Smart factories as lean manufacturing evolution

Smart factories as lean manufacturing evolution

The purpose of lean is to adopt a disposition to remove inefficiencies from any production phase in the company and never produce more products than those required by demand.
This proposition drives the culture and configures the whole production system and thus impacts not only processes but the whole factory design and architecture. The convenience of IoT, cloud platforms and massive spread of artificial intelligence impact the core of all these stages and call for a redesign of manufacturing processes, architectures, and ecosystem, creating a new generation of factories — the smart factories.

Optimizing the value stream

In lean manufacturing, the value stream is composed of all activities required to design, create and provide a given product to the user. In the quest to achieve a process without inefficiencies, we shall constantly think over activities in the process and understand which ones drive value and should be removed or can be optimized. Usual inefficiency sources in manufacturing are excess of inventory, overproduction, non-value-added processing, suboptimal trajectories, production of scrap parts or underutilized people.

Excess of inventory

Demand forecasting is one of the most important and complex issues regarding goods production. Demand depends on multiple factors, some inherent to the product and some contextual (competing products, economic context, and other factors). To provide accurate forecasts, we should act on all of them. Benefits will not only affect a given factory but propagate across the production chain reducing waiting times and the bullwhip effect [1].
Material stock inefficiencies are not only related to items quantity but also their diversity: the more homogenous the stock, the quicker it can be absorbed by demand of other products. This aspect can be addressed at the product design phase where models and heuristics can be used to identify and suggest alternative components for optimal designs [2]. These models may help not only to minimize material variety in stocks but also prioritize components that won’t become quickly obsolete, driving design and stock actualization as well as technology transfer across products. Technology transfer across products may have two additional advantages: product competitiveness since a correlation exists between the speed of technology transfer and how do products fare on market [3], and the ability to obtain competitive prices from providers by analyzing provider components, location, trends and prices for gradual design updates.

Mitigating defect related issues

Defect related issues have multiple dimensions: their recognition, traceability and even forecasting. All of them can be solved with proper information infrastructure and artificial intelligence models to assess current or expected item quality. Defects recognition aims towards fault-free processes through intelligent supervisory control systems (ISCS) where the defect detection may be formulated as a classification problem [4]. Defects traceability is enabled by IoT interconnectivity: as soon as a sequence of defective items are found, they can be traced to the manufacturing machine and take some action. Further production may be rescheduled to other machines, information feedback to the system about current machine state and behavior to associate it to defective items and eventually check if some person with required skills is available to repair it immediately or schedule it for maintenance. In some cases, defects may be predicted. Machine and contextual data can be used to foresee a high probability of defective items and take preventive measures to avoid incurring into them.

Reducing excess of motion

Excess motion can be mitigated using heuristics and AI models for layout optimization [5]. In a smart factory context, excess motion can acquire new meaning: we may no longer speak only about physical distance but also of steps to action. This analysis on how information is issued, presented and how people can act upon it may result in competitive advantage. Insights shall be accompanied with recommendations and eventually, this will evolve to action confirmation so that a user would just confirm the recommended action and this gets executed. Along with each recommendation, a human-understandable description should be issued. This way the person may understand context and concrete data on which the opportunity or issue was detected, some explanation why the model believes one should take the action as well as an estimate of the expected outcome.

Growing talented persons

Another important aspect of optimization is people’s talent and work time, to avoid under-utilized persons as well as to keep them always motivated and help them grow their skill set. To this end, people skills and characteristics can be analyzed to elaborate schedules that maximize overall working, productivity and skills transfer. Any event in the smart factory that requires human intervention may trigger re-schedulings to ensure best schedule configuration according to these goals. The scheduling system would require to understand which skills, seniority and personality traits work best together and project how knowledge is transferred among workers and find best arrangements that foster teamwork.

The use cases described above can be best implemented in a factory that properly integrates information flows and provides good means to store, process and serve information. How to best architect information systems and integrate data is still a matter of active research. Best data and information management practices in Industry 4.0 are evolving, but we can expect will do so quickly and benefit from the extensive experience at other software domains.

  1. Lee, Hau L., Venkata Padmanabhan, and Seungjin Whang. “Information distortion in a supply chain: the bullwhip effect.” Management science 43.4 (1997): 546–558.
  2. Jean Thilmany. “Putting Artificial Intelligence to Work in CAD Design”. Last retrieved 2019–07–23 in
  3. Cusumano, M.A., Nobeoka, K., 1998. Thinking Beyond Lean. The Free Press, New York.
  4. C. A. Escobar and R. Morales-Menendez, “Machine Learning Techniques for Quality Control in High Conformance Manufacturing Environment,” Advances in Mechanical Eng, vol. 10, no. 2, 2018.
  5. M.-Y. Cheng and L.-C. Lien, A hybrid AI-based particle bee algorithm for facility layout optimization, Engineering with Computers, vol.28, pp.57–69, 2012.

Read more articles

Industry 4.0: improving with predictive capacity

Industry 4.0 • Oct 16, 2019

Industry 4.0: improving with predictive capacity

Industry 4.0: a growing opportunity!

Industry 4.0 • Jul 11, 2019

Industry 4.0: a growing opportunity!

Want to read a whole series of QLECTOR articles?

Read a whole series of Industry 4.0 and Continuous Delivery at QLECTOR, reachable on Medium.

Keep in touch

Always be one LEAP ahead!