Partial Automation

There does seem to be another question begging. So many of the carrots in the two tails of the distribution are in the trouble areas (clogging or wrong-end chopping) that it appears that the automated in-feeder is not doing a satisfactory job. In the real processing plant from which Case study 3.5 was drawn, this was precisely the case (except that the vegetable was okra, not carrots). The reliability of the product forced the process engineers to compromise and use a partially automated work station. Space for human operators on both sides of the shaker tables was provided so that vegetables that are seen to fall incorrectly could be turned over manually before entering the chopper area.

Partial automation is not limited to food processes where product variability is the chief problem. Orientation and alignment of various kinds of products and machine parts may require intelligent vision or some sophisticated capabilities which in a given manufacturing process maybe infeasible either technically or economically. Partial automation, where the worker and machine each do their parts according to their capabilities is an alternative that should not be overlooked by the automation engineer. This website has considered the problem of moving the material or piece parts position for processing or assembly. Although the operations have been seen to be complicated and some even ingenious but none has resulted in useful change to the process itself. Parts feeding and orientation was without doubt important but we are now ready to consider the automation of machines that process or assemble these parts into useable products.

The success of an assembly automation project depends not so much on the achievable motions of industrial robots handling product parts as it does on the tiny subtleties of positioning and orienting of piece parts for assembly.  Sometimes these detailed problems can be avoided by preserving parts orientations to begin with, either during their manufacturing or at the vendor’s site for purchased parts. Lab automation news has an example video and case study of a partial automation device.

Given that parts orientation must be accomplished, clever mechanical devices are required for this purpose with the principal representative being the vibratory bowl with parts selectors in the discharge track. Those parts that the bowl system is unable to select and orientate correctly are rejected back into the bowl for a retry. The rejected parts are not lost; generally, the only things lost are time and production rate. The amount of decrease in production rate from rejection of mis-oriented parts is related directly to the efficiency of the vibratory bowl and part selector track. The overall efficiency of the system can be quantified from the matrix of individual rejecter probabilities along the selector track. High production rates are necessary, vibration rate can be sped up with corresponding higher feed rates up to a point. The science of orientation and feeding applies not only to small metal parts, but also to plastic bottles, moldings, castings, and even food products. Food products present special problems for automation engineers due to the variation in item sizes and shapes.

Food Product Orientation

In the canning industry, the sorting, orienting, and feeding of food products can present more complex problems of analysis due to the random variables, which cannot be avoided. Automation research by Kwei at the University of Arkansas has led to the development of an optimum in-feeder design for the chopping of vegetables in a frozen food process (see Figure 3.18). At the point where the carrot was attached to the top of the plant is a tough portion that is considered undesirable. Mechanical choppers are capable of chopping off the tops if the proper end is fed to the chopper. The picture shows an in-feed sorting and conveying mechanism, the objective of which is to advance the carrot towards the mechanical choppers. The carrots are dumped from bulk conveyors onto a “shaker table,” which is full of holes. The top end is heavier and will likely fall through the holes first regardless of carrot orientation on the shaker table. What happens after the carrots fall depends principally upon the relationship between carrot length and the size of the gap between the in-feed conveyor and the shaker table. Other factors are the speed of the in-feed conveyor and the downward slant of the entire assembly. If the carrots could be counted upon to be of constant length, the problem would be much easier. Analysis of the feed system design must consider the random variables of carrot size. Case Study 3.5 illustrates the computations for a simplified case.

Design of a Frozen Vegetable In-Feeder

Figure 3.18 shows part of an automated system for preparing frozen carrots. In the diagram, a carrot is falling through a hole in the shaker pan onto the in-feed conveyor with a correct orientation. However, if the carrot is too short for the gap between shaker pan and in-feed conveyor, it is clear that the carrot might fall such that the point of the carrot will be forward and the cap to the wrong end. On the other hand, if the gap is too small, the carrot will not feed and will clog the system. Clogging causes disruption of the production line and is considered twice as serious a difficulty as “wrong end” orientation. As an approximation to the true situation, assume that clogging always will occur if the gap is one inch less than carrot length and that clogging will never occur if the gap is larger than one inch less than carrot length. Also, as an approximation, assume that a carrot will always fall incorrectly if the gap is greater than the carrot length. Assume that carrot length is normally distributed with mean 3 in. and standard deviation of 2 in.

1. What size gap should be specified?
2. What percent of the incoming carrots would clog the in-feeder?
3. What percent of the carrots would be chopped incorrectly?

Solution

Let L length of carrot (normally distributed, t = 3 in., u = ‘/ in.)

F[L] = cumulative distribution function of L

G = gap in inches — optimum gap

Q(G) = penalty function = Prob (carrot will be chopped wrong end] + 2 Prob [carrot will clog]

Q(G*) = min Q(G)

Q(G) Prob[G  L] + 2(Prob[G  L — 1])

The function can be minimized either by standard search techniques or differential calculus. Intuitively, the gap should be greater than 2 in. because clogging is a more serious problem than wrong end chopping. The minimum for the penalty function is found to lie in the vicinity of a 2.7 in. gap.

The carrot processing case study required some understanding of random variables and statistical theory for analysis but even for those readers without a background in these techniques, the similarities between such a problem and machine parts feeding problems should be evident. Just as in the slot feeding problem there is a working range to be established. For a nominal 3-in, carrot the in-feed conveyor can be set anywhere from 2 in. to 3 in. with no problem. The complication is that carrots vary in size, and a working range for a 3-in, carrot is not a working range for a 5-in, carrot. Fortunately, most real world statistical distributions have a central tendency, and automation system designs can be aimed for the bulk in the mid-range of the distribution. This was done in Case study 3.5, except that the upper tail of the distribution was favored somewhat over the lower tail because of the higher penalty associated with clogging the in feeder.

Part Wear and Damage

A final consideration may be that the parts selector will overwork the parts oscillating and vibrating them and kicking many of them back for retry after retry. If the effectiveness is virtually 100 percent, it is possible to use the system efficiency to calculate the chances that a part will be tossed back k times before reaching an acceptable orientation. Thus, nearly one out of a hundred parts will be kicked back ten times before achieving an acceptable orientation. The automation engineer must decide whether or not this kind of treatment will damage the product. The average number of kickbacks for every pad entering the 100 percent effective selector system is also a function of efficiency.

Stainless steel stampings are used to jacket and support pulleys used in the assembly of blocks of various designs used on sailboats. The stainless steel has a polished appearance and the quality of the surface finish is a factor in the acceptability of this costly product. Suppose these stamping are oriented for subsequent assembly using a bowl feeder having an effectiveness of 100 percent and an efficiency of 50 percent. What are the chances that a given stamping will be tossed back into the bowl three times before a correct orientation will present it for assembly? What is the average number of times a typical stamping will be tossed back into the bowl for this operation?

Having completed a discussion of the rather precise ways in which a parts selector can be analyzed, it should be knowledge that the mechanization of parts handling and orientation is as much an art as it is a science. The fabrication of selector mechanisms and escape devices is usually done in a very experimental way in which the skilled artisan cuts and tries many different configurations until an effective design is perfected. Although many scientific principles govern the behavior of parts orientation and selection processes trial and error experimentation can be used to develop practical working systems without the science. The primary benefit of the analysis is to understand the dependence between successive steps in the orientation process and the importance of efficiency and effectiveness to the overall success of the automation project.