During the second semester in my second year as an Electrical Control Engineering option student, I saw how the final year students struggled with their research work. So, I started asking questions about the final year research work and its related problems. Upon seeing the enormity of the task in the coming semester not many months away, I started gathering information about my intended research work.
The research was on how we can transform existing hardwire systems into two digital systems using Programmable Logic Controllers (PLCs) etc. and the resulting benefits. In the late ’90s, not many Industrial Automation companies were visibly present in Ghana.
To be able to do my research work, my search led me to Mr. Bonsu who was the country-head of the ABB, and Mr. Peter Franks, the country representative for Rockwell Automation. Both were very happy with my taking a bold step to research into Automation, because according to both representatives many of my peers were not so enthused about researching into new areas.
It was a grey area for many students. However, of these two pioneers in the industry, Peter was the one who was interested in my work. Besides, he was more than willing to let me use the PLC ladder logic programme to convert the hardwired schematic designs into a digital one, and also do test simulations. I later worked with him after graduation. Peter Franks was one of the people who have positively influenced my career path, and also introduced me to a lot of other highly-trained individuals in the field.
Control systems engineering is very interesting, and the roles of Artificial intelligence (AI), Machine learning (ML) and industry 4.0 are making it better every day.
Maintaining growth, the desire of every enterprise
The Sigmoid function below is beautiful for academics, but to the ordinary man it means nothing. The function has a lot of lessons to offer. Graphically, the Sigmoid function is an elongated S-like shape lying on its side. This figure corresponds to three phases of growth.
During the learning phase, which begins at the bottom of the lower part of the ‘S’ shape, it corresponds to the initial Learning phase. This involves hard work, and during this phase little seems to be accomplished.
However, with persistence, we can move to the next stage, which is the Growth phase. During this stage, things seem to be in an upward swing as if there was no struggle in the beginning. Gradually, the growth curve becomes steeper. This leads to the Saturation phase, and then finally into the Decline phase. At this phase, things seem to lose steam. With insightful analytics from ML and AI technologies, meaningful changes can be made to keep the growth going.
Insightful Analytics is the way to growth
Understanding your environments is the way to maximise growth. To get a fair understanding, analytics is the medium of getting this needed information. Gartner classifies analytics into four groups, and these are;
Descriptive Analytics – is the form of analytics that initiates Business intelligence. Descriptive analytics help management to acquire hindsight information about what happened in a particular situation.
Diagnostic Analytics – answer the question of why a particular thing happened. This form of analytics relies on data to understand the causation of events and performances.
Predictive Analytics – focus on statistical methods as a form of analysis, which leads us into forecasting; that is, finding a mutual or reciprocal relationship between two or more things. The outcomes can be used in building predictive models that are based on historical data. Information about future possible outcomes can be predicted using these analytics.
Prescriptive Analytics – As the name suggests, this form of analysis helps us to answer questions like “What should we do; what are the best ways to achieve our desired goals; and how can we manage our available resources?” (Hagerty, 2019)
Machine learning (ML) and Artificial intelligence (AI)
Artificial intelligence (AI) and Machine learning (ML) are the buzz-words in recent times. There are so many articles on ML and AI in the soft and print media. But most of these articles create more confusion than make these concepts clear to some higher and lower management decision-makers who have little knowledge about these trends to make informed decisions.
So, let me try to simplify the terms so the ordinary man can be at the same level devoid of much academic jargon. According to Gartner, all the analytics can be broadly classified into four categories: namely Descriptive, Diagnostic, Predictive and Prescriptive Analysis.
Machine language (ML) is simply a written programme that is used to acquire ‘Knowledge’ about the manufacturing processes or anything a company produces. Artificial intelligence (AI), on the other hand, is a programme that is designed specifically to acquire ‘wisdom’ about your manufacturing processes based on the ML processes.
So, combining this with Industry 4.0, the machine learning systems will acquire accurate knowledge that will be used by the Artificial intelligence (AI) system to give an optimal solution to an industrial manufacturing need (Hack, 2019).
Example and Application of AI and ML in an industrial setting
Take for instance a manufacturing setting, where a conveyor system transports finished products to a warehouse. A piece of machine learning algorithm may be incorporated to acquire certain specific knowledge of the conveyor transportation processes, which will be useful in the manufacturing execution system. With such a system in place, the machine learning programmes over time can acquire knowledge about how a certain quantity of packaged products, at a certain specific weight, are transported in a specific duration.
Also, another machine learning algorithm can be added to monitor performance of the motor’s speed and the frequency of breakdowns. This acquired ‘knowledge’ can be used by the AI system to predict the failure of a piece of manufacturing machinery on the plant floor. The AI system will be instrumental in the planning of maintenance of schedules as wise insightful information is generated.
This acquired information can inform management about Mean Time Between Failures (MTBF), Mean Time to Failure (MTTF) and Mean Time to Repair (MTTR), which are very essential for any manufacturing operation. Planned scheduled maintenance is better than fault-driven maintenance. The cost in terms of monetary value is sometimes too colossal to bear. Take for instance a heater in a boiler system that circulates hot water at certain specified temperatures to assist in the flow of liquified chocolate-mass. When that heater fails in operations without the preparedness of replacement, the operational downtime can be too much to bear in terms of cost. The frozen chocolate-mass can create problems for adjoining systems, too. Also, the cost of paying staff for doing nothing and the penalties for delayed orders will result in huge unprepared spending.
How ML and AI is used in the Industry 4.0 setting
In the field of ML, some argue that the more data set we gather, the better the predictive model. However, some industry experts think better algorithms can lead to better outcomes. I think proper cleaning of the data set to remove all the noise [corrupt data set] will ensure high-scoring predictive models.
The Cloud computing platforms, whether public, private or hybrid, allow modern industry to store a huge amount of data. So, by leveraging on the industrial Internet of Things (IoT) devices on the manufacturing floor, specific data can be sent directly to the Cloud storage which can be used for future analytic work.
- From this storage, the data can be prepared – which is removing all noise from the data. The next very important step is creating a good model.
- The next phase is training the model, which includes supplying the model with the cleaned data that then undergoes evaluation processes, and fine-tuning the selected parameters to increase accuracy.
- Deploying the model is the next logical step. During this stage, the trained model is integrated into an existing production environment to commence its usage to make a practical business decision based on data.
- The next process is often ignored but very critical in ensuring a continuous, accurate, predictive modelling process that is monitoring and refining the process.
In the end, based on the acquired ‘Knowledge’ model, optimal measures can be determined by the AI system which relies on the acquired knowledge from the ML model. This acquired ‘wisdom’ from the AI system can be applied directly to a piece of machinery, or the outcome can be used by management to make informed decisions.
In this simplified explanation of AI and ML in manufacturing, I have tried to eliminate a lot of terminologies like regression, clustering, supervised learning and unsupervised learning, classification, convolutional neural network (CNN), deep learning, recurrent neural network (RNN), etc. It is an attempt to demystify AI and ML technologies without using academic jargon, and to make it simple for the average person.
I believe that the use of these technologies in manufacturing will go a long way to benefit manufacturing settings in Ghana, in the areas of:
- Accurate prediction of the useful life of a hardware or a piece of machinery
- Helping to ensure a good supply chain management system
- Effective cost-reduction through predictive maintenance
- Enhanced Cobotic integration in the manufacturing process
- Consumer-focused manufacturing
- Ensuring high-quality control standards
- Cutting waste in manufacturing processes
To achieve all the enumerated benefits above required a lot of work under the old paradigm. This involved a lot of trial and error, as well as continuous capital injections to achieve these set targets. However, using ML and AI we can achieve the above benefits using less effort and less budget. In an article written by Haponik entitled ‘Reduce Operating Costs and Improve Efficiency Using AI’, the author outlined many benefits of the AI and ML technologies (2019)
Hack, M. (2019, July 20). Industry 4.0: Definition, Design Principles, Challenges, and the Future of Employment. Retrieved from https://www.cleverism.com/: https://www.cleverism.com/industry-4-0/
Hagerty, J. (2019, July 22). Gartner Technical Professional Advice. Retrieved from https://www.gartner.com: https://www.gartner.com/binaries/content/assets/events/keywords/catalyst/catus8/2017_planning_guide_for_data_analytics.pdf
Haponik , A. (2019, August 5). Retrieved from Adepto: https://addepto.com/reduce-operating-costs-and-improve-efficiency-using-ai/