Gains:
- Ability to explain MPS, MRP and workshop scheduling concepts at a basic level
- Ability to compare scheduling rules and scenarios (SPT, EDD, critical ratio) with AI
- Ability to check and correct capacity and constraint violations in AI output
Production planning is the backbone process that organizes "what, when, with which resources" you will produce. A good plan meets deadlines, doesn't leave machines idle and doesn't inflate stock; A bad plan leads to chaos, overtime and late deliveries. AI is powerful in this area at comparing scheduling rules, generating scenarios, and listing constraints; However, it can also overlook capacity and priority constraints and produce plans that are "nice on paper but unimplementable". In this unit, we will cover MPS/MRP concepts, scheduling rules, and how to validate AI output.
Planning Hierarchy: MPS and MRP
Production planning is a hierarchy from general to specific.
Level
What does
time horizon
S&OP / Bulk plan
Roughly establishes the demand-capacity balance
months
MPS (Master Production Schedule)
Divides the finished product into periods
weeks
MRP (Material Requirements Planning)
Removes component requirements from the BOM
weeks/days
Workshop scheduling
Places jobs on machines/clocks
Days/hours
MRP logic is essentially simple: Finished product demand (MPS) is taken, exploded with the bill of materials (BOM), on-hand stock and open orders are deducted, and the net requirement is shifted back by the lead time (offsetting). AI is good at explaining this logic and setting up example tables.
Role: You are an industrial engineer who specializes in production planning. Task: Set up a simple MRP table for the following product. Show rows for gross requirement, planned delivery, on-hand stock, net requirement and planned order release weekly.Product: A (gross requirement weekly: {{ values }})Starting stock: {{ x }}, Lead/production time: {{ y weeks }}, Lot size: {{ lot }}Rule: Briefly explain how each cell is calculated, apply offset correctly. Producing fabricated value.
Tip: The most common error in MRP output is incorrect lead time offset. Check line by line that the AI's planned order release line is before the exact lead time of the net requirement.
Scheduling Rules (Dispatching Rules)
The order in which you handle waiting jobs on a machine directly affects the average latency and flow time. The most common priority rules:
- FCFS (First come, first served): Fair but not performance oriented.
- SPT (Shortest processing time): Reduces average streaming time and average latency; Postpones long tasks.
- EDD (Earliest delivery date): It is good at reducing maximum delay.
- Critical Rate (CR): (Remaining time / remaining work time); dynamically balances delivery pressure.
A small example. Four jobs, one machine, processing times and delivery dates:
Business | Processing time | DeliveryA | 4 | 6B | 2 | 4C | 6 | 8D | 1 | 3SPT order (by time): D(1), B(2), A(4), C(6)Completion times: D=1, B=3, A=7, C=13Sum of flow times = 1 + 3 + 7 + 13 = 24 → average = 6.0EDD order (by delivery): D(3), B(4), A(6), C(8)Completion times: D=1, B=3, A=7, C=13Delays: D:0, B:0, A:1, C:5 → maximum delay = 5
Different rules improve different purpose: SPT average flow, EDD delay peak value. "Which is correct?" The answer to the question depends on the purpose, and just because the AI says "this is best" does not make that decision.
Weak Prompt / Strong Prompt
Weak prompt:
Sort these jobs in the best possible way.
"Best" is undefined; AI gives a random order without knowing which purpose it is optimizing for.
Powerful prompt:
Sort these jobs using both SPT and EDD rules. Calculate completion times, average flow time and maximum delay for each row and make a comparison table. My goal is to "minimize missing deadlines". Tell me which rule is more suitable for this purpose, with reasons.Works: {{ list }}. Show account steps.
The second prompt clarifies the intent, compares multiple rules, and makes the account visible; Without these three, chart decision is unreliable.
Capacity and Constraint Control
If a plan exceeds capacity, it is not valid. Be sure to check the AI-generated schedule against the following constraints: machine capacity (hours), shift limits, setup times, precedence relationships (one job must finish before another), and material availability.
Check the following chart for constraint violations:- Does the total load for the week exceed the capacity (40 hours/week) on each machine?- Are setup times included?- Is the priority constraint (job B must finish before A) maintained?Chart: {{ ... }}. If you find violations, list them in a tabular form and suggest corrections.
Caution: AI often “forgets” the capacity constraint and loads a machine with 45 hours of work. Before making a production decision, collect the weekly total load of each machine manually or with a table and compare it with the capacity.
Mini Case: The Bottleneck Machine and the Wrong Rule
In a metalworking shop, deliveries are constantly delayed. Industrial engineer Cansu sees that jobs are sorted with FCFS. It gives the AI the current backlog, processing times, and deadlines and asks it to compare SPT, EDD, and critical ratio. AI calculates metrics for all three rules: SPT is best at average flow, but a few big jobs are too late; EDD significantly reduces maximum latency. Cansu switches to EDD on the bottleneck bench. But the first schedule the AI suggested violated the priority relationship of the two jobs (assembly requires body first); Cansu catches this at the control stage and fixes it. The result: delays are decreasing. AI made the rule comparison; It was the human audit that caught the constraint violation.
Common Mistakes
- Charting without defining the goal: Saying "best"; whereas SPT and EDD improve different purposes.
- Not controlling capacity: Thinking that AI's overcapacity plan is feasible.
- Ignoring setup times: Neglecting setup times and coming up with a plan that doesn't actually work.
- Bypassing priority constraints: Issuing mandatory priorities, such as assembly order, without checking them.
- Fixation on one rule: Using the same rule in every situation; When the bottleneck and delivery pressure change, the rule should also change.
In summary
- Production planning is a hierarchy from general to specific: S&OP → MPS → MRP → shop scheduling.
- The most critical mistake in MRP is lead time shifting; Check the AI output line by line.
- Scheduling rules (SPT, EDD, CR) improve different purposes; The choice of rules depends on the purpose.
- Check every schedule the AI produces against capacity, setup, and priority constraints.
- AI generates comparison and scenario; It is the human who detects the constraint violation and makes the decision.
Application task
Construct a workshop scheduling problem for at least 5 jobs (with processing times and deadlines). Have the AI sort these jobs with both the SPT and EDD rule; Calculate the average flow time and maximum delay for each queue. Critical validation: manually calculate completion times and metrics for at least one rule and compare with AI. Then add a priority constraint (e.g. "job C must finish before job A") and check whether the AI's schedule violates this constraint. Finally, justify in one sentence which rule you will choose based on your purpose (average flow or peak delay).