Glass breakage is commonly acknowledged as a major nuisance in parenteral manufacturing. Depending on when it occurs in the glass lifecycle, it causes a plethora of problems, such as disruption of production flow, particle contamination and loss of sterility, to name a few.
Even though glass is a material known since ancient times, it still puzzles people, giving rise to erroneous theories and stances—everybody has an opinion on glass breakage!
Contrary to popular belief, glass does not accumulate stress. Stress arises in the glass when it is subjected to external loads and disappears again immediately when the external load is removed. Glass breakage always requires both the presence of a flaw and tensile stress.
And glass strength is determined by the properties of its flaws only—it is not a material constant. Flaws in glass are intrinsic and inherent from the manufacturing process as well as extrinsic—acquired during the entire lifecycle—primarily due to glass-to-glass contact. In other words, glass is weakened throughout the manufacturing processes due to infliction of new flaws. If a glass container is subjected to a load outside the normal load range, or the container is significantly weakened by a defect, it will break and a root cause can be assigned, and is, therefore, “special cause” and the outcome of a deterministic model.
If the load is within the normal load range and the container strength is within the Weibull Modulus (a parameter used to measure variation in the strength of brittle materials), breakage can appear, where there is an overlap between the two (see Figure 1), and therefore “normal variation breakage” and the outcome of a probabilistic model.
Figure 1 Overlap of Process Load and Weibull Modulus Creating Normal Variation Breakage
It helps to characterize and rank all the manufacturing processes one-by-one on how much each contributes to the risk of glass breakage. Figures 2–3 below illustrate how Statistical Analysis Software (SAS) can calculate the differences between the two populations (before the process and after the process). The example is generic and can be applied to all glass breakage events.
A Review of Deming
First, understanding glass breakage requires a review of Deming, who said that the key to knowing variation lies in understanding that everything that is measured includes normal variation from a system’s flexibility as well as special causes that result in that result in defects. Achieving quality requires eliminating special causes while controlling normal variation. Making changes in response to normal variations just makes the system worse (1).
Glass breakage events can, at the outset, be considered as quality variations—a normal variationsuperposed with a special cause variation. In order to control variations in a given process, the nature of the variations must be known. See Figures 1 and 2.
The ratio between normal and special cause breakage in any process depends on the load characteristics applied by the process and the glass strength variation (Weibull Modulus). If the process loads are significantly below the glass strength, only special cause breakages occur.
Figure 2 Aggregated Number of Breakages
Establishing a Shewhart process control chart on normal variation breakage is beneficial for the individual process steps as well as for the entire process. The Shewhart control chart can be used for monitoring the processes for verification of process improvements. See Figure 3.
Figure 3 Normal Variation Breakages Shewhart Control Chart
Glass breakage can be described mathematically as a mix of deterministic and probabilistic models. In a deterministic model, it is always possible to predict the output of a process when the inputs are given. In a probabilistic model, the output of a process is always associated with probability or risk, also when the inputs are given.
As glass acquires more flaws during the entire lifecycle, the strength is irreversibly reduced and the probability of breakage increases.
Glass fractography is an essential tool in any glass breakage reduction campaign, as it enables the investigator to distinguish between normal variation breakages and special cause breakages. Glass breakage can further be divided into two subcategories: breakage caused by manual handling and breakage caused by automated process handling. Manual handling breakage can be addressed via behavioral programs (do’s and don’ts) and controlled by “go-look-see” process confirmations. Glass breakage caused by automated processes is handled differently, depending on the nature of the breakage.
Special cause breakages are addressed with commonly known methodologies like Systematic Problem Solving (SPS).
Normal variation breakages can only be reduced by:
- Increasing the glass strength (and durability)
- Reducing the loads to the glass
- A combination of these two
Reducing the loads to the glass to reduce the normal variation breakage requires that the culprit be known. A simple approach can be counting the piles of broken glass along the process flow; however, most of these breakages are likely to be special cause breakages or “delayed” breakage of glass that is weakened in upstream processes by infliction of flaws, making it too weak to withstand loads within the normal range.
A better approach is to measure the relative strength reduction of the glass population in the different subprocess steps. By ranking the relative strength reduction in all subprocesses, a prioritized list of improvements can be established.
Consider the following methodology for an automated process variation. First, determine the most common breakage morphology for the normal variation breakages by fractography; second, establish a glass breakage method and use equipment that emulates the most common normal variation breakage morphology with a quantifiable, calibrated output of the breakage strength. Third, sample immediately before and immediately after pertinent, critical process steps; then, establish the Weibull Modulus for the two samples. And finally, compare the two Weibull Moduli using statistics. Here is an example of such a methodology.
- Sample size 150 items
- 10% quantile
- Nonparametric quantile regression tests (Wald and Likelihood Ratio)
- Estimated difference between 10% quantiles with 95% confidence limits
“Old” filling line infeed
- 12% strength reduction of 10% quantile
- Wald Test: H0 : Ref0.1= Scroll0.1, p < 0.0082
- Likelihood Ratio: H0 : Ref0.1= Scroll0.1, p < 0.0007
- Scroll0.1 – Ref0.1 = -8.93, [-15.581;-2.278]
- There is a significant difference
New, improved filling line infeed
- No strength reduction of 10% quantile
- Wald Test: H0 : Ref0.1= Scroll0.1, p < 0.8265
- Likelihood Ratio: H0 : Ref0.1= Scroll0.1, p < 0.7195
- Scroll0.1 – Ref0.1 = -0.320, [-3.192;2.552]
- No significant difference
As normal variation breakage is probabilistic by nature; it is not possible to predict a quantifiable effect of improvements in the process. If a Shewhart chart is established for the process as recommended, however, the effect of the improvement can be measured as a downward shift in average value or a reduction of standard deviation.
This approach to glass breakage offers a systematic and practical approach to an often misunderstood problem. Glass breakage is a challenge that will always be present, so using an effective tool to manage process variations is key.
Note: This article originally appeared in the PDA Letter, a publication produced by the Parenteral Drug Association. Jan 17, 2017
- Deming, W.E. The New Economics for Industry, Government, Education. Cambridge, MA: Massachusetts Institure of Technology, Center for Advanced Engineering Study, 1994.
Continue at: https://www.pharmaceuticalonline.com/doc/a-new-process-for-reducing-glass-breakage-0001?vm_tId=2060316&user=e9e14da2-2932-478a-b692-b12fe9b1d5e5&utm_source=et_6214180&utm_medium=email&utm_campaign=PHARM_04-05-2018&utm_term=e9e14da2-2932-478a-b692-b12fe9b1d5e5&utm_content=A+New+Process+For+Reducing+Glass+Breakage
The text above is owned by the site above referred.
Here is only a small part of the article, for more please follow the link