• Keine Ergebnisse gefunden

In the following section, the application of the developed method is shown for three manufacturing systems utilizing the practical procedure described in Section 7.4.

presses and a cutting machine in multiple machine operation, with a team lead to support quality checks and setups if needed.

Step 3: Measuring material waste at the material sinks: through process observation and staff interviews the material waste forms of the considered material sinks are determined and assigned to the machine module or its peripheral areas. Measurements of material waste are recorded on the shop floor and supplemented with employee protocols. Assignment of the material waste to operating state, operating transition, activity or event provides a breakdown of the material waste costs, as shown in Table 22.

Table 22: Material waste forms for aluminum part production

Material sink Waste form Activity % Material waste cost

1. Greasing Grease Work 1%

2. Greasing Lubricants Work 4%

2. Impact extrusion Defects Work--> Error 25%

2. Impact extrusion Defects Setup--> Work 8%

2. Impact extrusion Defects Work 5%

2. Impact extrusion Lubricants Work 0%

3. Cutting Chips (fixed) Work 53%

3. Cutting Lubricants Work 4%

5. Blasting Lubricants Work 0%

5. Blasting Shot Work 0%

5. Blasting Abrasive loss Work 0%

5. Blasting Defects Work 2%

Three “biggest hitter” waste forms, highlighted in light blue in Table 22, amounting to 80% of the total non-fixed material waste cost, are selected for a detailed regression analysis of the waste amplifiers and the waste quantity per incident or waste rate per time unit. However due to company policy, some

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

% Material Waste Cost (without fixed waste forms)

2. Impact Extrusion: Work--> Error: Defects 2. Impact Extrusion: Setup--> Work: Defects 2. Impact Extrusion: Work: Defects 3. Cutting: Work: Lubricants

2. Greasing: Work: Lubricants 5. Blasting: Work: Defects 5. Blasting: Work: Lubricants 5. Blasting: Work: Abrasive Loss 2. Impact Extrusion: Work: Lubricants 5. Blasting: Work: Shot

(~ 50 parts) is discarded after startups, because a 100% visual quality test is not economical (setup work defects). For the same reason, a large bin of parts (~500 parts) is discarded after tool inserts break (work error defects).

A regression analysis is only completed for extrusion defects in work state for this reason.

Regression analysis: The waste rate (defects per minute) is calculated for

~700 production orders based on shift protocols in a 3-month period. Methods for measuring the waste amplifiers (predicting variables), are described below and summarized in Table 23.

 Part discommonality factors rooted in part geometry is determined in a process-expert workshop, here two levels of subfamilies can be distinguished: The discommonality is set to 0 if the previous part ordered on the machine is identical, 1 if in the same subfamily, and 2 otherwise. This is completed for all ~700 observations.

 Poor-machine-and product variant-fit is calculated by first determining the average defect rate for each combination and an average defect rate for each product variant overall, as described in Table 23.

 Poor machine condition is modelled as the elapsed time since the last planned maintenance activity at the order start time for simplicity.

 The production run length is based on the order start-and end times in the shift protocols.

 Employee qualification data is taken from a skills matrix. Skills scores from both work center-specific qualification and general qualifications are summed. These represent both the technical qualification and the employee’s cost and environmental consciousness.

 Temperature and air humidity are frequently blamed for high defect rates. To investigate these effects, weather data from a local weather station is used, since the observations are made in summer months in a non-climate-controlled factory hall.

Table 23: Waste amplifier measurement Waste amplifier Measurement method

Discommonality of successive product variants

{x, y} A→ Discommonality = 0

{x, y} A {x, y} B → Discommonality = 1 {x, y} A {x, y} B → Discommonality = 2 where

x : current variant

y: previous variant on the same machine A: same part (same / very similar parts) B: same sub family (similar parts) Poor fit of machine

assignment and product variant

Poor fit degree = WRMachine / WRAll where

WRMachine: average waste rate for machine – product variant combination

WRAll: average waste rate for product variant on all machines

Poor machine condition Elapsed time since maintenance = end time of last PM activity – order start time

Long run length Run length = Order end time – Order start time (hours) Low employee

qualification and cost-consciousness

Skill score of responsible operator (Company skill matrix)

Unfavorable ambient conditions

Daily high temperature Daily mean temperature Daily average air humidity

Holding time Start time production – time of raw material receipt Start time production – time of greasing

In a multiple linear regression analysis, three of the waste amplifiers demonstrated a statistically significant correlation (P-value<0.05) with the average waste rate (kg/minute) for extrusion defects in work-state. These waste amplifiers are employee qualification (inversely proportional), run length (inversely proportional), and poor machine condition, represented by time since maintenance (proportional), as shown in Figure 50. Nonlinear correlations are also investigated (e.g. employee qualification squared), but yielded higher p-values than their linear counterparts in this case. However, it is recommended to investigate both linear and nonlinear regression models in practical application.

Normalized defect rate Normalized defect rate Normalized defect rate Overall employee

qualification

Run length Discommonality degree

Figure 50: Extrusion defect waste rates a function of waste amplifiers

Step 4: Parameterization of a simulation model: Expert interviews, shift protocols, power measurements, and ERP system data provide the master data for the production model. Along with cost data from controlling, the case specific data is integrated into the input spreadsheets (see Figure 81) linked with the Vensim™ model, so that the same base simulation model can simulate multiple case studies with few parameter adjustments in Vensim™.

Step 5: Verification of base line: To ensure the model accurately represents reality, the production of four two-month time periods are reproduced in the model. A sensitivity analysis is used to investigate the influence of model parameters on the model throughput and material cost efficiency.

KPI

Simulation results (% Static calculation)

Volume 110,7 %

Throughput time 91,3%

Waste 108,4%

Figure 51: Verification of aluminum parts simulation model

Step 6: Derivation of improvement measures and scenario creation:

Using the visualization tool and the improvement measure generator (see Figure 83) four waste reduction scenarios are identified in Table 24.

y = -0,0348x + 0,0524

0,20 0,40 0,60 0,80 1,00 1,20

0,0 0,2 0,4 0,6 0,8 1,0

y = -0,1203x + 0,0771

-0,20 0,20 0,40 0,60 0,80 1,00 1,20

0,00 0,20 0,40 0,60 0,80 1,00

y = 0,0106x + 0,0331

0,20 0,40 0,60 0,80 1,00 1,20

0,0 0,2 0,4 0,6 0,8 1,0

0%

20%

40%

60%

80%

100%

120%

Volume Throughput time

Waste cost Total cost

Table 24: Simulated scenarios for aluminum parts manufacturer

Improvement scenario Addressed waste

form Parameter adjustment 1. Larger extrusion lots for

less startup loss

Startup losses (setup work), defects in work mode

+ Production order sizes

2. Shorter reaction time to tool failure at extrusion (e.g.

increased employee presence)

Defects (work error)

- Lessen waste quantity

3. Higher employee qualification

Defects in work mode

+ Min employee qualification + Mean employee qualification 4. Sequence products by

similarity (campaigns)

Defects in work mode

+ Change part sequences or next variant logic

For Scenario 1, the production order sizes are varied from their current value to investigate their effect on waste generation and overall performance in the dynamic system, as shown in Figure 52. Due to the comparatively small material loss quantity through at setups, the scenario exhibits small savings in material waste cost, while inventory levels increase dramatically and the service level suffers.

Results (% base line) Lot size Material cost

efficiency

Average serviceability

Average

throughput time Other costs

100% 100,0% 100,0% 100,0% 100,0%

150% 102,7% 91,0% 117,0% 98,0%

200% 104,9% 88,5% 144,5% 99,1%

Figure 52: Results of lot size variation (Scenario 1)

In Scenario 2, shorter reaction times lessen the waste quantity from 500 pcs to 250 pcs or 100 pcs per incident. This reduction yields material savings of 10.000€ annually (see Figure 53). It is assumed that with better timing, the current staff could catch these breakages within 2 minutes; however, the cost savings would not justify another employee in the area or an automated solution.

130.000 € 135.000 € 140.000 € 145.000 €

100% 150% 200%

Annual Material Waste Cost

Lot size at extrusion (% base line)

Results (% base line) Reaction time Material cost

efficiency

Average service-ability

Average

throughput time Other costs

100% 100,0% 100,0% 100,0% 100,0%

50% 101,8% 99,5% 99,5% 98,2%

20% 104,8% 99,7% 98,7% 99,0%

Figure 53: Shortening reaction time after tool breakage (Scenario 2)

Scenario 3 investigates the effect of increasing employee qualification in impact extrusion on material waste costs and performance. A moderate increase in qualification leads to a total material cost savings of roughly

~5.000€ annually as shown Figure 54. Assuming high employee retention, a qualification package for training for the extrusion operators would pay for itself in 1-2 years.

Results (% base line) Mean skills

index Material cost efficiency Average service-ability

Average

through-put time Other costs

0,43 100,0% 100,0% 100,0% 100,0%

0,57 101,3% 99,7% 97,1% 98,7%

0,71 101,3% 99,8% 98,0% 99,4%

Figure 54: Results of employee qualification variation (Scenario 3)

130.000 € 135.000 € 140.000 € 145.000 €

100% 50% 20%

Annual Material Waste Cost

Reaction time to tool breakage (% base line)

130.000 € 135.000 € 140.000 € 145.000 €

0,43 0,57 0,71 Annual Material Waste Cost

Mean employee qualification index in extrusion

Scenario 4 examines the effect of higher commonality of successive parts on each machine on the extrusion defect rate. Due to the larger regression coefficient, larger material savings are seen in Scenario 4 than Scenario 3, though Scenario 3 presents fewer trade-offs with market performance.

Results (% base line)) Max

dis-commonality

Material cost efficiency

Average service-ability

Average

through-put time Other costs

2 100,0% 100,0% 100,0% 100,0%

1 102,7% 93,6% 116,1% 98,0%

0 105,3% 88,6% 129,5% 99,1%

Figure 55: Product variant sequencing for less discommonality (Scenario 4)

Overall, the first case study focused on the extrusion press process due to the quantity and cost of the waste generated in the work center, particularly due to the short service times of the tooling inserts. This indicates that limiting defect rates, especially during instable periods, is the largest control lever for material efficiency for some manufacturers.

After implementing an employee-training program to increase the overall qualification of employees in the impact extrusion area, the material cost savings attained were roughly 30% greater than the simulation results.