I think there are multiple objectives and several levels for auditing, e.g.

1. Pre-audit. Looking for low hanging improvement opportunities. These are a good path to addressing root causes of process instability. There's little value in performing a detailed sampling if the process is unstable or operating at a low performance level for obvious causes.

2. Ensure the auditing objectives are well defined. I've seen several examples of people auditing with ill-defined objectives. Is the goal to identify improvement opportunities, to develop models, etc?

Otherthoughts?

Defining the problem thoroughly is already half of the solution... With all the "big data" enabling technologies out there you could stay busy for eternity looking at "interesting trends" without actually solving your problem. However, online data which is thoroughly thought through and effectively translated from numbers to information forms an indispensable part of any plant audit.

Two possible types of audits (there are several) are:

1. Is my mill drawing the power it should?

2. Is my throughput what it should be (for a given ore hardness).

Answering these two questions require different protocols and sampling. Audit type #1 requires a mill crash stop and grind out to generate parameters for mill models. One of my papers describes such modeling for SAG mills and how the parameters are used (https://www.sagmilling.com/articles/20/view/?s=1). Audit type #2 requires grind ability testing on ore samples, mill power draws (corrected to the mill shell/pinion) and F80/P80 sizes. I don't have papers specifically addressing this type of audit, but the math for turning grind ability test results into specific energy consumption is given in this paper (https://www.sagmilling.com/articles/12/view/?s=1).

Other rigorous audit types are functional performance of ball mill circuits and hydrocyclone surveys. You can also perform multiple of these audit types simultaneously, doing Audit #1 & #2, for example.

I audit comminution systems for a living. I do 15-20 per year. Have done 150+ in the last 6 years. In fact just done 2 this week. It is surprising how often I see the same common faults. Many times people know what to do but lack the experience or confidence to implement what they believe. Let me know if I can help?

Very often, audit will start with good sampling campaign of the grinding circuit that you intend to audit, to measure actual performances vs design. I was lucky to perform some of these campaigns in SA in the early 90's under the supervision of McGill University. A lot of preliminary works, among others, to manufacture properly designed sampling devices to conduct the sampling.

In a comminution circuit always there are some important signs for a visual audit like umbrella shape for hydro cyclones or well distributed screen feed which help you to understand the performance of units before any sampling. There are lots of certain facts in comminution that are not respected for long times in some plant. These faults easily can be found out with a visual audit. What do you guys think about these sings?

Often ore hardness variability is an ongoing challenge. Therefore to be sure that you understand the actual hardness today, it is proper practice to take a feed sample from the belt after a crash stop. The plant must be operating in stable mode with no new ore added to the feed bin prior to and during the test. Sampling this stopped belt quickly by selecting 25 kg of fist sized pieces in a 20 liter pail is important. Then, using a highly accurate SAG test to measure the hardness is critical because there is only one sample. A SAG Design test is perfect for this task. The crash stop also allows the mill load to be inspected so that the optimum loading, usually 26% by volume can be confirmed. Slurry samples and plant data including power draw and tonnage processed complete the survey. Sub par performance is quickly identified and ways to improve are often very obvious.

I agree with suggestion for a good sample of the mill feed. But then, if there is one sample taken and one accurate SAG test, does this give a reliable result? I often wonder if the variability we see in sample hardness/work index etc is partially a function of the testing procedure and not wholly the ore variability. I suggest at least two samples and two tests be conducted to get a feel for the ore feed to the mill. Also make sure that the material feeding the mill during the past hour or more, which forms the basis for the mill charge, is consistent and well represented by samples taken at the crash stop.

Good point. 

There's been some good work on the underlying variability of:
* the comminution tests themselves, e.g., Bond, drop weight, etc.

* the sampling and sample preparation tests

Looking at these sources of variation, it is possible to start understanding the threshold required for a difference between test results and plant results to be meaningful.

I was recently informed by one of my friend about a new book about auditing a mineral processing plant. I hope to study it on google book.http://smemi.personifycloud.com/PersonifyEbusiness/Store/ProductDetails.aspx?productId=121502