数字镜像：制造业里的模拟与真实
The Gemini makers
Millions of things will soon have digital twins
The Economist(July 13th. 2017)-Print Edition-Business
未来的工厂会满是机器人生产机器人的景象。Amberg，巴伐利亚的一个小镇里，一座工厂虽然还没有达到这样的程度，但是已经很接近了。德国的工程巨擘西门子运行的这座工厂生产工业数控系统，用于一系列自动化系统中关键配件的生产，包括工厂自己的生产线。
Amberg里的这座工厂宽敞、明亮，而且极度洁净。目前，这个工厂里每年生产1500万个配件单元，从1989年成立以来，在没有扩大厂房并且维持1200个员工（实行三班倒班制度）的情况下，产量已经是最初的10倍。生产过程的75%实现自动化，因为西门子认为，有些工作由人来完成是最好的。目前生产的配件的次品率接近于0，其中99.9988%都不需要进行调整；考虑到它们生产1000种以上不同的配件，这真的是一个令人吃惊的成就了。
这样的成就很大程度上归功于“数字镜像”，就是计算机系统中对实体设备建立了一个虚拟版本，像一个孪生工厂一样。“数字镜像”中，实体工厂的每个细节都被模拟，用以设计控制单元并进行测试，模拟如何生产以及操控设备。所有设备都运转完善后，数字模型就会交付给实体工厂用来进行实际生产。
数字模型并非新近发明。配对的概念追根溯源的话要追溯到太空旅行的早期，NASA为了能在飞行器发射后仍然能够监控和控制太空飞行器而建立了模拟模型。在计算机计算能力提高后，模拟模型转向了数字化。
这种强大的系统的出现整合了计算机辅助设计和建造，模拟，过程控制，产品周期管理等多种元素。有些数字模型还增加了人工智能和虚拟现实能力。这些数字模型可以对已经卖出的产品帮助实现远程监控和提供售后服务，西门子数字化工厂部的负责人Jan Mrosik说“这是整个价值链的数字模型。”
西门子不是唯一的使用数字镜像的公司，GE也在使用。这两家公司，连同这个领域里的一家法国公司，Dassault Systèmes，都在出售各自的数字镜像软件。客户来自航空，国防，自动化，消费品，能源，重工业，医药等多个领域。
使用镜像的动机之一是让产品以较低的成本快速投向市场。数字镜像可以在虚拟环境无限重复设计过程，不需要停掉生产线去看生产出的产品是什么样的。数字镜像统一可以建立工人的工作模型，以提高功效。菲亚特克莱斯勒汽车品牌之一的玛莎拉蒂，使用数字镜像，让Ghibli跑车仅用16个月即可在意大利Grugliasco投产，而正常周期则需要30个月。
数字镜像的广泛使用会冲击供应链。比如，买方可能要求卖方提供所售产品的数字镜像，用于在正式投产之前在虚拟条件下的生产模拟。这一点已经成为Amberg工厂的要求，他们需要上游供应商将数字镜像连同正式产品一同提供以辅助安装。
当产品配备了连接互联网数据的传感器，数字镜像就变得更加灵敏。F1赛车装备了大量这样的传感器，参赛车队使用传感器的数据给自己的车创造镜像，以便在大赛中间的一两个星期里更快的设计、测试和生产上百个所需要的更换配件。GE构建风力涡轮与喷气发动机的数字镜像，以便监测工作状态和定期维修。
甚至复杂度甚低的量产产品最终也可能需要数字镜像。数字镜像的建立有助于追踪和验证产品，这样的用途对于食品工业以及制药业愈加重要。挪威的一家利用算法生成产品安全码的公司首席执行官Thomas Körmendi认为，即使没有完全普及数字镜像，几乎任何产品也都要有与生存日期相关的唯一识别标识。这家公司的安全码可以用手机去扫描，通过联网，进而像产品位置和使用情况这样的信息就会交换给数字镜像。比如说，在伦敦的一位顾客要查验一瓶好酒，通过数字镜像他可以知道的葡萄的原产地，如果酒瓶是掉过包的，则会有警示信息。这样的用途则是同每个人息息相关的。
This article appeared in theBusinesssection of the print edition under the headline"The Gemini makers"
THE factory of the future will be a building stuffed full of robots making robots. A factory in Amberg, a small town in Bavaria, is not quite that, but it gets close. The plant is run by Siemens, a German engineering giant, and it makes industrial computer-control systems, which are essential bits of kit used in a variety of automated systems, including the factory’s own production lines.
The Amberg plant is bright, airy and squeaky clean. It produces 15m units a year—a tenfold increase since opening in 1989, and without the building being expanded or any great increase in the 1,200 workers employed in three shifts. (Production is about 75% automated, as Siemens reckons some tasks are still best done by humans.)The defect rate is close to zero, as 99.9988% of units require no adjustment, a remarkable feat considering they come in more than 1,000 different varieties.
Such achievements are largely down to the factory’s “digital twin”. For there is another factory, a virtual version of the physical facility that resides within a computer system. This digital twin is identical in every respect and is used to design the control units, test them, simulate how to make them and program production machines. Once everything is humming along nicely, the digital twin hands over to the physical factory to begin making things for real.
The digital twin is not a new invention. The concept of pairing traces its roots to the early days of space travel, when NASA built models to help monitor and modify spacecraft that, once launched, were beyond their physical reach. As computer power increased, these analogue models turned into digital ones.
The powerful systems that have since emerged bring together several elements—software services in computer-aided design and engineering; simulation; process control; and product life cycle management. Some digital twins are gaining artificial intelligence and virtual-reality capabilities, too. They can also help to monitor remotely and provide after-service for products that have been sold. “It is a digital twin of the entire value chain,” says Jan Mrosik, the chief executive of Siemens’s Digital Factory Division.
Siemens is not alone in equipping its factories with digital twins. Its American rival, GE, is doing the same. Both companies also sell their digital-twin software, along with firms such as Dassault Systèmes, a French specialist in the area. Customers come from industries ranging from aerospace and defence to automotive, consumer products, energy, heavy machinery and pharmaceuticals.
One motivation for twinning is to bring products to market faster and at a lower cost. The digital twin allows endless design iterations to be tried in the virtual world without having to stop the production line to see how they can be made, says Mr Mrosik. The twin can also model people working in a factory to improve their ergonomics. In one example, Maserati, which is part of Fiat Chrysler Automobiles (whose chairman is a director ofThe Economist’s parent company), used a digital twin to put its Ghibli sports saloon into production in Grugliasco, Italy, in just 16 months instead of the typical 30 months.
The spread of digital twins could shake up supply chains. For example, suppliers could be asked to submit a digital twin of their product so that it can be tested in a manufacturer’s virtual factory before an order is placed. It is already a requirement at the Amberg plant for suppliers to deliver a digital twin along with their product to help installation.
Twins will become more responsive still as products are increasingly fitted with sensors that relay data to the internet. Formula 1 cars are full of such sensors; racing teams use these data to create digital twins of their cars so that they can rapidly design, test and manufacture parts needed to make hundreds of changes in the week or two between races. GE creates digital twins of its wind turbines and jet engines to monitor their performance and carry out preventive maintenance. Data transmitted from a jet engine while planes are in the air can provide 15-30 days’ advance notice of potential failures.
Even mass-produced goods that are far less complex are likely to end up having digital siblings. This would help with product tracking and verification, which is increasingly important in food manufacturing and pharmaceutical production. Just about any product could have a unique identifier that links to production data, if not a full digital twin, reckons Thomas Körmendi, the chief executive of Kezzler, a Norwegian company that produces secure product codes using an algorithm.
The firm’s codes can be scanned with a smartphone, which then connects over the internet so that information can be exchanged with a digital twin on things like a product’s location and use. A consumer in London checking the provenance of a bottle of fine wine, for example, could confirm the vintage, or be alerted to the possibility of counterfeiting if the bottle had actually been dispatched to a different country. That’s something everyone can raise a glass to.
