this I needed the basic Java and this

specific software called analogic and

you should understand how objectoriented

programming works all right thank you so

much for your answer jar I I I believe

that uh your answer actually um is

related to one of our learner questions

uh Mario la was asking about the agent

step and I think this is kind of related

to the uh to what you just mentioned uh

so maybe if you can explain a little bit

better that step when you relate the

agents to nodes and also for example to

the trucks like to the different

elements in the simulation because some

of our Learners are still wondering like

what that means yeah okay um so very

basic thing you're probably if you're

familiar with programming you know the

between functional programming and

objectoriented programming if you're not

familiar in very basic words in in Java

for example you create objects and OB

objects here you see the Distribution

Center it is an object it has certain

parameters and certain behaviors and

Manufacturing site is another agent and

it has its own behavior in other

simulation paradigms like this discrete

event or functional programming you

don't have this concept of objects you

create a function for example a truck

movement function and you define there

but here at high level you create a

truck agent and inside it you define

what's going to happen with this so

objectoriented programming takes this

idea and applies it to here let's say I

have a manufacturing site right now I

have not model anything inside this but

let's say you have a thousand

manufacturing sites and they have

certain production process going in so

the good part of this is when you double

click inside the manufacturing site you

can Define what is going to happen with

this agent the same with the with the

trucks lores distribution centers let's

say inside the manufacturing you you

have certain um let's say um uh orders

arriving then you put a source block

here right now I generated random demand

so just random numbers but if you have a

certain demand pattern and you have

certain processing times and inside this

manufacturing site agent you can Define

what is going to happen with it again

you can have thousand of them and their

processing times can be different that's

totally fine you can Define this inside

your Manufacturing side agent and then

for example you have some resource pools

you can drag and drop and say hey I have

here Associates and they for example the

the capacity which means the number of

Associates I have in Warehouse it's 100

they have certain schedule um of of um

working you can Define inside these

agents what is happening so some of

these come with a pre-built bu Behavior

like the truck it has origin and

destination it moves in between these

two so um when I create the the the the

truck agent it has this idea that it

needs to move from origin and

destination and put I put their origin

as our manufacturing site and

destination is randomly selected between

our um distribution centers and then

they they start moving in between so

this is the um the strength of the

objectoriented programming where you

define the high level agent and then

inside of the agent you can Define what

they are going to do how are they going

to behave yeah thank you so much Jer I

think that clarifies a lot of our

Learners uh questions so yeah appreciate

it Paulo you want to take the next yes

we have one more here so many Canan is

asking what are the common pitfalls that

we need to avoid while making the

simulation and I know that you already

um told us in what situations we should

not apply the simulation but assuming

that we start a simulation what would be

the common pitfalls to avoid yeah so new

uh practitioners usually when they start

working on a project they think that I

can model the whole complexity from the

first shot and if you have a very

complex supply chain my suggestion is

start simp

like a very build a very small prototype

that you that it works and then you can

add complexity As you move forward and

at each step you need to test whether

the system behaves as it should do and

for example here it's visual if the

trucks are going in the correct

direction it means they are behaving

correctly and sometimes when you do this

like you you can have logic errors we

are none of us are are like perfect we

make mistakes uh here you need to be

able to debug what's going on wrong

wrong but you need to do it

incrementally instead of um doing

everything at once and then getting

maybe hundreds of errors here if I put

something illogical here it's going to

throw an error and when you do this with

a complex system you get a list of let's

say 50 errors and welcome how how are

you going to to debug that right right

this is one thing and then uh try to

communicate with the stakeholders people

want to know how they they don't want

you to treat it this system as blackbox

you need to give them visibility on how

your uh system is working and talk

somebody is will be consuming your

results your model runs and so on stay

in in close touch with them communicate

and get approvals like sign offs that

this is what they're expecting these are

the two main things um that I would

suggest awesome great recommendations

thank you so muchel do we have time for

one more yeah H maybe one or or or two

and let's see I I can do the next one um

and then we we can decide because we

have a lot of questions so thank you so

much to all our Learners and the

audience for bringing every uh all those

uh like super nice questions but we are

not going to have time to answer them

all H so I think one that is really

interesting is um because we've talked a

lot about simulation but we we all know

and you mention it jashar that a lot of

the times simulation uh Al is done

together or in parallel with

optimization or or you simulate and then

you optimize or whatever so H when you

have high uh variants like I don't know

what tools do you use or what's the

process to actually merge uh and put

together simulation plus optimization

yeah um so these are the set of tools

right AI ml is is a set of

tools mathematical optimization and it

has also sub branches like mixed integer

linear programming pure linear

programming nonlinear programming

dynamic programming which they also

offer a lot of tools for you and this

simulation is another type of tool now

you it might be that the if somebody

comes to you with one terabyte of data

and they're asking for insights they're

P purely hinting at a data science based

solution if somebody brings you uh some

fixed uh demand figures and the transit

times and let's say um the supply

capacities and is asking you what is the

cheapest way of fulfilling this demand

this is clearly a mathematical

optimization

problem simulation is not it has its own

domain so if you have limited number

number of options to set to test

simulation is the way to go but if you

have billions of different options

simulation can't really tell you which

one is the best so depending on the ask

you need to find out which is the right

tool to use but they are used in

conjunction with each other let's say

you build a supply chain you optimize

your very complex supply chain and then

you want to validate it with simulation

right usually um complex optim ation

problems they use mainly deterministic

ones like mixed integer linear

programming where they say my demand is

1,000 tons and Supply is 2,000 tons

there is usually no variability because

it it would take like um ages to

optimize um um a stochastic system in

that way so you can take that optimal

Network create different scenarios and

test if it is really um answering your

your needs might be that if the demand

goes up by 10% your supply chain is

going to explode right nowadays we also

want to model the uh weather disruptions

like strikes and lots of things going on

in our supply chain it's not a flat and

and you know many things can happen you

can test different scenarios with

simulation some people use uh machine

learning uh to feed the parameters of

the simulation right inside the

simulation you need to make a decision

for example which supplier to choose

they have a machine learning model they

connect it to each other and every time

the simulation needs to make a decision

it calls that API and it responds like

in this situation this is the best way

to go so yeah um you can use in in

conjunction with each other these

different methodologies well thank you

so much jashar for for the answer uh of

course for for being here uh we are

going to wrap it up now because it's

been uh 50 minutes it's been a super

insightful session but we want to be

really respectful with everybody's time

um again thank you so much to everyone

everybody who decided to join us today

and to learn more about simulation and

how it can solve really complex problems

in supply chain H in particular I think

the the overview of the types of

simulations at the beginning the

discussion of when to use and when not

to use simulation and and also bringing

that real example uh it was great uh I

think uh everybody uh got a really nice

understanding of this uh simulation in

Supply Chain management so um I would

say that H before saying goodbye uh this

was the last life event of the Fall

series that Paul and I co-host so it's

been a pleasure to share this experience

with you guys and second as we mentioned

before several scx courses are still

open H for those completing SC1 X and SC

3x soon is going to be uh important to

not that sc2x and forx are going to uh

open right after the Christmas holidays

so you guys can take a break uh during

that special time recharge your

batteries and then continue your path to

complete the microm Masters in Supply

Chain management it's going to be

January 3rd when the both courses open

if I remember quietly so we encourage

you to to check them out again thank you

so much to everyone uh of course thank

you Paulo for co-hosting this with me

thank you jashar so much for joining if

you want to share any final words the

floor is you guys yeah um thanks a lot

for the invitation it's um my pleasure

to be here thank you both thank you so

much it was awesome thanks to our

audience I just want to remind that this

um session will be uploaded to ctl's

YouTube channel have a great week have a

great time thank you so much

[Music]

everyone