Twenty additional supplier representatives demonstrated software systems and equipment at the concurrent Automation Expo, followed by a demonstration of the pilot MES/performance-analysis system jointly developed by Kraft Foods and Raytheon CSI for the Kraft Maxwell House Coffee plant at Jacksonville, Fla.
Conference attendees also toured the contract-manufacturing plant operated by Thermo-Pac Inc. (a unit of H.J. Heinz Co.) at Stone Mountain, Ga., and the fluid-milk products plant operated by Mayfield Dairy Farms (a unit of Dean Foods) at Braselton, Ga. Conference highlights follow.
Kraft is close to totally integrating its 60 North American manufacturing and distribution locations into a single enterprise with common systems and work practices, said Sherriff. Next step: Integrating more than 70 contract manufacturers as well as key packaging suppliers, customers and distribution channels into an "extended enterprise." The supply chain is no longer a chain but a network of customers and partners. "The proposition I lay-out to you is this: Each one of us must e-connect, so we can leverage the capabilities of business-to-business (B2B) communications in the food industry," said Sherriff. "We must electronically connect our business practices."
Improved productivity is a continuing priority at Kraft, Sherriff continued. Kraft's target is to reduce costs by 4 percent per year, which means reducing conversion costs by $1.2 billion over the next three years.
The key to driving down costs: Performance-monitoring applications that integrate equipment with information repositories, thus empowering plant operators with the performance-improvement data and knowledge they need to drive continuous improvement. These applications feed a repository that stores the data in a standard format. "In our plants, we're working toward people having real-time knowledge of what they do, and the tools to do the diagnostics for improvement." The operator is connected to both the machine for line performance, and to the enterprise for monitoring compliance with the production schedule, manufacturing costs and delivery standards.
"This will not be an easy journey, but we're going to bring it all together," Sherriff declared. "We know what we have to do to complete the missing links."
He outlined Kraft's integration approach: At the machine or plant-floor level, Rockwell SQL (structured query language) captures real-time performance and automation knowledge from PLCs. At the operator/plant level, performance applications (measuring yield, downtime, weight control, SPC, etc.) capture raw transaction data from the PLCs and feed it to the MES repository, which is linked to the operator, to the enterprise and to external partners. (Kraft has to date catalogued 30 applications and posts 15 as "Big Board" real-time business-performance measures at headquarters.) Enterprise links connect the MES repository with external business groups for functions such as forecasting, customer orders, production planning, production scheduling and continuous replenishment. Kraft will rely on Web tools to build the first level of connectivity to business partners, said Sherriff, and XML-type languages for advanced applications.
Sherriff's vision: "Business-to-business integration for the first time, largely due to connecting machine and plant to the enterprise." Needed, he added: communications standards to connect with business partners.
Successful automation includes motivating, training and empowering people with software support, Hussain continued. ConAgra organized Statistical Process Control (SPC) and Quality Improvement (QI) teams to develop new SPC systems. These teams "control the process by eliminating the causes of defects." To start an SPC/QI automation project, Hussain recommended, select a small, self-contained and relatively trouble-free production module, which nevertheless needs internal improvement. Then select an automation partner offering strong consulting capabilities, technical service and training.
ConAgra empowers its QI teams with NW Analyst SPC software from Northwest Analytical Inc. Using SPC, QI teams measure variation, then eliminate its root causes. During one three-month period last year, for example, operators reduced tare-weight variation on a sliced-meat product to only 63 defec tive packages from more than 3,000 the previous quarter, reducing product "giveaway" for projected annual savings of $10,000.
Integrated PMIS modules are phased in. One production team takes a module to prototype; configures and tests it; trains the operators; and runs the module in parallel with the existing system to prove it out. Then the module is "rolled-out" to the other two systems. "The goal of project management is maturity, where technology and management processes are efficiently integrated," Williams observed.
Hatfield presented simple system architecture being implemented to control the bottling and secondary-packaging lines operated by self-directed work teams at Riverside. An Intellution FIX SCADA system running on Windows NT integrates Allen-Bradley and Siemens PLCs, Filtec bottle inspectors and Filtec case inspectors. In support of this system, real-time efficiency models such as TRAK-SYS can diagnose the cause of a problem right away.
At Simplot, production models in the MES software layer tell a production-capability model what resources are needed to make a particular SKU. The production-capability model combines that information with inventory, sales forecast and plant-capacity information and sends the results to a master production schedule, which in turn determines the weekly line-load schedule for the plant. The models are continually updated to reflect variables such as seasonality and yields.
Plant architecture integrates Wonderware Factory Suite at the PLC level, Wonderware Protean at the MES level for line-loading and Chesapeake (Aspen Technology) MIMI to generate the production-capability model and master production schedule (MPS). SAS Focus Forecast, a forecast editor, adds historical sales data to develop the most probable future scenario, which is modified monthly by sales and fed into the production capability model.
But forecasts are always wrong, Friend pointed out. So he started broadcasting forecasts versus actual sales to the Simplot executive committee, and the gap between forecasts and sales started to close. Result: For Simplot's fiscal 2000 first quarter, the gross error in forecast was reduced by a factor of 10.
Optimizing the supply chain requires modeling both opportunities and constraints, and solution time must match the business cycle, Chambers pointed out. BAF's architecture integrates SCT's FYGIR Advanced Planning module for modeling and optimizing the plan; SCT's FYGIR Advanced Scheduling module for refining schedules; and SCT's Adage, an ERP program with central repository for executing the plan. PlannerWeb, a tool developed in-house, coordinates data and publishes results to the organization. Models are not the real world, Chambers cautioned, but can identify optimal solutions without costly experimentation.