A Simulation of the Proposed Hardware Configuration for the Expansion of a Boxed Beef Processing Plant

Abstract

In order to prove that their design would work, Cirrus Tech, Inc. conducted an independent simulation for Kuecker Equipment Co., Inc. Kuecker is a hardware and material handling systems fabricator and engineering company. The simulation was provided to them for their customer. Companies with experience in the applications of simulation are often brought in as consultants to conduct studies on existing or future plants. In such cases the scope of the project and its deliverables are well defined. Direct interaction between the consultant and the hiring company provides easy access to data required for the simulation and client participation in the project is high.

In this simulation application, Cirrus acted as an independent third party in a proof of concept for an engineering design and hardware layout for a beef processing expansion plant. This simulation was done at the bidding stage of the project. Cirrus played the role of independently verifying that the material handling solution proposed by Kuecker would meet the requirements of their customer. Through a series of hardware layout iterations and the use of AutoMod^®, Cirrus proved that Kuecker design would work.

At an early stage of the simulation project it was realized that the hardware layout would be changing dynamically, as a result of the bidding process negotiations. The plant layouts were submitted to Cirrus, in the form of AutoCAD^® drawings. Somehow, conversion had to be done to turn the plant layout into AutoMod^® entities. Because of the dynamic nature of the layouts, effort was put towards the development of software to convert these AutoCAD^® drawings to AutoMod^® automatically. The result was a conversion utility written in C++. The conversion process was semi-automatic; it extracted information from AutoCAD^® drawings and wrote it to an AutoMod^® .asy file. This data was then merged with the AutoMod^® models manually. This tool proved itself invaluable as it helped Cirrus cope with four major layout changes. In addition, a set of drawing practices were developed to facilitate the conversion process for future projects.

An important contribution to the success of this project was Cirrus' experience with the control of carousels. Cirrus had developed carousel control software in C++ prior to this simulation project. The carousel is essentially a state machine; sending messages to the carousel controls it. It responds with state information by sending back messages. Cirrus added code to their carousel control software to time stamp messages. By analyzing time stamped message logs, detailed timing information was available for carousel operations. Using this timing data Cirrus developed a model for carousels in AutoMod^®. Cirrus did not see any added value in using kinematics to model carousels in this project, hence a black box approach was used. However, even though no graphical animation was presented for carousels other than a change of color for busy or idle, the model is based on accurate real-world timing data. It takes into consideration, pick-up and deposit as well as communication delays.

Cirrus presented the simulation to Kuecker and then to their customer. At this stage the model’s results were basically based on the combined effect of machine capacities and conveyor speeds. The customer’s target throughput requirements were met. However, The customer expressed their concern on a number of issues. The model did not take into consideration the sporadic nature of cartons of product arriving at the storage facility from the fabrication side of the plant. It did not take into account equipment failures. In the case of machine failures, the model did not dynamically reroute product to avoid blocking conveyor paths. There was no consideration for the current inventory at the beginning of a simulation run. There was little user interaction with the model; no user interface was presented for “what if” experiments.

Cirrus realized soon that the customer wanted a lot more than a simulation based on machine capacities and that the level of sophistication of this model had just risen exponentially. Cirrus also realized that it was inevitable that further requests for more realism, through actual inputs from their current operation would be made. To handle this anticipated increase in functionality, Cirrus introduced the use of its application framework to perform pre-processing of these additional inputs. The reasons for this approach were to keep the simulation model simple and to minimize the impact of additional requirements by making them external to the simulation. In addition, this opened the door for easier integration of the Warehouse Management System functions of this application. Cirrus also planned to take this project from Simulation to Emulation should Kuecker win the bid. Cirrus’s application framework was ideal for this task. This framework is what Cirrus had been using to deploy MHCS and WMS applications. Using the wizards developed for its application framework, Cirrus was able to quickly generate a user interface for the simulation. This helped move some of the pre-processing tasks outside the simulation model allowing concurrent development of the model and the inputs.

The customer provided production profile data that Cirrus pre-processed and interfaced with the simulation through data files. This provided great flexibility as it allowed the user to specify the type and production rates for cartons arriving into the storage area. It also added great credibility to the results since the model was now running actual customer supplied data. Similarly error logs from the customer’s existing plant were pre-processed and injected into the model in the same manner. This approach had the benefit of accounting for both real-world error conditions as well as user specified machine errors. One can create any scenario of machine failures and submit it to the simulation. Additional logic was added to the model to provide the dynamic re-routing of product when specific error conditions were detected. The addition of flow control into the model prevented conveyors from getting blocked by too many cartons. The customer also supplied spreadsheets for their inventory. This data was pre-processed and loaded into the model at startup.

These additions greatly enhanced the realism of the model and showed the dramatic impact of errors and fault conditions on the throughput and recovery through the dynamic routing algorithms. However, last but not least, a request to include the customer’s own order profiles was made. Again the Cirrus framework accommodated these changes. Order profiles submitted by the customer were pre-processed and fed into the model.

In the first iteration of the plant layout, Kuecker Equipment Company used a total of 44 carousels and two pallet-sequencing machines. Early simulation results had shown that 44 carousels were excessive and this number was cut down to 38. Significant savings in hardware and installation costs were the direct result of the simulation process. The purpose of the sequencing machines was to allow retrieval of cartons from carousels in any order without any sequencing restriction. The sequencing machine would assume the task of staging cartons in the order required, to build pallets. However because of cost issues these sequencers were removed from the layout on the second iteration. The removal of the sequencers meant that cartons had to be retrieved from the carousels in sequence without mixing them until they arrived at the Palletizers. This moved the complexity that was once handled in hardware, into the model as a sequencing algorithm issue.

A challenging part of this simulation was, correctly sequencing the retrieval of cartons from 38 carousels according to the customer’s order profiles and yet meeting the throughput requirements while dealing with unforgiving machine errors and faults. Essentially, Cirrus was to show that the carousels could be made to play the role of sequencing machines through the use of smart control algorithms, providing for both present and future order profiles. This justified the cost savings associated with the removal of the sequencing machines from the bid and gave the Kuecker proposal a significant edge over other competitors.

When presented to their customer, they wanted to know exactly how Cirrus managed to meet the requirements with the sequencing algorithms. Fortunately for Cirrus, this issue was anticipated. While developing the sequencing algorithm emphasis was placed on making every carton movement traceable. Detailed information was written to log files, allowing one to verify that cartons were indeed arriving at the proper place and were not mixed. Cirrus supplied a document explaining the details of the sequencing algorithms.

Through this simulation application experience, Cirrus has learned that it is sometimes not only the simulation model itself but the whole package of solutions built around it that can make the difference between winning or loosing a bid. The throughput requirements were met and even exceeded when subjected to no machine faults and errors. However, with the addition of faults and errors as per the customer’s error logs, the throughput was in some cases just under that required. On a request from Kuecker, Cirrus was able to quickly add an additional conveyor system to show that the throughput requirements could be guaranteed at additional hardware cost. This was presented as an option to the customer. The error logs supplied were of the worst-case conditions and were not reflective of new equipment.