TOP100 Case Studies "predictive maintenance"

Each year, the field of science and technology which technologies and products are becoming indelible "mark" and "symbol"? Domestic and foreign scientific and technological circles and what people and organizations like the most points? What's most worth learning and innovation cases Replay?
TOP100 Case Studies "predictive maintenance"

Sponsored by Maisi Bo (msup) Co., Ltd., to "develop strategic evolution of the times of artificial intelligence," the main direction of its sixth global software case studies summit, the summit from 100 years of outstanding software development practices worldwide case of 2017 the overall development of the industry conducted a recovery disk. Yibai case in this year's list, from Silicon Valley, Seattle nearly 20 cases of research and development, technology leader at home and abroad together to bring a four-day feast for the technical field of more than 1700 participants.

Neusoft Group's product development platform product first division RealSight IoT Things intelligent analysis platform product manager Jiang Zehao, networking, big data analytics case by two industrial customers was about the networking of intelligent analysis of physical development and application of predictive maintenance, the implementation process problems encountered, solutions, results achieved and bring value to the customer.
TOP100 Case Studies "predictive maintenance"
About Things

The value of the data mining analysis can help companies achieve competitive differentiation, drive new business model of things to achieve. Jiangze Hao believes that analysis of data and applications is the ultimate expression of the value of things, while giving users the value of the entire networking industry chain is rapidly collecting statistical data from the shift in the direction of data analysis and intelligence applications.

McKinsey said in a recent release of "ARTIFICIAL INTELLIGENCE THE NEXT DIGITAL FRONTIER", the asset-intensive companies, to ensure the normal operation of complex systems reduce downtime current artificial intelligence is a major opportunity. A European power plant is able to determine the health of a transformer, and thus change its maintenance mode to reduce cash costs by 30% within five years of 20 kinds of variables remotely.

关于设备的维护,江泽浩总结有三个阶段,分别是事后维护、预防性维护和预测性维护。麦肯锡的报告指出,人工智能“在未来”将促成从预防性维护到预测性维护这一转变。江泽浩说,预测性维护是最经济的维护。
TOP100 Case Studies "predictive maintenance"
案例说明:通过以下两个案例,江泽浩对当前物联网的数据应用情况做出来分析并总结启示。
案例一:对消费家电智慧云平台建设,为其提供故障隐患识别和异常检测智能应用。
案例二:为某风力发电企业,提供数据清洗、发电损失评估、风机健康评估等智能应用。

启示总结1 “有用”的数据为什么没有发挥作用?

关于数据源的选择,有些传统观点认为,针对特定问题只有小部分数据真正有用,同时考虑成本因素,仅保留认为“有用”的数据。但如果这样做,在数据分析前就已经将分析对象束缚在传统对于机理认知的范围内,即便采用多么智能和高级的算法,其结果仍然很难超出传统基于阈值、机理模型、统计算法的效果。

只有突破传统的对设备机理的认知,从数据本身出发来分析其中的关联关系,数据维度的选择通过数据分析来决定,才能最大限度洞察数据中的价值。

启示总结2 为什么使用了高级算法,效果并不显著?

受互联网领域的影响,部分企业或工业者迷信于深度学习、强化学习等算法。深度学习作为机器学习的一个分支,虽然在互联网领域有非常成功的实践,但对于传统领域的物联网数据分析,其效果并不一定比回归、分类等传统机器学习算法好,但也并非排斥深度学习。

物联网数据分析应以解决问题和产生价值为根本目的,针对不同的场景选择合适的算法,而非为了使用更“高级”的算法而去做数据分析。
TOP100 Case Studies "predictive maintenance"
启示总结3 预测性维护是通用的么?

对于预测性维护的应用场景,由于设备的不同、应用场景的不同、操作行为的不同,预测分析模型很难做到行业通用,大多数需要根据用户数据和应用场景重新构建或调优。

在物联网智能分析发展的初期,建议企业从已经能够采集到的数据出发构建智能应用示范,即先在较小的成本基础上构建1-2个有效的范例,再逐步扩大应用范围。
 
启示总结4 预测性维护是不是徒有虚名?

目前预测性维护的应用并不广泛,在先期阶段概念炒作的背后,行业里出现两个极端:要么打着预测性维护的牌子,做的仍然传统的事后报警和交互式诊断辅助;要么把预测和分析的概念定义得过高,过度估计和宣传了预测性维护的价值,但实际产生的效果一般。

事后维修、预防性维护、预测性维护,是三种不同的设备运维方式,虽然预测性维护代表更先进的设备运维方式,但对于不同的设备及不同的应用场景,其它的运维方式仍会长期存在,如对于易损易耗件,事后维修(或替换)始终是最好的选择,对于特种设备(如电梯),定期定量的预防性维护在相当长的一段时间内仍会存在。

启示总结5 如何开展预测性维护?

企业对设备运维的成熟度也可以分为不同的阶段,包括1)无数据采集的阶段、2)有数据采集和监控的阶段、3)有数据分析的阶段,处于不同阶段的企业应有规划的构建运维策略,如对于1)阶段的企业,应先采集数据,并实施如基于阈值的报警、基于统计的分析、构建专家库等传统的运维方式,在达到2)3)阶段后再实施预测性维护策略。

精彩总结

根据Gartner2017技术成熟度曲线,物联网、机器学习、深度学习等技术处于“Peak of Inflated Expectations”阶段,迈向稳定发展的过程中必然经泡沫期。企业应正确评估自身所处的阶段,策划符合自身业务的设备预测性维护策略,既要利用更先进的技术和物联网发展机会构建差异化竞争力,又要充分借鉴传统方式和其它行业经验,在风口中成长,在泡沫中存活。

RealSight IoT 物联网智能分析平台

It is based on the monitoring, predictive analysis and optimization of large-things device data platform. The use of Big Data technologies gain insight from massive multi-source heterogeneous data equipment, the environment, business systems in real-time integrated monitoring equipment, predictive analytics and optimization to improve, improve operational efficiency, reduce operational risk and cost savings. More Things content Notes

Reproduced in: https: //blog.51cto.com/14337720/2407593

Guess you like

Origin blog.csdn.net/weixin_34095889/article/details/93036611