With advances in technology driven by artificial intelligence (AI) and the creation of data lakes, organizations are coming to recognize their value to industrial production.

Enterprise AI can be embedded in fundamental business models to augment decision-making. It focuses on outcomes rather than the technology itself, enabling an organization to turn data into valuable insights for creating continuous customer value.

The metal industry, one of the oldest in human civilization, has been the backbone of modern industrial growth. Steel is the most popular metal in use today, and iron, the fourth most common element in the Earth’s crust, is its key constituent.

According to the Worldsteel Association, global crude steel production increased from 189 million metric tons in 1950 to 1.8 billion mt in 2018. Rapid growth over the past two decades came from excess capacity produced in China, which contributes nearly 50 percent of the world’s steel production. The mismatch has caused major disruptions to industry, especially in the western world, as Chinese manufacturers began exporting their excess inventory at low prices.

While this imbalance is likely to continue, companies are working to improve efficiency through modernizing their iron and steelmaking technologies. In the process, they have gradually reduced dependency on human labor, in favor of automation.

A modern steel plant employs far less human labor than 25 years ago. During a period when the world’s steel production grew by two and a half times, the industry has shed more than 1.5 million members of the workforce.

The steel supply chain contains some unique elements that are core to the industry:

• A multi-sourced inbound supply chain. Mines generate a continuous flow of raw materials. However, iron ore comes in a variety of forms and quality that often requires additional processing before moving into end-product processing. This can result in various steel grades that aren’t necessarily pegged to specific customer demands.
• Fault-sensitive production. The iron and steel manufacturing process requires an unbroken flow of materials between production stages, including blast furnace, basic oxygen furnace, continuous caster and rolling mill. Shutting down and restarting a given operation during the steelmaking process can be costly. Hence, production and inventory flow need to be balanced to avoid reheating cost, minimize changeovers, and eliminate unwanted accrual of work-in-progress inventories.
• A complex finished-product storage and distribution network. Storage, tracking and distribution are vulnerable to inefficiencies due to the varied grades, weight and size of end products. Moreover, there are limitations to the use of tracking technologies for the steel industry, such as radio frequency identification (RFID) tags that interfere with steel’s physical properties.
• Multiple sales channels. Traditionally, steel companies have relied on a variety of indirect sales channels, such as dealers, agencies and service centers, all targeting the same markets. Steel original equipment manufacturers (OEMs) have limited control in the marketplace, with minimum visibility to end-consumer requirements. Moreover, indirect channels slow the selling process due to multiple hand-shaking and the accumulation of overhead, such as agency commissions. With the advent of internet selling and direct sales channels, e-marketplaces and e-auctions have become a popular means of enhancing transparency, shortening sales cycles and reducing overhead. At the same time, e-marketplace platforms have given customers ready access to market data and competitive quotes for specified grade requirements. This has resulted in a proliferation of steel grades, 75 percent of which have been developed in the last 20 years. Meeting customers’ requirements with the shortest order-fulfillment cycle and most competitive price have become keys to the selling process.
• A commoditized and volatile market. In the steel supply chain, both raw materials and finished product are commoditized. Hence, the business is exposed to price volatility at both demand and supply points, resulting in diminished profitability.

Enterprises are generating large volumes of data daily, and it’s growing exponentially. Data comes in both structured and unstructured forms. As in-memory computing, storage, and digital technologies become reliable and affordable, many metal companies are using them to develop advanced analytics and gain process insights. Up to now, however, most of those efforts have lacked organization-wide vision in the form of integrated supply chain strategies. The steel industry has significant room to benefit from improving its digital prowess.

A digital twin is the virtual replica of physical supply-chain processes, and the backbone for cyber-physical integration. It ensures the seamless transmittal of data between digital worlds and physical entities. To enable enterprise AI, the following attributes of digital twins are necessary:

• An ecosystem commerce platform for information exchange with all internal and external business partners, through commercially available off-the-shelf software.
• Listening platform and information subscription, to capture information beyond the boundaries of direct control.
• Physical equipment connectivity and event capture through Internet of Things (IoT) devices. The digital twin ensures continuous, real-time data collection at various supply-chain nodes, such as ore storage (by miners, suppliers and vessel operators),  production (by coke oven, sinter plant, blast furnace, caster and mill), product storage and distribution (by yards and freight transporters), and sales channels (including e-marketplaces, service centers and dealers).

A big-data lake is the single place of storage for all enterprise data in its native format. It can be used for a variety of purposes, such as data science-driven advanced analytics and machine learning. For steel companies, a big-data lake can store unrelated business data from various supply-chain nodes, including pits, yards, blast furnaces, casters and mills, in raw formats. Big data can be used to obtain insights in the following areas:

• Market intelligence, composed of information on macroeconomics, monetary policies, tariffs, metal exchanges, commodity price fluctuations, competitor information and geopolitical situations.
• Steel mill data, providing details of capacity and operations at various stages, such as iron and steelmaking, yard management, and transportation.
• Business plan data, including production and shipment plans.
• Partner ecosystem data, generated by external stakeholders such as customers, agencies, service centers, miners, freight forwarders and vessel operators. The partner ecosystem should provide access to data in a multi-enterprise business network (such as an ecosystem commerce platform) from external organizations with whom steel companies do business.

Enterprise AI comprises the following functions:

• Sensing of events at various stages in the steel supply chain. Before reaching the consumer’s doorstep, a steel product has to undergo a complete manufacturing lifecycle. As bulk iron ore is converted into discrete steel products, raw materials move through multiple equipment and process steps. Any disruption or change to any part of the supply chain will have a major impact on other phases of production. A digital twin, with associated attributes such as IoT, will immediately recognize the changes and collect the data for further analysis.
• Analyzing events and determining their impact on key performance indicators (KPIs) at different time horizons. Once event data is collected, an advanced analytic platform is triggered to identify possible influences on planned activities. This step creates numerous “what-if” scenarios in a fraction of a second, allowing for the comparison of outcomes from possible changes throughout the supply network. The assessment can determine impacts on various KPIs within the planning horizon.
• Recommending alternative solutions. While data collection and analysis are essential, the real value of enterprise AI lies in its ability to analyze the full extent of impacts, and provide associated recommendations. If the impact is beyond the threshold of certain KPIs, business rules and lessons from prior cognitive experience can help enterprise AI to recommend solutions that deliver desired business outcomes. Such recommendations must consider influences across the supply-chain network, and recommend optimal plans.
• Optimizing outcomes through continuous cognitive learning. A big-data lake provides insights through data science. Enterprise AI, in turn, uses the information to enable the continuous optimization of outcomes. A big-data lake is a mass of unrelated information that would take the lifetime of a human to comprehend. Without a structure, this information cannot be used for business purposes. Data-science techniques can filter unstructured data within specific business dimensions, such as time frame, geography and product, and unearth hidden connections to enable continuous self-learning.

Enterprise AI drives reliability, efficiency and productivity in the steel industry through reduction of manual labor, replacing it with machine-to-machine connectivity and prescriptive analytics. It can sense elements such as market insight, demand volatility, and disruptions in production and supply. The industrial use of AI technologies, along with investments in big-data lakes and digital twins, promises to transform steel companies into more responsive and profitable operations. A pragmatic view of enterprise AI can dramatically increase steel supply-chain efficiencies, leading to reduced inventory carrying costs and shortening time to market in the volatile steel marketplace.

Hiranmay Sarkar is a managing partner with Tata Consultancy Services’ (TCS) Consulting and Services Integration Practice.