Introduction and Company background
Owens-Illinois is one of the largest glass manufacturers in the world. Owens-Illinois (O-I) with 62% market share is the largest glass bottle maker in the world (O-I, 2015). The company has a century-long history of glass making. O-I has introduced the automation in the glass bottle making process in early 20th century. Presently, O-I has presence in four continents (APAC, EU, LA and NA) and has over 80 manufacturing plants all over the globe (O-I, 2015). O-I supplies glass bottles to many major beer companies and wine companies. O-I has posted a revenue of $6.9 billion in 2014-2015 with a profit of $320 million (O-I, 2015). Owens-Illinois is headquartered in Perrysburg, Ohio in the USA. Regional headquarters are located in Sao Paolo, Geneva, and Melbourne. Regions operate relatively independently. The growth of the company has been non-existent for the last 10-15 years. In fact, after the 2008 recession and a bad investment decision in China, the revenue of the company has come down. Coupled with that, increasing pressure on raw material cost has increased the operational cost and reduced profit. Apart from increased focus on revenue diversification by introducing new products, the company is also trying to renovate the manufacturing process. Currently, Latin America and North America are driving the topline and bottom line of the company whereas Europe is pulling down the profitability mainly due to a reduction in revenue and less capacity utilization of the furnaces (O-I, 2015). However, O-I continues to be the largest player in the market. This essay will primarily look at the operational issues faced by Owens-Illinois in its production plants and based on operational theories, suggest actions that can improve the situation.
Operational Problems
There are many problems faced by the manufacturing plants in O-I. All the problems faced by plants can be categorized into three major heads.
Demand Variation
Figure 1: Weekly Demand Fluctuation for a Product at O-I
Most of the manufacturing plants receive demand data from customer service representatives who are either located in regional headquarter or regional sales offices. Every week, sales and planning team sends a demand report to the plants about the demand for different SKUs for the next three months. Based on this demand report and the inventory situation at the plant, plants decide the manufacturing schedule to minimize the overall setup changes for the furnace. However, often the demand that is sent to the manufacturing plant in the subsequent weeks vary considerably from the original plan. This weekly demand variation often creates major changes in the production planning process and often cause disruption. Although, some plants complain that the overall demand for different SKUs for every quarter only varies by less than 10% whereas the weekly variation is often as high as 70% which is a major concern for production. Demand forecasting team creates a forecast based on requirements received from the beer or wine manufacturers and they say that the variation can be almost fully attributed to the demand received from the customers.
Capacity Utilization
Figure 2: Production Process of Glass Bottle
O-I plants have huge variation when it comes to capacity utilization of furnace. A typical production process has three steps 1) Furnace to melt glass 2) Hot End to mold molten glass into bottles 3) Cold end to quality check and pack glasses into final boxes. Furnace runs for 24X7 and needs to produce molten glass all the time except when it is totally shut down. Furnaces are shut down only when there is a huge maintenance required (change of refractory etc.) or demand has gone down substantially. However, these two factors do not lead to the low capacity utilization. Whenever there is a change in color of the glass required, furnace requires 3-7 days (depending on the color) to slowly change from one color of molten glass to another. This 3-7 day is the lost production. Therefore, frequent color changes lead to lower capacity utilization.
Loading and Unloading
Most of the plants have two loading and unloading docks connected to the plant warehouse. For most of the plants, the total capacity of the two loading and unloading plants are more than sufficient to handle all the unloading and dispatching. Plants operate two shifts in the loading and unloading bay with no loading and unloading done on Sunday. However, when the trucks will arrive at the plants is unknown. Sometimes many trucks arrive towards the end of second shift and the second shift people cannot handle all the trucks in time and some trucks need to wait overnight in the loading bay. This delays the delivery as well as cost more money. O-I is thinking to introduce a third loading and unloading bay in some plants but wants to understand for which plant it will be beneficial.
Analysis and Recommended Solution
Demand Time Aggregation
Figure 3: Weekly and Monthly Demand of a Product
As can be seen from the above graph that the monthly variation in demand is very low, whereas the demand variation at the weekly level is substantial. Therefore, it makes more sense for the company to create a production plan based on the aggregated demand of a month and not based on weekly demand (Hanna and Newman, 2012). This will definitely reduce the number of changes in the setup in the furnace which will increase the capacity utilization of the furnace and will reduce the cost associated with a change of color in a furnace and lost production.
However, a monthly production may affect the customer serviceability adversely. Some deliveries to the customer may get delayed as some products will be produced in longer batches and less frequently (Slack, Chambers, and Johnston, 2012). This problem can be somewhat addressed by increasing the level of safety stock for those products so that customers continue to get serviced like before. The only thing that will go up is the inventory holding cost in this case (Greasley, 2008).
Decide on number of dispatch bays required using Queue Theory
The main problem with loading and unloading process in the plants is the uncertainty involved in the arrival of the trucks. Although there is an average arrival rate per day that is more or less stable but at what time of the day those trucks will arrive is unknown and this is the primary reason for some poor utilization of bays. For example, in some plants, most of the trucks arrive early in the morning and had to wait long in the queue. On the other hand, in other plants trucks arrive mostly during the evening and may need to wait overnight. The main problem for the management is to decide if adding one more bay in some plants will solve the issue or more than one bay is required. However, they cannot decide on the same because of the uncertainty involved. In fact, top management argued that even adding multiple bays will not solve the issue and the capacity utilization will plummet.
The most used way to analyze such problem is to use the queuing theory. This situation may not fit the classical queuing problem but one can make some simple assumption to make this problem a queuing problem (Krajewski, Ritzman, and Malhotra, 2007). The process diagram is as below
Figure 4: Loading and Unloading Process at the glass plants
Let’s say for a plant truck arrives at a rate of 24 per shift. There is a probability of 50% that the time between two arrivals is .75 hours. There is a 30% probability that time between two arrivals is 1.25 hours and there is a 20% probability that the time between two arrivals is 1.75 hours. Service time for unloading is fixed (30 mins) and all trucks need to go through the unloading process. Loading time varies depending on the situation. There is a 30% probability that no loading is required and the truck leaves after unloading. There is a 50% probability that the loading time will be 2 hours. There is a 20% probability that the loading time will be 3 hours.
Figure 5: Average Waiting Time for 1 loading bay and arrival rate of 1 truck per hour
For a typical scenario like above, the average time spend in the system for each new truck arriving will increase as a single bay is unable to handle the given arrival rate. Now, in the actual scenario, there are 2 bays. Therefore, the average loading time will decrease (Kumar and Suresh, 2009). Assuming that average loading time is halved and looking at a 16 hour operation day we get that
Figure 6: Average Waiting Time for 2 loading bays and arrival rate of 1 truck per hour
With two bay, the maximum wait time for a truck comes out to be two hours in the system which is acceptable as per the current O-I standards. Therefore, plants that have arrival rate of 1 truck per hour or less, having two service bays for loading and unloading is sufficient. However, some of the bigger plants like Los- Angeles have arrival rate 2 trucks per hour. In such case, the two bay scenario provides the above wait time as below
Figure 8: Average Waiting Time for 2 loading bays and arrival rate of 2 truck per hour
The average time clearly goes up as the shift progresses. This means that the service rate is slower than the arrival rate. If we increase the number of bays from 2 to 3 then the average waiting time in the system becomes as below
Figure 9: Average Waiting Time for 3 loading bays and arrival rate of 2 truck per hour
Still, the average waiting time goes up as more and more trucks are served. This is an indication that current number of bays are not sufficient. If we increase the bay to 4, then the average waiting time becomes as below
Figure 10: Average Waiting Time for 4 loading bays and arrival rate of 2 truck per hour
The above figure shows that when there are four bays running in parallel in Los Angeles plant then the average time varies based on the arrival rate. Maximum waiting time (based on random arrival rate) is found to be 1 hour which is acceptable. Therefore, for plants where the average truck arrival rate is 2 per hour, the number of bays required for acceptable service level and low waiting time is 4 bays.
Increase capacity Utilization through Batch Optimization
As discussed in the previous section that the overall capacity utilization can be improved by looking at the aggregated demand of the SKUs at the monthly level rather than at the weekly level. However, demand aggregation at a monthly level will not solve the capacity utilization problem of the furnace. The main problem with the production line is the long downtime required to change a color in the furnace. Therefore, it is important to minimize the number of color changes in the furnace. For example, if bottle-A is of color amber, bottle-B and bottle-C are of color flint, bottle-D is of color blue and bottle-E is of color amber, then it is not only sufficient to aggregate demand at the bottle level. It is important to understand the demand for each color as that will be essential in reducing the color changes at the furnace. For example, if the furnace produces the monthly demand of bottle-A, followed by demand for bottle-B, bottle-D, bottle-E and bottle-C then there will be 4 color changes. However, if the sequence is bottle-A followed by bottle-E, bottle-B, bottle-C and bottle-D then total color changes required is 2.
Figure 11: Color Setup Optimization on Furnace
As can be seen in the above diagram that in the existing production planning the color changes are too frequent (the above scenario is illustrative and not actual). The first scenario tried to match the demand dates as much as possible. Looking at the demand, it is recommended that a color block level optimization schedule should be run to find out the best possible solution which minimizes color setup changes and at the same time minimizes delay (Brennan, 2011). In this example shown below, you can see that for amber color all the demands are produced on the demand date or well ahead. Therefore, the plant will carry an inventory of this color. On the other hand, blue color glass bottles are all produced and delivered after the original demand date. Therefore, customers of the blue bottle will get a delayed delivery. For flint (ash color) bottles, some of the demands are met on time whereas some of the production is done after the demand date.
Conclusion
Owens-Illinois is one of the largest company in the world in the glass manufacturing industry. The company has dominated the market for over a century. However, in last few years, the growth has stagnated and also the profitability has gone down for several operational issues. Frequent changes in weekly demand are causing the production planner to change the color setup often which results in lost production and low capacity utilization. Loading and unloading is another problem which is causing on time delivery for many orders mainly due to uncertainty around truck arrival. Demand variability can be improved if the production planners start looking at the aggregated demand and plan accordingly. Capacity utilization and color setup changes can be improved by doing an optimization between cost associated with a number of color setup changes required and delay cost. Finally, Owens-Illinois should do queue modeling for its loading and unloading docks at each plant separately to estimate the correct number of bays required to keep the waiting time within acceptable levels.
Bibliography
Slack, N., Chambers, S. and Johnston, R. (2012). Operations and process management. 3rd Edition. Harlow (GB): Prentice Hall, Pearson education.
Greasley, A. (2008). Operations management. Los Angeles: SAGE Publications.
Stevenson, W. (2005). Operations management. Boston: McGraw-Hill.
Krajewski, L., Ritzman, L. and Malhotra, M. (2007). Operations management. Upper Saddle River, N.J.: Pearson Prentice Hall.
Kumar, S. and Suresh, N. (2009). Operations management. New Delhi: New Age International.
Brennan, L. (2011). Operations management. New York: McGraw-Hill.
Jacobs, F., Chase, R. and Jacobs, F. (2010). Operations and supply chain management. New York: McGraw-Hill Irwin.
Hanna, M. and Newman, W. (2007). Integrated operations management. Ohio, OH: Thomson/South-Western.
Fast, L. (2012). The 12 principles of manufacturing excellence. Boca Raton, Fla.: CRC Press.
Bozarth, C. and Handfield, R. (2006). Introduction to operations and supply chain management. Upper Saddle River, NJ: Pearson Prentice Hall.
Yang, Y. and Ma, Y. (2011). Innovation manufacturing and engineering management. Durnten-Zurich, Switzerland: Trans Tech.
Moore, R. (2004). Making common sense common practice. Amsterdam: Elsevier/Butterworth-Heinemann.
Taghizadegan, S. (2006). Essentials of lean six sigma. Amsterdam: Elsevier.
Ortiz, C. (2009). Kaizen and kaizen event implementation. Upper Saddle River, NJ: Prentice Hall.
Berk, J. (2010). Cost reduction and optimization for manufacturing and industrial companies. Hoboken, N.J.: John Wiley.
Owens-Illinois (O-I). (2015). Annual Statement, 2014-2015.
Appendix: Calculation Excel
