Sponsors

 

As information solutions, materials requirements planning (MRP), manufacturing resource planning (MRP II), and enterprise resource planning (ERP) evolved to become the mother-of-all-manufacturing-databases. But as production management solutions, they left much to be desired. They were inherently unable to provide an accurate and current view of what was really happening on the plant floor.

So, how do you get your plant more responsive to such events? Use advanced planning and scheduling (APS). APS attempts to dynamically synchronize customer demand to the very real constraints on the production floor—typically inventory, machine capacity, and people—against a backdrop of business and production rules. Because of changes in technology—faster computers, more computer memory, and efficient scheduling algorithms—APS can synchronize these demands and constraints in a flash, using real-time customer and manufacturing data.

What comes out are plans that show what materials and resources will be used in the future, and schedules that execute customer orders by detailing who should make what, when to start and finish making the items ordered, and where and how to make those items (such as sequencing and batch lot).

With these, management now has a tool to analyze the affect of production's responsiveness on profitability, customer service, and the general competitive stance of the enterprise.

A Short History of Manufacturing Scheduling

Plant scheduling used to be some variation of order-point scheduling. Take an order's due date and work your way back through the bills of materials (BOM), subtracting the times associated with producing that order, including material delivery, production, and shipping. After all the subtractions, you wound up with a start date for that customer order. This process, called "backward scheduling," is an approach based on averages; it doesn't consider the daily fluctuations and fire drills on the factory floor.

Then came MRP in the early 1960s. MRP used backward scheduling to highlight material shortages and then generate production and purchase orders to avoid those shortages. Through the wonders of computerization, MRP automated a lot of the drudge work associated with material requisition.

MRP II added a layer on top of MRP called the master production schedule (MPS). MPS marked the end of ordering inventory based on past usage. Instead, MPS focused on sales and marketing's best guess of the future need for products. This best guess was then passed to the next planning function, namely, the next MRP run.

Both MRP and MRP II assume certain ideal characteristics about the imperfect world of production and the plant floor:

• Infinite resources (machine capacity and labor) are always available and never change on a weekly, let alone daily, basis. MRP and MRP II (and even ERP systems) typically let schedulers plan any job; the ones that can't be done become past due or violations of lead time.
• Material resources will arrive as scheduled in the right quantities. Any variances, or missed incoming shipments, were expedited manually until the next MRP run.
• Customer orders and products have the same priority. MRP can't differentiate customer orders from orders for safety stock and forecasted orders. Generally, MRP just aggregates demand (customer orders) into lots and outputs a bunch of numbers that essentially say "make all of these items."
• Lead times (production and material delivery) are fixed or proportional to lot size.
• Weekly buckets are good enough for scheduling purposes. The fact is, an MRP run took all weekend, so all a production planner could expect was a weekly schedule that was immediately out of date once printed.

Capacity requirements planning (CRP) tried to correct this situation by at least identifying under-utilization and overload conditions at a machine or work cell. CRP was a useful reporting tool, but it didn't provide the ability to fully model production and all its constraints.

The real problem was that the entire computerized scheduling process was flawed. The assumption was that production scheduling was a top-down, sequential process—first materials, then workcenter capacities. It doesn't work that way. Explains Gary Barton, partner at Deloitte Touche Consulting Group (San Francisco, CA), the master production schedule (MPS) assumes static requirements and it schedules existing capacity. "But, it's not enough for marketing and sales to give us a plan of demand once a month and then for us to cycle through the traditional distribution requirements planning (DRP), MPS, MRP, and CRP processes each week. By the middle of the week, the requirements plan that we put in place is usually busted. Our ability to predict is only so good."

Thus the need to constantly update the demand plan going to the manufacturing and distribution organizations. This would ensure that these organizations remain in lock-step with the "current" forecast of demand. However, with an MRP run typically taking an entire weekend to complete, scheduling was a one-shot, batch affair that did not loop back to reconsider different options and multiple constraints.

Finite capacity scheduling (FCS) systems in the 1980s made the scheduling problem a mathematical sequencing problem that tried to ensure against having more than one job vie for the same capacity resource. The result was a combinatorial problem involving multiple orders, multiple sequential and often interdependent manufacturing operations, and multiple production and business constraints.

FCS emphasized maximizing production resources. Over-capacity at a workcenter was a no-no. If capacity exceeded 100% for a predetermined time, the system could automatically extend some production dates or reprioritize the production order based on priorities associated with the customer or customer order. Or, the planning logic could evaluate the factors leading to a constraint or the confluence of constraints, generating a set of recommendations to rebalance those factors toward an optimized schedule.

Around this same time, Optimized Production Technique (OPT), developed by Eli Goldratt and others, introduced the concept of constraints as bottlenecks on the factory floor. OPT attempted to identify only production bottlenecks, and then eliminate them through proper scheduling.

In the early 1990s, a separate planning system, generically called "Fast MRP," became available. Fast MRP loaded all of the scheduling logic and the entire MRP database—customer orders, materials, workstation capacities, and labor resources—into the computer's memory. There, data processing ran fast—in minutes—enabling users to run several material requirements and production scenarios. Scheduling changes resulting from the best scenario would then be uploaded and integrated into the MRP II database. Some MRP II vendors provided interfaces to "embed" Fast MRP into their enterprise-wide system, even though the Fast MRP still ran as a separate batch planning system.

Eventually, rules-based logic was introduced to let users capture their business and production strategies and priorities as algorithms. Couple these rules with a true model of their factory floor and a finite capacity view of production, and you have APS. Now users could answer a question such as, how would 97% on-time performance save the company money?

APS in Operation

APS systems are appropriate to all suppliers in the automotive industry, regardless of tier. An i2 Technologies (Irving, TX) product backgrounder explains it best:

• High- and low-volume assembly: APS coordinates material purchase, arrival, and flow better than MRP alone, overcoming infinite capacity and fixed lead time assumptions of MRP while accommodating material and capacity constraints.sequence-dependent setups and lot sizes.
• Batch process: APS addresses the trade-offs between long and sequence-dependent setups and lot sizes.
• Job shop: APS manages resource conflicts where complex routings and tight due dates exist.
• Make-to-order: APS supports available-to-promise by estimating order completion dates and flagging potential delays in existing orders.
• Build-to-stock: APS significantly reduces finished goods inventory by reducing production cycle time and, more than other scheduling techniques, by reacting faster to market changes and unforeseen manufacturing fluctuations (such as poor quality or late arrival of raw materials, machine breakdowns, and sick days).

Particularly where manufacturing enterprises compete on their abilities to meet available-to-promise, APS is king. For example, says D'arcy Elliott, vice president of Sales and Marketing for ShivaSoft Inc. (Edmonton, Alberta, Canada), APS can perform front-end sequencing of auto assembly orders in preparation for line assembly—supplier environments where color constraints and changeovers, model year changes, and just-in-time operations are critical.

Regardless of the manufacturing environment, APS requires lots of data: constraints, behavior rules for when a given constraint may be altered, priorities, and strategic corporate objectives. Both "static" and "dynamic" data are required. Static data are typically build data, including BOMs, process routings, cycle times, setup penalties, and inventory levels. These data are automatically updated periodically through file transfer from other systems. Dynamic data are the daily demand data resulting from sales, including customer orders, scheduled shop orders, work orders, production runs, and completion dates. Dynamic data is updated every time the APS runs. Other data required include:

• Parts routings and process descriptions
• Manufacturing production rates
• Workcenters, including machine identification numbers and the number of shifts the machines work
• Production rules, such as for machine state changeover, set-up optimization, lot splitting and grouping, order prioritization, and dynamic in-process materials buffering
• Dynamic information, including the MRP forecast, customer orders, and shop floor status

Add to these data the information about the constraints in your enterprise, such as:

• People, including skills, job classifications, qualifications, and work schedules
• Equipment and associated tooling, their setup and maintenance requirements, operating capacities, maintenance, and setup and calibration specifications
• Inventory, including raw materials, work-in-process, and finished product availability and sequencing
• Transportation, including warehousing and logistical requirements, shipping rates and discounts, and availabilities
• Business rules, including customer service requirements, environmental considerations, management reporting structure, supplier relations, and product year changeovers
• Financials, including cash-on-hand, rate of cash flow, profit margins by product or customer, and return-on-investment.

All of these data then run through one of five categories of constraint-based planning engines:

• Heuristic engines rely on forward and backward scheduling to balance orders against resources
• Theory of constraints engines identify resource bottlenecks and then prioritize activities to minimize the bottlenecks
• Simulation engines model the shop floor, process local heuristic scheduling rules, and then generate a statistical output about the flow of work through the various work centers on the shop floor
• Knowledge-based engines use explicit rules about customer demand, work flow, resources, and constraints to balance incoming customer orders against desired delivery dates
• Optimizer engines use mathematical programming and branching techniques to minimize scheduling and resource conflicts in meeting individual customer orders.

In operation, APS systems can qualify the requested delivery date in a customer order, such as by best lead time or exact date. It can then determine which set of rules are needed to process individual orders and calculate the capacity needed in each production work cell for that order. The APS can then match that capacity against a database of manufacturing constraints and available capacity by work cell. The output would be production orders that initiate jobs in time to meet the requested delivery date.

Moreover, APS systems reschedule material and other resource capacities simultaneously, not sequentially. Also, they can perform both backward and forward scheduling, which better synchronizes upstream and downstream resources as changes in demand, availability, and capacity are made by the APS to those resources. However, forward scheduling introduces the problem of prioritizing resources against the backdrop of both demand and business strategy; hence the need for a rules-based approach to scheduling, which can capture and apply those priorities as appropriate.

The balancing act in conflicting production goals. (Source: Baan Company)

In addition to detailed scheduling, APS features decision support including extensive modeling, workflow simulations, mimic diagrams, and "wizards" to prompt the user through "what-if" scenarios. Graphical user interfaces provide real-time, color-coded views of order status and production data to visualize timeliness, identify bottlenecks, and analyze trends. And there are all the reporting tools for both standard and ad-hoc inquiries.

All of us have seen the cartoon of the employee doubled over in laughter saying, "You want it when?" With APS systems, your colleagues can not only answer that question, but deliver as promised in their answer.