Advanced Analytics · Data Science · Productivity
Data Analytics: sustainable improvements to productivity
Modelling large volumes of production data and transferring these to algorithms that generate productivity gains is one way to achieve automatable improvements that are sustainable over time.
Improvements in technology were traditionally achieved through methodologies focused on human capital. In other words, the aim was for people to actively participate in and manage processes as identifiers and drivers of improvements, based on process data and diverse methodologies for ongoing improvement. This had the advantage that people participated in the process, but it also had certain disadvantages such as the loss of continuity due to staff rotation, loss of trained staff, loss of critical process information, lack of follow-up on the impacts of improvements over time, and underutilisation of process control systems.
Advanced analytics is a different type of effort, which corrects some of these disadvantages.
It builds on the experience of those who operate the processes, whose participation and involvement in the process is equally important. The only difference is that the process does not depend on them.
It looks to use data intensively: this makes it possible to develop improvement hypotheses, identify patterns, and develop algorithms that imply ongoing improvements.
It aims to support these improvements in software development that maintain and automate improvements over time, without necessarily depending on the operator or people involved in the process.
It promotes the active use of control systems and eventually generates the basis for the development and application of additional industrial Internet technologies (such as Interoperability, Smart Factories, Internet of Things, Internet of Services, Virtualisation, Decentralisation).
However, an advanced analytics project that makes it possible for the data acquired to optimise time and resources and even make decisions autonomously requires a very well-planned development cycle and methodology from start to finish.
A normal work methodology involves: Understanding the problem, obtaining data, separating data into training sets / validation / testing, creation of architecture and model development, training the model with training data, evaluating with test data, adjustments, and final implementation.
To achieve this, work is undertaken in cells that include business experts who understand the specific operational process, Scrum masters who ensure the correct application of the practices within the scrum framework and achieve the effectiveness of the equipment, data scientists who analyse and develop algorithms to predict results and recommend actions, software developers, data engineers, systems engineers, and others.