Sreedher Kadambi and Badri Narayana, Flame TAO Knoware Pvt. Ltd.
Our previous blog elaborated on the ‘defining’ stage of the Six Sigma process in our associated apparel company. Here, we’ll be continuing the explore the DMAIC project which we implemented. Specifically, let’s look into the ‘M’ of DMAIC – measuring the current state!
MEASURE: The Current State
The Quality Head stated that while inspection took place daily, the results were not documented. Therefore, we had to develop a data collection plan immediately. Firstly, we asked the team to brainstorm and define the defects that can be found in the output of an embroidery process. Secondly, we asked them to make a swatch of the defect and visually display them in the section. This could act as a good visual training for new inspectors on the job.
Furthermore, the brainstorming process while measuring the current state uncovered another detail! This was that defects can also be contributed by bad input to the process. Therefore, the team decided to compliment the display board with ‘unacceptable’ inputs (cut panels) for the embroidery process, as shown in Figure 1.
Additionally, the people who engage in preparatory processes were trained on what to accept and reject as an input.
Figure 1 – Output and Input Defects displayed in the Embroidery Section
Check Sheets
Firstly, we needed to explore a tool called a check sheet. It’s one of the 7 ancient quality-check tools applied while collecting data. Its constituents depend on what one wants to know about. Specifically in this case, we were interested in the following:
- Firstly, the date on which data was collected.
- Secondly, the name of the inspector who carried out the inspection.
- Thirdly, the total number of units
- Fourthly, the number of rejected pieces from the total unit.
- Lastly, the reason for rejection.
Table 1 (below) shows the design of the check sheet used for data collection in this case:
Table 1 – Data collection check sheet
Stratification
Secondly, stratification is another QC tool that goes in hand with check sheets. These are the factors which experimenters are interested in from an analytical stand point. For example, stratification refers to shifts, machines, line numbers, and operators to which data belongs to.
Moving ahead, we decided to collect data for at least a week before the team met again to discuss the next steps. Afterwards, we plotted a control chart to baseline the current status. This chart is used to monitor a specific process. It suggests whether the process is in control or not. In other words, it monitors its stability. A control chart is characterized by control limits, which are of 4 types. Firstly, there’s the UCL, which is the upper control limit. Next, there’s the LCL – i.e., the lower control limit. Lastly, we’ve got the average and the trend line. Figure 3 shows a skeletal structure of a control chart. Subsequently, based on the quest and type of data, control charts are selected. In this figure, the x axis represents the subgroup numbers collected in order of time, and the y axis represents the measurements against.
Figure 2 – Elements of a Control Chart
Based on the type of data at hand, there are different types of control charts:
- Control charts for continuous data – (i.e., Average and Range, Average and Standard Deviation, Individual and Moving Range) charts can be used. For instance, you could use charts in monitoring critical measurements.
- Control charts for discrete data – P, NP, C & U (i.e., Fraction defective, Number defective, Number of defects, and Number of defects per unit) can be used. Keep in mind that individuals control charts are often used when there is only one observation per subgroup or time interval. For example, use ‘U’ chart to monitor incoming fabric quality, use ‘P’ chart to monitor sewing line performance, and so on.
These control limits are also called the ‘voice of the process’. However, these are different from specification limits – which is the voice of the customer. Control limits outline the current process capability. The figure below shows the ‘individual’s’ control chart that we plotted using the collected data.
Figure 3 – Individual’s control chart used to measure the current state
This control chart shows us the following things:
- The process is in the state of statistical control and stability – Here, we used a statistical software called Minitab to plot the chart. Consequently, it runs 8 tests to check the process stability, as shown in Figure 4. While the default values of ‘k’ are given, they can be changed if required. If one or more of these tests fail, it leads to the process being out of control – instances of which are highlighted in the graph.
Now, what does we mean when a process is in a state of statistical control? To understand this, you need to focus on 2 broad causes of variation:
- Chance causes: This gives rise to variations in the product, despite everyone consistently operating the process as per universal standards. Additionally, these causes aren’t under technical control. However, note that they are usually unavoidable. Furthermore, the variation that they produce is called ‘controlled variability’.
- Assignable Causes: These produce abnormality, which leads to large variations within products – known as ‘uncontrolled variability’. For instance, they may take place when something beyond or against work standards occurs. Nevertheless, these can be eliminated if everybody focuses on improving issues related to man, machine, methods etc. However, unlike chance causes, these are avoidable in nature. Note that the perfection of work standards under the presence of uncontrolled variability would mean that workers were not performing as per instructions.
Therefore, a process in ‘statistical control’ refers to it being influenced by chance causes of variation. In this case, everyone in the embroidery section was following the instructions and SOPs as per the guidelines.
Figure 4 – Tests performed by Minitab to check if process is in control or not
- The process is capable of producing more than 95% quality – This is clearly evident in the control chart (figure 3), where we observed that the LCL is 96.96% – i.e., the worst quality that the process will produce.
The endgame is to facilitate a stable process that meets a target. Going by this, the process seemed to be fairing well! We shared these views with the team and suggested that there was no need to carry out a project. However, the department head said, “What we have done so far is interesting. But why not aim for 100% quality?”
This propelled us to continue the journey. However, we now had a revised target of achieving 100% quality! Hence, when quality accounts for everything, why settle for anything less that 100%? That’s our motto.