Tata Steel is the second geographically diversified producer in the steel industry that operates across five continents. The company has been recognized as one of the largest steel manufacturers and suppliers with an annual crude steel capacity of 33 million tonnes per annum. Tata Steel focuses on a wide range of segments like automotive, consumer goods, infrastructure, etc.
ProblemOne of the company’s steel manufacturing units showed potential for increasing the overall unit throughput by optimizing the internal logistics systems. The opportunity was in improving crane and ladle handling by introducing optimized process flows and testing different layout configurations.
The steel melting team saw that the utilization rate of cranes was unequally distributed throughout the production process. As a result, the throughput was impaired: there was an underproduction of heats (batches of steel), the final product of this unit. The possible reason for this was that most of the crane handling was done manually and independently. Therefore, less human decision making was required.
In the steel manufacturing unit in question, the production process starts with tanks (torpedoes) with hot metal arriving at the plant, specifically at the torpedo station. This metal is then transferred to a ladle which, in turn, is placed on a transfer car. Then a torpedo bay crane picks it from there and transfers it to the desulphurization station. After the metal is purified, it is moved by the transfer cars to the charging bay, picked by another crane, and taken to an LD converter, or vessel. Once the metal is poured into the vessel, the crane returns the empty ladle to the torpedo station.
The objective of the project was to find out if changing the existing rules of handling the internal logistics system would increase throughput. The engineers wanted to find a simple rule of thumb to operate cranes and optimize the overall production process by decreasing human dependency in decision making.
One of the reasons the team decided to go for simulation modeling as a solution was the complexity of interdependencies between the elements of that system. In addition, there were many variables with much randomness (e.g. processing time of vessels and desulphurization stations, downtime of cranes and transfer cars, etc.) that only simulation models could handle.
The first stage was to collect the input data. The team conducted a field study, where they collected the data on every piece of equipment in the unit for the past year. Furthermore, they analyzed the pattern of failure, specifically MTBF and MTTR, and process flow.
Using AnyLogic as plant simulation software, they built a model representing the process from hot metal unloading into ladles at torpedo stations to vessel charging. To set up the model’s logic, the team used the following parameters collected from the field study:
- Vessels are available 87.5% of simulation time.
- Metal processing time at a desulphurization station is ten minutes. After that, a ladle with the purified metal leaves the station.
- Vessel charging time is around ten minutes and the value varies according to a probability distribution (maximum time is 15 minutes).
- A ladle with hot metal should be transferred from a desulphurization station to a vessel within five minutes.
The team created three different what-if scenarios with the aim to reduce waiting time for the vessels and modify the crane operation logic:
- Both charging bay cranes are used for vessel charging.
- One crane works between the desulphurization station and vessel areas while the other crane mainly unloads metal from ladles into the vessels. If the second crane is unavailable, the first one takes over its functions.
- Scenario 2 combined with a possibility to hang empty ladles at the height of nine meters above the floor.
For each scenario they designed 90 experiments, with variations in desulphurization station and vessel processing time, to observe how the quantity of heats produced per day, crane utilization rate, and waiting time of vessels would change. Every experiment simulated ten days and gathered statistics.
Statistical analysis of the experiment results allowed the team to determine the optimal scenario from the standpoint of crane utilization, waiting time of vessels, and throughput in heats (1 heat = 1.65 tonnes). It showed that scenario 3 had a significant advantage of a minimum two heats/day which could save the company several millions of dollars per year. In addition, although vessel waiting time decreased and the throughput improved, the charging bay crane utilization was still around 80%, which was unwanted. To avoid unplanned downtime, the engineers would need to introduce a preventative maintenance plan.
Due to the opportunity to experiment with the model in a safe digital environment, using AnyLogic as plant simulation software, all the necessary changes could be implemented without disrupting production.
To further optimize the unit, the team’s goal is to incorporate the scrap charging part of the production process into the simulation. As a result, they would be able to use simulation modeling to improve the company’s steel manufacturing system as a whole. Furthermore, plans exist to incorporate AI into the model to improve policy making.
Watch the video about this case study presented by Tata Steel at the AnyLogic Indian Conference 2019