APPLICATIONS OF DATA-SCIENCE IN MANUFACTURING

Manufacturing is the backbone of every other industry. Manufacturers use heavy machinery, equipment, tools, etc. to manufacture products. To compete in the market, they need to analyze performance, reduce errors in production, adapt to the changes in the market trends, and upgrade the production system using new technologies. 

Nowadays, the word ‘smart manufacturing’ is becoming a buzz word globally. If any manufacturer wants to compete at global level and increase their scale and speed, they have to adopt data-driven manufacturing (Often called Industry 4.0 or fourth industrial revolution).

The global smart manufacturing market will surpass $303.0 billion by 2026, at a 6.4% Compound Annual Growth Rate (CAGR) from $171.5 billion in 2018¹. Data analytics is a key piece of smart factory. Today at least one- fifth of the large manufacturers rely on embedded intelligence built on ML, AI and IIoT.

Future of manufacturing is in a connected factory² where all elements are able to communicate, facilitating enhanced visibility into each process. The IIoT makes this vision possible through sensors and devices interconnected with machine software and applications. Data analytics system is a vital element of connected factory. 

In smart manufacturing, technology enables machines, people and sensors to share information seamlessly and automatically across shop floor. Connected equipment produces vast amounts of data, and with edge connectivity and computing, this data can be analyzed and understood in transformative ways. Streaming information allows improvement in manufacturing processes, as well as full insight into troubleshooting assessments and predictive maintenance. 

So, the manufacturing industry is undergoing a huge transformation supported by today’s digital age that requires greater agility for the customers, business partners, and suppliers. 

Currently, data-driven manufacturing is new. However, as the chart below shows, many industries have fast adopted it and other related technologies. New applications are being discovered every day, and various solutions are invented constantly. Companies must take this as an opportunity. A well executed manufacturing analytics project can generate ROI in around a year; although average time for getting ROI is 5-7 years. So, sooner a company starts, better for it and if data-science consultancy is sought the ROI comes sooner.

The global ‘Manufacturing Analytics’ market was estimated for USD 5.85 Billion in 2019 and is expected to achieve USD 25.8 Billion landmarks by 2025, with a CAGR³ of 22%. Following are granular breakup statistics for manufacturing analytics market : 

ADVANTAGES OF MANUFACTURING ANALYTICS

1. Profit maximization.
2. Risk minimization.
3. Productivity enhancement.
4. Sustainability, reduced emissions, energy-saving.
5. Improved quality control.
6. Decreased labor cost.
7. Continued operability and high-speed processing capacity.
8. Smart and better-informed decision-making.
9. Better Management Of supply-chain risk.

USE-CASES OF DATA-SCIENCE IN MANUFACTURING

1. RAW-MATERIAL DATA ANALYTICS:
This involves  mainly analysis of price and availability of raw material required for a specific manufacturing use-case. In addition, one needs to analyze world markets for possible alternative raw material that is better/cheaper based on research in material science. 

It also involves Geo-political analysis or epidemic analysis in which usual source country of raw material has to be quickly replaced by alternative. Not only raw materials but by-products and waste generated during a manufacturing process also needs to be analyzed for their efficient usage/disposal. 

Material Data Analytics also involves analysis and prediction of  expected emission norms as manufacturers have to be future-ready for upcoming changes. 

Raw material Analytics is highly specific to industry, very complex and usually one data analytics firm specializes in one industry. Still, the diagram below shows the general process.

However, some large consultants like Deloitte cater to hundreds of manufacturers and can do analysis for almost every raw material used in world. As example see this 179 page report⁴ from Deloitte. Raw material analytics is critical for industries like pharmaceuticals and processed food makers.

2. ANALYSIS OF MACHINE DATA, PROCESS DATA AND  PRODUCT DATA

(A) Machine Data Analytics: 

Both machine-to-machine (M2M) and human-to-machine (H2M) interactions generate machine data. Machine data is generated continuously by every processor-based system (E.g.HVAC controllers, smart electrical meters, GPS devices, Sensor Data, Geotag Data and RFID tags). Analysis of this data helps in reducing energy-usage, controlling of temperature, pressure etc. in boilers, designing of efficient logistics as well as doing predictive and preventive maintenance.

(B) Process Data Analytics:

Every business process (particularly when use of computers is involved) generates its trails. E.g. Application Data,  call detail records, click-stream data, website activity logs, system configuration files, alerts, and tickets. 

Analysis of this data helps to make business processes more efficient, faster and allow businesses to scale.

(C) Product Data Analytics:

Feedback obtained from customers by surveys, polls etc. used to be  the main source of product data earlier. However, nowadays, most consumer-oriented systems like mobile devices,automobiles and medical devices with embedded electronic devices continuously analyze every interaction done by user with device and send data continuously. This data is ‘Truthful’ data (unlike survey responses which can be reactive or biased) and it can be collected in real time, in a cost-effective manner and constantly updated automatically. 

Analysis of this data helps in improvement products and services,  improvement of  cyber-security and makes compliance with laws easier.

3.DEMAND FORECASTING AND INVENTORY MANAGEMENT:

Demand-forecasting shares strong relation with inventory management which makes the two fields depend on each other for smooth functioning. Demand forecasting involves massive work for data scientists & domain-specialists as it requires analysis of big data (including the data from the supply chains) for efficient decision making.

Demand forecasting is crucial to the efficient management of production system for a manufacturer. The opportunity to control the inventory simply by analyzing the data reduces the cost incurred in storing items you may never need. Further as the data input can be continually updated, forecasts will always be relevant to the current market situation,current production environment and material availability. Demand forecasting improves the relationship between supplier and manufacturer and both parties can regulate their activities more efficiently.

In several industries, there are seasonal boom-periods requiring an increase in production. Data science can help in forecasting when and how much production increase is needed to meet the demand in the market. Similarly lean periods can be predicted and accordingly raw-material orders and production can be scaled down. The chart below shows basic practices of inventory forecasting.

4. PREDICT SALES AND PROFIT:

Manufacturing companies require significant capital investment and then further investment for technology up-gradation and scaling business. In this case it is important to be able to forecast future sales and profits based on past data. 

Time-series analysis is particularly suitable for such forecasting as it includes components of trend and seasonality both and it also enables isolating ‘noise’ or random factors which can not be controlled. Using such analysis, company can deploy its investments in strategic manner and scale the business optimally.

5. PREDICTIVE  AND PREVENTIVE MAINTENANCE:

There are four types of maintenance in increasing order of sophistication as shown in the figure:

Downtime due to equipment-failure can be very costly. For many industries there are certain costly and critical machines which are made in limited amount by their producer company and even after giving order it takes months for them to arrive. This will be disastrous for the ordering company. Hence, for such equipment, most companies use real-time monitoring and try to predict fault or breakdown in advance. This help to order replacement in time or sometimes timely repairs can save the costly machine. Such preventive maintenance: can be of two types time-based or usage-based.

The collection of data from operators and machines is used to create a set of KPIs for Overall Equipment Effectiveness [OEE]. This enables a root cause analysis of scrap and downtime driven by data. Data-science also enables a proactive and responsive approach to machine maintenance and optimization. Analysis of production data can also help in detecting and reducing outdated technology/machines, improving product quality, designing new production techniques and optimize production cost. 

If your firm wants to adopt predictive maintenance then you should use the following path:

Some important benefits of predictive and preventive maintenance are as below:

6. AUTOMATION OR ROBOTIZATION:

Artificial intelligence, machine learning, and autonomous systems are a critical part of Industry 4.0, and these features can not not exist without critical elements like data-analytics and data interpretations. The big move towards automation requires big investment. Hence the engineers use data science as a guide leading them to effective allocation of resources and significant productivity gains. Data scientists employ predictive tools to determine the best cost-saving opportunities yielding optimum benefits. The insights are then used by the engineers in their mode of operation and allow the manufacturers to make the best decision while investing their money in robotics and automation technology. This is how data science provides a new way of approaching design and optimization in some of the best production facilities operating today. The use of real-world data to understand effect on production caused by new technology, designs and machinery have been revolutionary for the manufacturing industry. The standard RPA workflow model is shown in the figure.

Robots are changing the face of manufacturing. Nowadays, it is a common cause to utilize robots for performing repetitive, high precision or dangerous tasks. The manufacturers are investing more and more money into robotization. The AI-powered robot models help to satisfy the ever-increasing demand. Moreover, industrial robots largely contribute to increasing the quality of a product. Every year, the upgraded- yet more cost effective- models come to the production floor to revolutionize the production lines. An actual production-line photograph is shown in figure.